diff --git a/agent/orchestrator/SKILL.md b/agent/orchestrator/SKILL.md index 1b68bdb22..fbcc2fbf4 100644 --- a/agent/orchestrator/SKILL.md +++ b/agent/orchestrator/SKILL.md @@ -10,8 +10,8 @@ description: Create and coordinate native agent sessions (Codex, Claude, Qwen, K A session started from CanvasTTY's launcher with the **Orchestrator** role has the `canvastty_agents` MCP tools. Use them in this order: 1. `list_providers`: which agents CanvasTTY can launch here. Each entry's `id` is the exact `spawn_agent.provider` value; prefer entries that are `available` with `signIn: "ok"` (`unknown` only means CanvasTTY has not checked yet). When plugins offer launch options (for example an accounts plugin's `list_routes` tool), the answer names that tool. -2. `spawn_agent` once per part of the task: a known `provider` id, an absolute `cwd` and a self-contained prompt. An unknown provider is refused with a pointer back to `list_providers`. **If the person names a model, pass it as `model`** in the format `list_providers` gives for that provider (`model.format`; OpenCode takes `provider/model`, for example `zai-coding-plan/glm-5.3-flash`, and `model.known` lists what `opencode models` reported); without `model` the CLI uses its own default. `effort` sets the reasoning effort where the CLI has one (`efforts`). Without `profile` the subagent gets your own launch profile, or the next lower one its CLI has (never more than yours, never YOLO); `list_providers` shows each provider's `profiles`. `cwd` must be this project's folder or a folder inside it. A refusal says why (a folder outside the project, a profile above yours, a limit the person set): change the request, do not retry it. -3. `wait_for_agent` for each subagent (`timeoutSeconds` up to 600; call it again after `timeout`) instead of polling. `needs_approval` means the person must answer a prompt in that card: never answer it yourself. +2. `spawn_agent` once per part of the task: a known `provider` id, an absolute `cwd` and a self-contained prompt. An unknown provider is refused with a pointer back to `list_providers`. **If the person names a model, pass it as `model`** in the format `list_providers` gives for that provider (`model.format`; OpenCode takes `provider/model`, for example `zai-coding-plan/glm-5.3-flash`, and `model.known` lists what `opencode models` reported); without `model` a subagent of your own CLI keeps your plugin launch options (for example your model account), and any other CLI uses its own default. A plugin you do not name in `launchOptions` keeps your own values; passing `model` next to inherited options that already set the model is refused, and a subagent of another CLI is refused while your model comes from a plugin's options (spawn one of your own CLI instead). The answer's `servedBy` names the CLI, the model asked for, and whether each plugin's options were `inherited` or `explicit`. `effort` sets the reasoning effort where the CLI has one (`efforts`). Without `profile` the subagent gets your own launch profile, or the next lower one its CLI has (never more than yours, never YOLO); `list_providers` shows each provider's `profiles`. `cwd` must be this project's folder or a folder inside it. A refusal says why (a folder outside the project, a profile above yours, a limit the person set): change the request, do not retry it. +3. `wait_for_agent` for each subagent (`timeoutSeconds` up to 100; a call returns within 50 seconds unless you are Claude Code; `timeout` means it is still running, so read `output` as progress and call it again) instead of polling. `needs_approval` means the person must answer a prompt in that card: never answer it yourself. 4. `get_agent_result` to read what the subagent produced; check the actual files before reporting success. Do not explore the filesystem, `PATH` or config folders (for example `~/.local/bin` or `~/.config/`) looking for agents or their configuration: `list_providers` is what CanvasTTY can launch, and delegation goes through `spawn_agent`. diff --git a/docs/ARCHITECTURE.md b/docs/ARCHITECTURE.md index aa004a203..218c35143 100644 --- a/docs/ARCHITECTURE.md +++ b/docs/ARCHITECTURE.md @@ -70,7 +70,7 @@ Electron main process - `src/main/services/BrowserService.ts` is the only owner of the built-in browser's `WebContentsView` tabs and shared persistent partition. Remote pages have no preload or Node access, keep context isolation and sandbox enabled, and cannot request hardware, location, notification, clipboard-read, certificate-bypass, or external-protocol capabilities. HTTP(S) popups are adopted as internal tabs; other schemes are rejected. - `src/main/services/browser/` contains the browser kernel. `BrowserStore` atomically persists only tab order, active tab, and safe restore URLs. `BrowserPolicyService` centralizes URL, permission, download, and upload rules; validated uploads are copied through an already-open no-follow file descriptor into private staging before Chromium sees them. `BrowserAutomationService` attaches Electron's internal debugger to the existing live tab without a remote-debugging port. `BrowserCommandDispatcher` adds revisions, revision-bound refs, mutation request deduplication, per-tab FIFO mutation lanes, bounded concurrency, typed errors, and redacted fail-closed audit for agent mutations. - `src/main/services/agent-browser/` exposes the kernel only through an authenticated user-local Unix socket (`0600`) or Windows named pipe. The Windows pipe is created by the bundled native host with a protected DACL containing only the exact current-user SID and rejects remote clients. Each agent PTY receives a one-use bootstrap capability through its child environment. A successful authentication rotates it to a session-scoped reconnect capability held only in helper memory; duplicate bootstrap authentication is accepted only while the same `connectionId` is already live, and every capability is revoked when the PTY ends. The bundled stdio MCP helper is the only protocol adapter; no TCP listener, cookie/storage endpoint, arbitrary evaluation tool, or raw CDP surface exists. -- `src/main/services/agent-runtime/` is a separate lifecycle boundary and is not controlled by the Browser access switch. When CanvasTTY status hooks are enabled, every agent PTY receives a distinct capability for a protected user-local socket/pipe. Provider command hooks and the OpenCode event plugin may report only the fixed status enum, bounded event name, and optional opaque turn/prompt ID; prompt text, responses, tool input, and arbitrary telemetry are rejected by the exact gateway schema. On macOS and Linux the helpers are the native `canvastty-helper` (`native/canvastty-helper`, Go, standard library only; subcommands `mcp-browser`, `mcp-orchestration`, `permission-gate`, `hook`), which speaks exactly the wire protocols of the `.mjs` helpers (`tests/native-helper-*-parity.test.mjs` run both) without an Electron-as-Node process per agent and per tool call; `src/main/services/agentHelpers.ts` falls back to the `.mjs` helpers when the binary is absent, on Windows, or with `CANVASTTY_HELPERS=node`. Electron helper commands carry `ELECTRON_RUN_AS_NODE=1` inside the exact hook command only; the provider PTY never inherits that process-mode flag, so a provider cannot accidentally launch a second CanvasTTY GUI instance. The user can revoke this capability from Agents settings, immediately returning live agent status to `unavailable`; re-enabling requires a new/restarted PTY. Explicitly trusted plugin hooks use a separate process runner which re-checks the private PluginManager registry on every invocation and strips CanvasTTY internal capabilities before passing the provider payload to third-party code. Provider-native review remains an independent gate; CanvasTTY does not bypass Codex hook trust globally. +- `src/main/services/agent-runtime/` is a separate lifecycle boundary and is not controlled by the Browser access switch. When CanvasTTY status hooks are enabled, every agent PTY receives a distinct capability for a protected user-local socket/pipe. Provider command hooks and the OpenCode event plugin may report only the fixed status enum, bounded event name, and optional opaque turn/prompt ID; prompt text, responses, tool input, and arbitrary telemetry are rejected by the exact gateway schema. On macOS and Linux the helpers are the native `canvastty-helper` (`native/canvastty-helper`, Go, standard library only; subcommands `mcp-browser`, `mcp-orchestration`, `permission-gate`, `hook`), which speaks exactly the wire protocols of the `.mjs` helpers (`tests/native-helper-*-parity.test.mjs` run both) without an Electron-as-Node process per agent and per tool call; `src/main/services/agentHelpers.ts` falls back to the `.mjs` helpers when the binary is absent, on Windows, or with `CANVASTTY_HELPERS=node`. Electron helper commands carry `ELECTRON_RUN_AS_NODE=1` inside the exact hook command only; the provider PTY never inherits that process-mode flag, so a provider cannot accidentally launch a second CanvasTTY GUI instance. The user can revoke this capability from Agents settings, immediately returning live agent status to `unavailable`; re-enabling requires a new/restarted PTY. Explicitly trusted plugin hooks use a separate process runner which re-checks the private PluginManager registry on every invocation and strips CanvasTTY internal capabilities before passing the provider payload to third-party code. Provider-native review remains an independent gate; CanvasTTY does not bypass Codex hook trust globally. A Codex or OpenCode subagent's final answer (Codex `Stop` hook, the OpenCode plugin at `session.idle`) travels on the same hooks, so it exists only while status hooks are on for that launch: whether a launch has them is decided when it is planned (a deferred plugin launch included), and a launch planned with them off has no lease and no lifecycle hooks at all. `wait_for_agent` and `get_agent_result` then return `answerUnavailable: { reason: "status_hooks_off", message }` and the orchestrator reads the output instead. An answer is kept only for its own turn (`src/main/services/agent-runtime/applyRuntimeSignal.ts`, the glue main runs for each signal): OpenCode names each turn, a submitted input other than an approval answer or an empty Enter in an idle card starts a new one, and a signal of the current turn that predates that input still ends the wait but brings no answer. - Lifecycle adapters use launch-only settings for Claude, Codex, Qwen, and OpenCode. Kimi, Hermes, and Grok, whose hook discovery is home-config based, receive ownership-checked temporary entries shared across live CanvasTTY sessions. Kimi and Hermes keep recovery journals and exact backups; Grok uses a dedicated owned hook file. Cleanup restores exact original bytes when no concurrent edit occurred and otherwise removes only CanvasTTY-owned entries. - `TerminalManager` injects the MCP helper per launch without leaving permanent provider configuration. Claude Code, Codex, and Qwen Code receive CLI arguments; Qwen gets one inline `--mcp-config` entry that overrides only the CanvasTTY server name and leaves unrelated user servers available. OpenCode receives a merged launch-only `OPENCODE_CONFIG_CONTENT` entry plus one scoped browser-tool permission; Kimi uses its per-run MCP configuration when supported. Older Kimi versions receive a compare-and-swap temporary CanvasTTY entry and one exact permission rule with an atomic recovery journal. Hermes receives a temporary `mcp_servers.canvastty_browser` entry in `HERMES_HOME/config.yaml` (defaulting to `~/.hermes/config.yaml` on POSIX or `%LOCALAPPDATA%\hermes\config.yaml` on Windows); sensitive capability values stay as child-environment placeholders. Temporary Kimi and Hermes configuration remains until the final owning PTY session ends, then exact original bytes are restored when safe. A journal repairs an interrupted Hermes launch at the next CanvasTTY startup, while compare-and-swap checks preserve concurrent user edits. Unrelated MCP entries, credentials, and file/shell permissions are preserved. Qwen, OpenCode, and Hermes YOLO remain launch-only and do not change persistent permission settings. - `src/main/services/providerCliRegistry.ts` is the single owner of provider CLI discovery. During main-process startup it creates a shared snapshot for every provider in `PROVIDER_CLI_DEFINITIONS` — each definition declares the executable command names it may install as (which may differ from the provider ID, e.g. a provider shipping as `mcode`) and optional known home-relative or Windows LOCALAPPDATA-relative directories — by checking smoke-only overrides, the inherited `PATH`, platform defaults, and those known directories in that order. Available entries retain an absolute executable, launcher kind, and supplemented child `PATH`; POSIX entries must be executable files and Windows entries must be supported native or batch launchers. `TerminalManager`, `LimitsService`, agent-browser probes, and provider smoke tests consume that same snapshot and never repeat command lookup. Missing entries produce a failed session with copyable checked-path diagnostics before PTY or temporary browser configuration creation, while the limit adapter reports `cli-not-found`; its HOME row is hidden until explicitly selected after CLI detection. Launchers keep a provider without a local CLI visible when an account bound to a remote computer or a saved container profile provides another route. CanvasTTY never reads shell startup scripts. Agents settings can recheck candidate paths, atomically replace the registry snapshot, reconcile saved launcher and limit selections, and refresh CLI-bound adapters without restarting the app. Existing sessions keep running; a newly found CLI remains disabled until selected. diff --git a/docs/ARCHITECTURE.ru.md b/docs/ARCHITECTURE.ru.md index ae66a88d0..d7e0054a4 100644 --- a/docs/ARCHITECTURE.ru.md +++ b/docs/ARCHITECTURE.ru.md @@ -48,7 +48,7 @@ Electron main process - `src/main/services/PluginSecretsService.ts` сериализует запись секретов каждого плагина, шифрует весь ограниченный payload через Electron `safeStorage`, отклоняет plaintext-only backend и удаляет зашифрованный файл при uninstall. `ProviderSecretsService.ts` применяет ту же архитектуру к API-ключам провайдеров для CLI с BYOK: значения остаются в main-процессе, а renderer-контракт раскрывает только флаги `configured` и операции set/clear. Записи настроек `ApiProfile` именуют model-бэкенды (протокол, HTTPS base URL, ссылка на секрет) для тех же BYOK-рантаймов; это не agent providers, а normalizer настроек отбрасывает невалидные профили вместо «ремонта». - `src/main/services/PluginMediaService.ts` сохраняет разрешения только после нативного выбора папки, скрывает абсолютные пути, пропускает symlinks и отдаёт аудио с HTTP Range. Чтение плейлистов остаётся внутри разрешённых библиотек; ограниченная атомарная запись разрешена только в `Playlists/`. - `src/main/services/BrowserService.ts` владеет вкладками встроенного браузера в `WebContentsView`. Удалённые страницы используют отдельный persistent partition с выключенным Node, включёнными context isolation/sandbox и отклонением website permissions по умолчанию. Это core service, а не возможность runtime-плагина. -- `src/main/services/agent-runtime/` — отдельная всегда включённая lifecycle-граница, не зависящая от переключателя Browser access. Каждый agent PTY получает собственный capability для защищённого user-local socket/pipe. Provider command hooks и OpenCode event plugin могут передать только фиксированный status enum, ограниченное имя события и необязательный opaque turn/prompt ID; точная schema Gateway отклоняет prompt text, ответы, tool input и произвольную telemetry. При завершении PTY capability и временные файлы отзываются. +- `src/main/services/agent-runtime/` — отдельная всегда включённая lifecycle-граница, не зависящая от переключателя Browser access. Каждый agent PTY получает собственный capability для защищённого user-local socket/pipe. Provider command hooks и OpenCode event plugin могут передать только фиксированный status enum, ограниченное имя события и необязательный opaque turn/prompt ID; точная schema Gateway отклоняет prompt text, ответы, tool input и произвольную telemetry. При завершении PTY capability и временные файлы отзываются. Финальный ответ субагента Codex или OpenCode (хук Codex `Stop`, плагин OpenCode на `session.idle`) идёт через те же хуки, поэтому он есть только пока status hooks включены для этого запуска: это решается при планировании запуска (включая отложенный запуск с плагинами), и запуск, спланированный с выключенными хуками, не получает ни lease, ни lifecycle-хуков. Тогда `wait_for_agent` и `get_agent_result` возвращают `answerUnavailable: { reason: "status_hooks_off", message }`, и оркестратор читает вывод. Ответ хранится только для своего хода (`applyRuntimeSignal.ts`): OpenCode именует каждый ход, отправленный ввод (кроме ответа на запрос подтверждения и пустого Enter в idle-карточке) начинает новый, а сигнал текущего хода, начатого до этого ввода, завершает ожидание, но без ответа. - Claude, Codex, Qwen и OpenCode получают lifecycle hooks только на текущий запуск. Для Kimi, Hermes и Grok, которые ищут hooks в home-конфигурации, используются ownership-checked временные записи с совместным владением живых сессий. Kimi и Hermes используют recovery journals и точные backups, Grok — отдельный owned hook file; cleanup восстанавливает исходные байты или удаляет только записи CanvasTTY при конкурентных изменениях. - `TerminalManager` подмешивает MCP helper, не оставляя постоянных изменений в provider-конфигах. Claude Code, Codex и Qwen Code получают CLI arguments; Qwen получает одну inline-запись `--mcp-config`, которая переопределяет только имя сервера CanvasTTY и не скрывает сторонние user servers. OpenCode — объединённый launch-only `OPENCODE_CONFIG_CONTENT` с одной scoped browser-tool permission, Kimi — per-run MCP config или временную запись с compare-and-swap и recovery journal для старых версий. Hermes получает временную запись `mcp_servers.canvastty_browser` в `HERMES_HOME/config.yaml` (по умолчанию `~/.hermes/config.yaml` в POSIX или `%LOCALAPPDATA%\hermes\config.yaml` в Windows); чувствительные capability-значения остаются ссылками на окружение дочернего процесса. Временная конфигурация Kimi и Hermes живёт до завершения последней владеющей PTY-сессии, после чего исходные байты точно восстанавливаются, если файл не менялся параллельно. Journal восстанавливает Hermes после прерванного запуска при следующем старте CanvasTTY, а compare-and-swap сохраняет одновременные пользовательские изменения. Сторонние MCP-записи, credentials и file/shell permissions не затрагиваются. Qwen, OpenCode и Hermes YOLO остаются launch-only и не меняют постоянные permission-настройки. - `src/main/services/providerCliRegistry.ts` — единственный владелец обнаружения provider CLI. При запуске main-процесса он создаёт общий snapshot для каждого провайдера из `PROVIDER_CLI_DEFINITIONS` — каждое определение объявляет имена executable-команд, под которыми провайдер может устанавливаться (они могут отличаться от ID провайдера, например провайдер с командой `mcode`), и опциональные известные каталоги (относительно home или Windows LOCALAPPDATA), — последовательно проверяя smoke-only overrides, унаследованный `PATH`, системные каталоги платформы и эти известные каталоги. Доступная запись хранит абсолютный executable, тип launcher-а и дополненный дочерний `PATH`; POSIX-кандидат обязан быть исполняемым файлом, а Windows-кандидат — поддерживаемым native или batch launcher-ом. `TerminalManager`, `LimitsService`, agent-browser probes и provider smoke используют один и тот же snapshot и не повторяют поиск команды. Недоступный CLI создаёт failed-сессию с копируемой диагностикой проверенных путей до создания PTY или временной browser-конфигурации, а адаптер лимитов сообщает `cli-not-found`; строка HOME скрыта до ручного выбора после обнаружения CLI. Провайдер без локального CLI остаётся в панелях запуска, если для него есть аккаунт на удалённом компьютере или сохранённый контейнерный профиль. CanvasTTY не читает shell startup scripts. Настройки агентов позволяют повторно проверить пути, атомарно заменить snapshot registry, согласовать сохранённые списки запуска и лимитов и обновить зависящие от CLI адаптеры без перезапуска. Работающие сессии продолжаются; найденный позже CLI остаётся выключенным до ручного выбора. diff --git a/docs/ARCHITECTURE.zh-CN.md b/docs/ARCHITECTURE.zh-CN.md index 34dcc04e0..1f7d0b66e 100644 --- a/docs/ARCHITECTURE.zh-CN.md +++ b/docs/ARCHITECTURE.zh-CN.md @@ -49,7 +49,7 @@ Electron main process - `src/main/services/PluginSecretsService.ts` 串行化每个插件的机密写入，通过 Electron `safeStorage` 加密完整的有界 payload，拒绝 plaintext-only backend，并在卸载时删除加密文件。`ProviderSecretsService.ts` 将同一架构应用于面向 BYOK CLI 的服务商 API key：值只留在 main 进程，renderer 契约只暴露每个 key 的 `configured` 标志与 set/clear 操作。`ApiProfile` 设置项为同一批 BYOK 运行时命名 model 后端（协议、HTTPS base URL、secret 引用）；它们不是 agent provider，且 settings normalizer 会丢弃而非修复无效 profile。 - `src/main/services/PluginMediaService.ts` 仅在原生目录选择后保存授权，隐藏绝对路径，跳过 symlink，并以 HTTP Range 提供音频。Playlist 读取限制在授权媒体库内；写入受大小限制，并且只能原子写入 `Playlists/`。 - `src/main/services/BrowserService.ts` 管理内置浏览器的 `WebContentsView` tab。远程页面使用独立 persistent partition，禁用 Node，启用 context isolation/sandbox，并默认拒绝网站权限。这是 core service，不是 runtime 插件能力。 -- `src/main/services/agent-runtime/` 是独立且始终启用的 lifecycle 边界，不受 Browser access 开关控制。每个 agent PTY 都为受保护的 user-local socket/pipe 获得独立 capability。Provider command hook 与 OpenCode event plugin 只能提交固定 status enum、受限 event 名称和可选 opaque turn/prompt ID；Gateway 的精确 schema 会拒绝 prompt text、回复、tool input 与任意 telemetry。PTY 退出时 capability 与临时文件都会被撤销。 +- `src/main/services/agent-runtime/` 是独立且始终启用的 lifecycle 边界，不受 Browser access 开关控制。每个 agent PTY 都为受保护的 user-local socket/pipe 获得独立 capability。Provider command hook 与 OpenCode event plugin 只能提交固定 status enum、受限 event 名称和可选 opaque turn/prompt ID；Gateway 的精确 schema 会拒绝 prompt text、回复、tool input 与任意 telemetry。PTY 退出时 capability 与临时文件都会被撤销。 Codex 或 OpenCode 子 agent 的最终回答（Codex `Stop` hook、OpenCode 插件在 `session.idle` 时）经由同一组 hook 传递，因此只有该次启动开启了 status hooks 时才存在：是否开启在规划启动时决定（包括延迟的插件启动），在关闭状态下规划的启动既没有 lease 也没有任何 lifecycle hook。此时 `wait_for_agent` 与 `get_agent_result` 返回 `answerUnavailable: { reason: "status_hooks_off", message }`，编排者改为读取输出。回答只归属于它自己的 turn（`applyRuntimeSignal.ts`）：OpenCode 为每个 turn 命名；除审批回答和空闲卡片中的空 Enter 外，提交的输入会开始新 turn；在该输入之前开始的当前 turn 的信号仍会结束等待，但不带回答。 - Claude、Codex、Qwen 与 OpenCode 使用仅本次启动有效的 lifecycle hook。Kimi、Hermes 与 Grok 只能从 home 配置发现 hook，因此使用由实时 CanvasTTY 会话共享、带 ownership 检查的临时条目。Kimi 与 Hermes 使用 recovery journal 和精确 backup；Grok 使用独立 owned hook 文件。Cleanup 在无并发编辑时逐字恢复原文件，否则只移除 CanvasTTY 自己的条目。 - `TerminalManager` 注入 MCP helper 时不会留下永久的服务商配置变更。Claude Code、Codex 与 Qwen Code 使用 CLI 参数；Qwen 使用一个 inline `--mcp-config`，只覆盖 CanvasTTY 服务名，不隐藏无关用户服务。OpenCode 使用合并后的、仅本次启动有效的 `OPENCODE_CONFIG_CONTENT` 和一条 scoped browser-tool 权限；Kimi 使用 per-run MCP 配置，旧版本则使用带 compare-and-swap 与 recovery journal 的临时配置。Hermes 会在 `HERMES_HOME/config.yaml` 中获得临时 `mcp_servers.canvastty_browser` 配置项（POSIX 默认路径为 `~/.hermes/config.yaml`，Windows 默认路径为 `%LOCALAPPDATA%\hermes\config.yaml`），敏感 capability 值仍以子进程环境变量占位符保存。Kimi 与 Hermes 的临时配置会保留到最后一个所属 PTY 会话结束；若文件未被并发修改，则精确恢复原始字节。若 Hermes 启动意外中断，journal 会在 CanvasTTY 下次启动时修复配置，compare-and-swap 则保留用户的并发修改。其他 MCP 配置项、凭据和文件/shell 权限不会受影响。Qwen、OpenCode 与 Hermes 的 YOLO 都不修改持久权限设置。 - `src/main/services/providerCliRegistry.ts` 是服务商 CLI 发现的唯一职责边界。main 进程启动时，它按 smoke-only override、继承的 `PATH`、平台默认目录、已知用户/服务商目录的顺序，为 `PROVIDER_CLI_DEFINITIONS` 中的每个服务商创建一个共享快照——每个定义声明该服务商可能安装的 executable 命令名（可以与服务商 ID 不同，例如命令为 `mcode` 的服务商），以及可选的已知目录（相对 home 或 Windows LOCALAPPDATA）。可用条目保存绝对 executable、launcher 类型以及补充后的子进程 `PATH`；POSIX 候选必须是可执行文件，Windows 候选必须是受支持的 native 或 batch launcher。`TerminalManager`、`LimitsService`、agent-browser probe 与 provider smoke 共用该快照，不再各自查找命令。CLI 不可用时，系统会在创建 PTY 或临时 browser 配置之前生成 failed session，并提供可复制的已检查路径诊断；限额适配器报告 `cli-not-found`；HOME 行保持隐藏，直到检测到 CLI 后由用户手动选择。若某服务商绑定了远程计算机上的账户或已保存的容器配置，即使本地没有 CLI，启动器仍会显示它。CanvasTTY 不读取 shell startup script。Agents 设置可重新检查候选路径、原子替换 registry 快照、调整已保存的启动器和限额选择，并在无需重启的情况下刷新依赖 CLI 的适配器。运行中的 session 保持不变；新检测到的 CLI 需手动启用。 diff --git a/docs/agent-orchestration.md b/docs/agent-orchestration.md index 0e8b6e3f4..bf3c4a6b6 100644 --- a/docs/agent-orchestration.md +++ b/docs/agent-orchestration.md @@ -9,10 +9,10 @@ A session launched from the desktop with the **Orchestrator** role gets the `can | Tool | What it does | | --- | --- | | `list_providers` | The agent providers CanvasTTY can launch as subagents: `id` (the exact `spawn_agent.provider` value), `name`, `installed`/`available` (from the provider CLI registry resolved at startup or on a recheck), `signIn` (`ok`, `signed_out`, `expired` or `unknown`, from the last usage-limits read; it never starts a read and never reads credentials), `subagent` and `orchestrator` support, `model` (whether the CLI takes one, its format, and for OpenCode the models a cached `opencode models` listing reported: run in the background with a 5 s timeout, refreshed after 10 minutes, never awaited) and `efforts`, and plugin launch options (`launchOptions`, plus `launchOptionTools` such as an accounts plugin's `list_routes`). Fast: cached state only, no network calls. | -| `spawn_agent` | Launches a subagent (profile: see below). `provider` must be a `list_providers` id (the schema lists them); an unknown one is refused with `INVALID_REQUEST` naming `list_providers`. Optional `model` goes to the CLI's own `--model` for that run (OpenCode `provider/model`; Codex, Claude, Qwen, Kimi alias, Grok, OMP, Pi, Cursor; Hermes, MiniMax, Devin and Antigravity take none) and optional `effort` to its reasoning effort (Codex `-c model_reasoning_effort`, Claude `--effort`, Grok `--reasoning-effort`). Both are checked per provider, kept on restart and restore, and never written to the CLI's configuration. If the person names a model, the orchestrator passes it as `model`. An OpenCode model its own `opencode models` list does not contain is refused before launch with up to five closest listed ids (OpenCode would otherwise stop with only "Unexpected server error"); when that list cannot be read, the model is passed as given. A subagent that exits anyway reports the last lines of its screen as `exitLines` in `wait_for_agent`, `observe_agent` and `get_agent_result`. | -| `wait_for_agent` | `{ sessionId, timeoutSeconds ≤ 600 }` (default 55). Returns when the subagent is `idle` (after a prompt: only once a turn that started after the latest delivered prompt has ended, so the CLI's startup idle does not count), `needs_approval`, `done`/`failed` (exited), `quiet` (reports no status and its screen stopped changing), `closed`, or on `timeout`, with `status`, `exitCode`, `waitedMs` and the masked terminal tail (masked before it is cut). Only this orchestrator's own subagents; it stops at once when the call is canceled or the orchestrator disconnects. | -| `get_agent_result`, `observe_agent` | `get_agent_result`: `answer`, the final reply of the subagent's last turn as the agent itself reported it (Codex through its Stop hook, OpenCode through CanvasTTY's plugin reading the session's last assistant message at `session.idle`; at most 4,096 characters with the end kept, masked, in memory only and cleared when the next turn starts), plus `status`, the exit state and the masked terminal tail. Other providers have no `answer`: their tail is raw screen output. `observe_agent`: the current status and tail. | -| `send_to_agent`, `cancel_agent`, `list_agents` | Follow-up prompts, disposing a subagent, listing this session's subagents. | +| `spawn_agent` | Launches a subagent (profile: see below). `provider` must be a `list_providers` id (the schema lists them); an unknown one is refused with `INVALID_REQUEST` naming `list_providers`. Optional `model` goes to the CLI's own `--model` for that run (OpenCode `provider/model`; Codex, Claude, Qwen, Kimi alias, Grok, OMP, Pi, Cursor; Hermes, MiniMax, Devin and Antigravity take none) and optional `effort` to its reasoning effort (Codex `-c model_reasoning_effort`, Claude `--effort`, Grok `--reasoning-effort`). Both are checked per provider, kept on restart and restore, and never written to the CLI's configuration. If the person names a model, the orchestrator passes it as `model`. Plugin launch options the call does not name are inherited from the orchestrator for a subagent of the same CLI (see Plugin launch options below); `model` next to inherited options that set the model is refused, and so is a subagent of another CLI while the orchestrator's model comes from a plugin's options. The answer's `servedBy` is `{ provider, model, launchOptions: { : "inherited" | "explicit" } }`. An OpenCode model its own `opencode models` list does not contain is refused before launch with up to five closest listed ids (OpenCode would otherwise stop with only "Unexpected server error"); when that list cannot be read, the model is passed as given. A subagent that exits anyway reports the last lines of its screen as `exitLines` in `wait_for_agent`, `observe_agent` and `get_agent_result`. | +| `wait_for_agent` | `{ sessionId, timeoutSeconds ≤ 100 }` (default 50). For an orchestrator other than Claude Code a call returns after at most 50 seconds, because its MCP client (Codex, OpenCode and others) ends a tool call after about 60 seconds; `timeout` then means the subagent is still running, with its progress in `output` and a `message` to call again. Returns when the subagent is `idle` (after a prompt: only once a turn that started after the latest delivered prompt has ended, so the CLI's startup idle does not count), `needs_approval`, `done`/`failed` (exited), `quiet` (reports no status and its screen stopped changing), `closed`, or on `timeout`, with `status`, `exitCode`, `waitedMs`, the masked terminal tail (masked before it is cut) and, for Codex and OpenCode subagents, `answer` (see `get_agent_result`). A turn that the person interrupted by submitting their own input in the subagent's card still ends the wait, without an `answer`: that answer is no longer for the orchestrator's prompt. With CanvasTTY status hooks off (Settings → Agents) a subagent has no hooks at all, so no `answer` can be captured: the result carries `answerUnavailable: { reason: "status_hooks_off", message }` instead and the wait ends `quiet`; read `output` (or `observe_agent`). Turning status hooks on applies to subagents started or restarted afterwards. Only this orchestrator's own subagents; it stops at once when the call is canceled or the orchestrator disconnects. | +| `get_agent_result`, `observe_agent` | `get_agent_result`: `answer`, the final reply of the subagent's last turn as the agent itself reported it (Codex through its Stop hook, OpenCode through CanvasTTY's plugin reading the session's last assistant message at `session.idle`; at most 4,096 characters with the end kept, masked, in memory only, kept only for the turn it ended and cleared when the next input is submitted or the next turn starts; an Enter with nothing typed in an idle card keeps it; an answer reported as the subagent exits still counts; `answerUnavailable` instead when status hooks are off), plus `status`, the exit state and the end of the subagent's screen as masked plain text (a headless terminal renders the redraws of a full-screen TUI such as OpenCode, so no control sequences remain; `wait_for_agent` returns the same). Other providers have no `answer`: read that text. `observe_agent`: the current status and the end of the screen as masked plain text, rendered like `wait_for_agent`'s `output` (at most `maxChars`). | +| `send_to_agent`, `cancel_agent`, `list_agents` | Follow-up prompts, disposing a subagent, listing this session's subagents. Prompts to one subagent arrive in the order they were sent. OpenCode drops keys typed while it starts, so a prompt for it (also `spawn_agent`'s first one) is typed only once its input field is up: the terminal control sequence that shows its text cursor on its full screen, or a lifecycle hook only 2 s after its first one (OpenCode's plugin can report shortly before the field takes keys). This wait is used on macOS and Linux; on Windows, whose pseudo console may not pass those sequences, OpenCode gets input as soon as its PTY exists, as before. `spawn_agent` (its checks, the launch and this wait together) and `send_to_agent` wait at most 45 s, under the MCP client's tool deadline; then the call fails with the reason and the session id, and nothing is typed. A send cancelled while it waits returns at once. | ### Subagent profiles @@ -43,9 +43,9 @@ The core applies these to every way an agent can start another one: `spawn_agent | Rule | What it guarantees | | --- | --- | | Profile ceiling | A subagent's profile is at most its orchestrator's, never YOLO. A saved subagent whose record says more is restored with its orchestrator's ceiling. | -| Project folder | A subagent's `cwd` must be the folder the person chose for the agent it descends from, or a folder inside it, compared as real paths in both Unicode spellings (NFC and NFD name the same folder on macOS). `/`, the home folder or another project is refused: "A subagent works only inside this project's folder …". A relative `cwd` is taken from the orchestrator's folder. | +| Project folder | A subagent's `cwd` must be the folder the person chose for the agent it descends from, or a folder inside it, compared as real paths in both Unicode spellings (NFC and NFD name the same folder on macOS). `/`, the home folder or another project is refused: "A subagent works only inside this project's folder …". CanvasTTY's own private data is refused even when the project contains it (a project at the home folder): its tokens, account homes and plugins' data ("A subagent cannot work in … it is CanvasTTY's private data"); a folder inside one plugin's data only when the project itself is a folder inside that same plugin's data (an environment's worktree). A relative `cwd` is taken from the orchestrator's folder. | | Limits | At most `orchestrationMaxDepth` levels of subagents below the agent the person started (2 by default) and `orchestrationMaxSubagents` live subagents in one orchestration, all levels together (8 by default). Only the person sets them, in Settings → Agents. Exited subagents do not count. | -| Plugin launch options | `launchOptions` from an orchestrator reach only plugins that declared `launch.delegable: true`; everything a plugin contributes still passes the core-owned flag and configuration checks (no permission, sandbox, hook or bypass flags, no OpenCode/Kimi configuration that sets approvals). | +| Plugin launch options | `launchOptions` from an orchestrator reach only plugins that declared `launch.delegable: true` (a plugin the spawn does not name keeps the orchestrator's own values for a subagent of the same CLI; repeating them needs no declaration; a plugin turned off since is left out, or refused when it set the model; changed plugin fields refuse with that reason); everything a plugin contributes still passes the core-owned flag and configuration checks (no permission, sandbox, hook or bypass flags, no OpenCode/Kimi configuration that sets approvals). | | YOLO | Enforced in the main process: only for a CLI the person acknowledged YOLO for in the launcher, never for a subagent, never from an orchestrator's control connection; a plugin with `sessions:launch` needs the same acknowledgement. | | Isolation | Every subagent and every agent a plugin starts runs inside agent isolation (below) when it is on; where there is no layer, a subagent runs in normal (it asks) and its card says why. | | No settings | The `canvastty_agents` tools and an orchestrator's control connection cannot change settings, base protection, profiles, plugin trust or isolation. | @@ -54,10 +54,11 @@ The core applies these to every way an agent can start another one: `spawn_agent Settings → Agents → Agent isolation (on by default) wraps an agent's whole process tree (the CLI, its commands, its MCP servers and hooks) in an operating-system layer. It applies to every subagent, every agent a plugin starts, and every agent in a profile other than normal (manual), whoever launched it. Inside it: -- files can be written only in the project folder (both spellings), the launch's own temporary folder (`TMPDIR`), the CLI's own state, configuration and cache folders, and the npm/bun caches; the project's git hooks and an existing repository's git config, and the CLI's own permission settings files (`~/.codex/config.toml`, `~/.claude/settings.json`, OpenCode's config) are not writable; -- SSH, cloud and package-registry credentials, other CLIs' folders (their credentials) and CanvasTTY's own tokens, secret stores, account homes and launch files are unreadable, except what this launch was handed (its own control connection, its account home); +- files can be written only in the project folder (both spellings), the launch's own temporary folder (`TMPDIR`; for Claude also `CLAUDE_CODE_TMPDIR`, so its Bash tool and editor files stay out of the `/tmp/claude-` folder all Claude sessions share), the CLI's own state, configuration and cache folders, and the npm/bun caches; the project's git hooks and an existing repository's git config, and the CLI's own permission settings files (`~/.codex/config.toml`, `~/.claude/settings.json`, OpenCode's config) are not writable; +- SSH, cloud and package-registry credentials, other CLIs' folders (their credentials) and CanvasTTY's own tokens, secret stores, account homes, launch files and plugins' private data are unreadable, except what this launch was handed (its own control connection, its account home, the one CLI home its launch plugin selected); - no other process can be signalled, no app opened through Launch Services, no Apple events sent (osascript driving Terminal), no preferences written through cfprefsd, no launchd job submitted, and Unix sockets reach only DNS, syslog, the launch's own folder and CanvasTTY's token-checked gateways (not Docker, tmux or an SSH agent); - network and the keychain's services work as before: the CLI reaches its provider and reads its own sign-in. +- when the project contains CanvasTTY's own data folder (a project at the home folder), that folder stays read-only as a whole (its settings, its plugins); only the CLI's own folders inside it (a selected account home) stay writable. Its parent folders inside the project (such as `~/.config`) cannot be renamed or replaced; entries can still be created in them. A project CanvasTTY keeps inside its data folder is writable as usual. The agent gets `CANVASTTY_ISOLATION` explaining the rule, so "Operation not permitted" outside it is understood rather than worked around. macOS uses `sandbox-exec` with a profile generated per launch; Linux uses bubblewrap (`bwrap`) when it is installed (read-only root, the same writable folders, empty tmpfs over what may not be read, its own PID namespace, the user runtime folder hidden). Windows has no layer yet. Where there is no layer (Windows, Linux without bubblewrap) a subagent runs in normal (it asks) and a CLI without an auto mode of its own gets no auto; the card shows the reason. Turning isolation off is the person's opt-in: subagents then keep their profile without the layer. The layer fails closed: when it cannot be set up the launch is refused with the reason, never started without it. @@ -67,6 +68,8 @@ Claude Code keeps its sign-in in the macOS login keychain and rewrites that file After an isolated session ends, is closed or is restored after a quit, CanvasTTY audits the repositories under its folder whose git folder changed: config keys and files git would execute outside the layer (`core.hooksPath`, filter and diff drivers, `fsmonitor`, `!` aliases, hooks, `info/attributes`) are listed in a notice, and the person can neutralize them or keep them. See [Installing and security](installing-and-security.md#agent-protection-layers). +Inside the layer, the files git reads its config and paths from are not writable in any repository of the project, existing or not: `.git/commondir`, `.git/config.worktree` and the linked worktrees' records in `.git/worktrees/` (so `git worktree add` fails there); objects, refs, the index, logs and `HEAD` stay writable, so add, commit, checkout, branch and stash work. A repository's `.git`, `.git/hooks` and `.git/info` folders cannot be renamed or removed inside the layer, so these rules cannot be sidestepped by moving the folder away and back. On Linux these are mounted read-only for the project's own existing repository; a `commondir` that does not exist yet cannot be mounted (an empty one would stop git), and repositories created during the session are left to the audit above. + ## Plugin environments A plugin environment declares in its manifest what it keeps (`keeps`): `launch` (the launch's arguments and environment reach the agent unchanged, so CanvasTTY's hooks and the profile's per-run settings work there), `isolated` (it does not run on this computer's files: a container or a remote host) and `confines`. Without `launch`, any profile but normal is refused and the card says that base protection and CanvasTTY's hooks do not reach the agent there. An `isolated` environment is not wrapped again (its own boundary applies; the card says so); a local one (a worktree) runs inside the layer with the worktree as its project. diff --git a/docs/installing-and-security.md b/docs/installing-and-security.md index c81480251..a372e9609 100644 --- a/docs/installing-and-security.md +++ b/docs/installing-and-security.md @@ -87,11 +87,11 @@ Agents work in their own mode; CanvasTTY's layers sit outside it as a quiet safe | Layer | Guarantees | Does not guarantee | | --- | --- | --- | | The agent's own permission system | What its CLI promises in the chosen mode (Claude Code, Codex, OpenCode, Grok…). | Anything the CLI's own classifier or reviewer lets through. | -| Base protection (hooks) | Denies elevation, pipes into a shell, disk commands, fork bombs, writes and deletes outside the project and access to CanvasTTY's private data, before the tool call runs, for agents whose CLI has hooks (Claude Code, Codex, Qwen Code, OpenCode). A decision plugin's "ask" is shown by Claude Code; for a CLI that cannot ask it is a deny with the reason. | It reads commands; code that hides what it does, or a CLI without hooks, is not covered. It is a guard, not a sandbox. | +| Base protection (hooks) | Denies elevation, pipes into a shell, disk commands, fork bombs, writes and deletes outside the project and access to CanvasTTY's private data, before the tool call runs, for agents whose CLI has hooks (Claude Code, Codex, Qwen Code, OpenCode). A decision plugin's "ask" is shown by Claude Code; for a CLI that cannot ask it is a deny with the reason. | Best effort: it reads tool calls, so code that hides what it does, a path built at run time, a script that walks folders itself, a recursive search started above CanvasTTY's data (a project at the home folder), or a CLI without hooks is not covered. It is a guard, not a sandbox: a project that contains CanvasTTY's data needs agent isolation to keep that data private. | | Delegation rules | A subagent never gets more than its orchestrator (never YOLO), works only in its orchestrator's project folder, within the depth and count limits the person set; orchestrators get a control connection of their own; no agent-facing tool changes settings, protection, profiles, trust or isolation; plugin launch options from an agent only where the plugin declared them delegable. | What a plugin the person trusted does with its own permissions. | | Agent isolation (OS layer) | On macOS (sandbox-exec) and Linux (bubblewrap), for subagents, plugin-started agents and every agent not in Manual: the whole process tree writes only in the project, its own temporary folder and its CLI's own folders; keys, other CLIs' credentials and CanvasTTY's tokens are unreadable; no other process, app, Apple event, preference write, launchd job or foreign Unix socket (Docker, tmux, SSH agent) is reachable. Fails closed. | Network is not restricted. Linux cannot filter Unix sockets, only hide the user runtime folder. The CLI's own sandbox cannot run inside the layer on macOS, so it is switched off there (approvals unchanged) and commands can use the network. A Claude launch may rewrite the login keychain file to save a refreshed sign-in, so it could also damage that file (other items stay behind their own access lists). Windows has no layer yet: subagents run in Manual there unless the person turns isolation off. | -Isolation keeps approval configuration read-only, including OpenCode's project and custom config files and its `agent`/`agents` directories. CLI state and caches remain writable. A moved CLI home is refused if its real path overlaps host credentials or private CanvasTTY data; a selected account folder below `account-homes` remains available. Credential exclusions use the host environment captured before launch contributions, even if the launch changes `HOME`. +Isolation keeps approval configuration read-only, including OpenCode's project and custom config files and its `agent`/`agents` directories. CLI state and caches remain writable. A moved CLI home is refused if its real path overlaps host credentials or private CanvasTTY data; a selected account folder below `account-homes` remains available. Plugins' private data (`plugin-data`) is unreadable as a whole; only the CLI home a chosen launch plugin started this agent with, an exact real folder inside that plugin's own data, is reopened, and a project inside plugin data is reopened as that one folder only where CanvasTTY placed the launch: a folder an environment prepared inside its own plugin's data, or the folder the person chose for the card (and its subagents' folders below it). Any other project inside plugin data, such as a subagent folder an agent asked for, is refused. On Linux, CanvasTTY's gateway socket folders are bound read-only, and only the sockets CanvasTTY's own gateways set up for the launch are: a `CANVASTTY_*_ADDRESS` from the person's environment or a plugin binds nothing, and a gateway socket folder that holds a hidden or writable folder is refused. Credential exclusions use the host environment captured before launch contributions, even if the launch changes `HOME`. OpenCode Auto inspects only regular local config files of at most 1 MiB. An existing source that cannot be inspected adds no Auto allowances. A launch contributor's uninspectable config is refused; only absent optional directory candidates are skipped. JSONC comments and trailing commas are parsed without changing quoted strings. diff --git a/docs/installing-and-security.ru.md b/docs/installing-and-security.ru.md index 0532080dd..939392b31 100644 --- a/docs/installing-and-security.ru.md +++ b/docs/installing-and-security.ru.md @@ -83,9 +83,9 @@ gh variable set SPARKLE_PUBLIC_ED_KEY --repo howdeploy/CanvasTTY \ Агент работает в своём режиме; слои CanvasTTY стоят снаружи как тихая страховка: не добавляют вопросов при обычной работе в проекте и останавливают опасное с короткой причиной. - **Режимы запуска**: Авто (по умолчанию), Вручную, Правки, План, Обход — только там, где их поддерживает CLI. Обход выбирает человек для каждого CLI, он проверяется в main-процессе, никогда не даётся субагенту и всё равно остаётся внутри слоёв ниже. -- **Базовая защита** (хуки) запрещает повышение прав, `curl | sh`, дисковые команды и запись вне проекта у агентов с хуками; это проверка команд, а не песочница. +- **Базовая защита** (хуки) запрещает повышение прав, `curl | sh`, дисковые команды и запись вне проекта у агентов с хуками, а также доступ к закрытым данным CanvasTTY; это проверка команд по возможности, а не песочница: проект, содержащий данные CanvasTTY (например, домашняя папка), защищает только изоляция. - **Правила делегирования**: субагент получает не больше оркестратора (никогда YOLO), работает только в папке проекта, в пределах глубины и числа, заданных человеком; агенты не могут менять настройки, защиту, профили и изоляцию. -- **Изоляция агентов** (macOS — sandbox-exec, Linux — bubblewrap): запись только в проект, свою временную папку и папки своего CLI; ключи и токены недоступны; другие процессы, приложения и демоны недоступны; при сбое запуск отклоняется. Сеть не ограничивается; на Windows слоя пока нет, и субагенты там работают вручную. Подробности — в [agent-orchestration.md](agent-orchestration.md). +- **Изоляция агентов** (macOS — sandbox-exec, Linux — bubblewrap): запись только в проект, свою временную папку и папки своего CLI; ключи и токены недоступны; другие процессы, приложения и демоны недоступны; при сбое запуск отклоняется. Приватные данные плагинов (`plugin-data`) недоступны целиком; снова открываются только папка CLI, с которой выбранный плагин запуска стартовал агента (точная реальная папка внутри данных этого плагина), и проект внутри данных плагинов — как одна эта папка и только там, где запуск разместил CanvasTTY: папка, подготовленная окружением внутри данных своего плагина, или папка, выбранная человеком для карточки (и папки её субагентов ниже). Любой другой проект внутри данных плагинов, например папка субагента, о которой попросил агент, отклоняется. На Linux папки сокетов шлюзов CanvasTTY подключаются только для чтения, и только сокеты, которые собственные шлюзы CanvasTTY подготовили для этого запуска: `CANVASTTY_*_ADDRESS` из окружения человека или от плагина ничего не подключает, а папка сокетов шлюза со скрытой или записываемой папкой отклоняется. Сеть не ограничивается; на Windows слоя пока нет, и субагенты там работают вручную. Подробности — в [agent-orchestration.md](agent-orchestration.md). **Аудит git.** Когда изолированная сессия агента заканчивается, закрывается или восстанавливается после выхода, CanvasTTY проверяет репозитории в её папке, у которых изменилась папка git. Если git теперь запустит что-то вне изоляции (`core.hooksPath`, фильтр или diff-драйвер, `fsmonitor`, файл хука, `info/attributes`), появляется уведомление со списком изменений; **Обезвредить** удаляет эти ключи и отключает файлы, **Оставить как есть** ничего не трогает. Другие изменения файлов в проекте аудит не откатывает. diff --git a/docs/installing-and-security.zh-CN.md b/docs/installing-and-security.zh-CN.md index 605deda0b..533c9913c 100644 --- a/docs/installing-and-security.zh-CN.md +++ b/docs/installing-and-security.zh-CN.md @@ -83,9 +83,9 @@ gh variable set SPARKLE_PUBLIC_ED_KEY --repo howdeploy/CanvasTTY \ 智能体按自己的模式工作；CanvasTTY 的各层位于其外部，作为安静的安全网：项目内的常规工作不会增加提问，只拦截危险操作并给出简短原因。 - **启动模式**：自动（默认）、手动、接受编辑、计划、绕过，仅在 CLI 支持时提供。绕过由用户按 CLI 确认，在主进程中检查，绝不交给子智能体，并且仍处于下列各层之内。 -- **基础保护**（hooks）为有 hooks 的智能体拒绝提权、`curl | sh`、磁盘命令以及项目外的写入；它检查命令，不是沙箱。 +- **基础保护**（hooks）为有 hooks 的智能体拒绝提权、`curl | sh`、磁盘命令以及项目外的写入，并拒绝访问 CanvasTTY 的私有数据；它是尽力而为的命令检查，不是沙箱：包含 CanvasTTY 数据的项目（例如主目录）只有在隔离下才受保护。 - **委派规则**：子智能体的权限不超过其编排者（绝不为 YOLO），只在项目文件夹内工作，并受用户设定的深度和数量限制；智能体无法更改设置、防护、配置档或隔离。 -- **智能体隔离**（macOS 使用 sandbox-exec，Linux 使用 bubblewrap）：只能写入项目、自身临时目录和其 CLI 的目录；密钥和令牌不可读；其他进程、应用和守护进程不可达；无法建立时拒绝启动。网络不受限制；Windows 暂无此层，子智能体在那里以手动模式运行。详见 [agent-orchestration.md](agent-orchestration.md)。 +- **智能体隔离**（macOS 使用 sandbox-exec，Linux 使用 bubblewrap）：只能写入项目、自身临时目录和其 CLI 的目录；密钥和令牌不可读；其他进程、应用和守护进程不可达；无法建立时拒绝启动。插件的私有数据（`plugin-data`）整体不可读；只重新开放两类路径：所选启动插件启动该智能体时使用的 CLI 目录（该插件自身数据内的一个确切真实文件夹），以及插件数据内的项目——仅作为这一个文件夹，且仅限 CanvasTTY 安排启动的位置：环境在其插件自身数据内准备的文件夹，或用户为卡片选择的文件夹（及其下方子智能体的文件夹）。插件数据内的任何其他项目（例如智能体要求的子智能体文件夹）都会被拒绝。在 Linux 上，CanvasTTY 网关的 socket 文件夹以只读方式挂载，且只挂载 CanvasTTY 自身网关为本次启动准备的 socket：来自用户环境或插件的 `CANVASTTY_*_ADDRESS` 不会挂载任何东西，包含隐藏或可写文件夹的网关 socket 文件夹会被拒绝。网络不受限制；Windows 暂无此层，子智能体在那里以手动模式运行。详见 [agent-orchestration.md](agent-orchestration.md)。 **Git 审计。** 隔离的智能体会话结束、被关闭或在退出后恢复时，CanvasTTY 会检查其文件夹下 git 目录发生变化的仓库。如果 git 现在会在隔离外运行某些东西（`core.hooksPath`、filter 或 diff 驱动、`fsmonitor`、hook 文件、`info/attributes`），会出现一条列出具体变化的通知；**Neutralize** 删除这些键并停用这些文件，**Keep as is** 保持不变。它不会撤销项目内的其他文件改动。 diff --git a/docs/plugins.md b/docs/plugins.md index 0047ad17b..28732f39a 100644 --- a/docs/plugins.md +++ b/docs/plugins.md @@ -196,7 +196,7 @@ Up to 8 fields; `kind` is `boolean`, `select` (1–16 options) or `text` (at mos A `select` with `"optionsFrom": "service"` also lists choices the service offers, such as its own accounts. When the launcher opens, CanvasTTY asks the service `canvastty.launch.options` `{ provider, fields: [keys] }` and waits at most 3 s; the answer `{ "": [{ value, label }] }` adds up to 64 choices per field after the declared ones (which stay required and are all the launcher shows when the service does not answer). Because such a list can change after a card was saved, its value is accepted as any text up to 200 characters without control characters, and `canvastty.launch.prepare` must check it and refuse a value it no longer knows. -Orchestrators pass the same values to `spawn_agent` as `launchOptions` (`{ "": { "": value } }`), checked exactly like the launcher's, but only to a plugin whose `launch` declares `"delegable": true` (its options are safe for an agent to choose; without it only the person chooses them, and `list_providers` does not list them); a plugin tool can hand them out (for example the account it picked). Whatever a contribution adds is checked the same way for every launch: no permission, sandbox, hook or bypass flags, and no configuration the CLI reads that sets approvals (an OpenCode `OPENCODE_CONFIG` file or `OPENCODE_CONFIG_DIR`, `OPENCODE_PERMISSION`, Kimi's `--config`/`--config-file`); a file CanvasTTY cannot read and check is refused. While a child's launch waits for its plugins (launch options, a launch policy, an environment), `spawn_agent` answers only after its `prompt` reached the started agent, and `send_to_agent` waits the same way. A refused, failed or cancelled launch fails the call with the reason and the session id (the card stays); the text is dropped, never kept for a later restart. The control CLI answers `NOT_READY` for such a card. +Orchestrators pass the same values to `spawn_agent` as `launchOptions` (`{ "": { "": value } }`), checked exactly like the launcher's, but only to a plugin whose `launch` declares `"delegable": true` (its options are safe for an agent to choose; without it only the person chooses them, and `list_providers` does not list them); a plugin tool can hand them out (for example the account it picked). A plugin the spawn does not name keeps the orchestrator's own values for a subagent of the same CLI (the person chose them; this needs no `delegable`), so a subagent never falls back silently to its CLI's default model; a subagent of another CLI is refused while a plugin's contribution sets the orchestrator's model (its CLI model argument or `thirdPartyModel`) unless the spawn names that plugin, and `spawn_agent` answers with `servedBy` (CLI, requested model, and whether each plugin's options were inherited or explicit). An inherited plugin that was turned off since the orchestrator started is left out, unless it set the model (then the spawn is refused); one whose fields changed refuses the spawn with that reason. For a subagent's launch, a contribution that chooses the model (`--model`/`--model=`, `-m` for OpenCode, Codex, Qwen, Kimi and Grok, and for Codex also `--profile`/`-p` and `-c`/`--config` setting `model`, `model_provider` or `profile`) refuses a launch that also asks for a model, and two contributions that both set the model are refused, instead of starting the CLI with two. A person's own launches and restores start as before. Whatever a contribution adds is checked the same way for every launch: no permission, sandbox, hook or bypass flags, and no configuration the CLI reads that sets approvals (an OpenCode `OPENCODE_CONFIG` file or `OPENCODE_CONFIG_DIR`, `OPENCODE_PERMISSION`, Kimi's `--config`/`--config-file`); a file CanvasTTY cannot read and check is refused. While a child's launch waits for its plugins (launch options, a launch policy, an environment), `spawn_agent` answers only after its `prompt` reached the started agent, and `send_to_agent` waits the same way. A refused, failed or cancelled launch fails the call with the reason and the session id (the card stays); the text is dropped, never kept for a later restart. The control CLI answers `NOT_READY` for such a card. Before the agent starts, the host sends the service a `canvastty.launch.prepare` request, which surfaces cannot send: @@ -222,6 +222,7 @@ Rules the host enforces, none of which is ever skipped: - Claude Code applies only its last `--settings`, so a plugin's inline `--settings` JSON is merged into CanvasTTY's own (objects such as `env` key by key, hook lists appended); one that sets `permissions`, `hooks`, `disableAllHooks`, `sandbox`, `defaultMode` or `apiKeyHelper` is refused. Both `--settings ` and `--settings=` are checked; a settings file is accepted only as one of the contribution's own launch files (`{launchFiles}/…`), which CanvasTTY reads, checks the same way and passes inline; any other file path is refused. `--bare`, `--safe-mode`, `--allowedTools`, `--permission-prompt-tool` and `--permission-prompts` are CanvasTTY's too. - No answer within 5 s, an error, an invalid answer, a missing secret, or a plugin that is disabled, removed or no longer trusted refuses the launch with the reason on the card. The agent is never started without a contribution the person chose. A restored card whose plugin is unavailable comes back stopped with that reason and keeps its record until the plugin returns or the card is closed. - A plain terminal takes no launch options. +- An isolated agent cannot read any plugin's data folder (`/plugin-data`). If an `env` value sets the agent CLI's own home variable to an exact absolute folder inside the plugin's own data folder (not the data folder itself; a trailing `/` is ignored), that one folder is reopened for this agent. Only these variables count, each for its own CLI: `CODEX_HOME` (Codex), `CLAUDE_CONFIG_DIR` (Claude Code), `GROK_HOME` (Grok), `HERMES_HOME` (Hermes), `KIMI_HOME` (Kimi), `OPENCODE_CONFIG_DIR` (OpenCode) and `QWEN_HOME` (Qwen Code); never `HOME` or `XDG_*`. It must be a real folder, with no symbolic link on the way. Nothing else in plugin data is reopened, and a plugin cannot name other folders to reopen. **Launch profiles.** `profile` is `normal` (the default), `yolo`, or `auto`. `auto` exists only for agents whose CLI has a native auto mode, checked against each CLI's `--help`: Codex `--approve-for-me` (its own reviewer, in its `workspace-write` sandbox), Claude Code `--permission-mode auto` with its sandbox (`{"sandbox":{"enabled":true,"autoAllowBashIfSandboxed":false}}` merged into the one `--settings`), and Grok `--permission-mode auto` (no Grok sandbox is relied on). Base protection and decision services keep answering in front of it (Codex, Claude Code). When any contributor answers `thirdPartyModel: true`, `auto` becomes the CLI's accept-edits mode in the same sandbox (Codex `--sandbox workspace-write --ask-for-approval on-request`, Claude Code and Grok `--permission-mode acceptEdits`): the native reviewer would be that same model, and a weak model's own classifier is not a safety boundary. @@ -286,7 +287,7 @@ Before a local agent's shell command or file write runs, CanvasTTY can ask a plu }] ``` -`pre-tool` is every shell and file-writing tool call, before it runs and in every permission mode, YOLO included: Claude Code, Codex and Qwen Code through their `PreToolUse` hook, OpenCode through CanvasTTY's OpenCode plugin (`tool.execute.before`). `appliesTo` limits the agents; all four when omitted. The host sends `canvastty.decide` (host-only) and waits at most `timeoutMs` (1000 to 60000; 3000 when omitted). The agent's call waits as long, so ask for more only when the answer needs it (for example a local model reading the command); CanvasTTY sizes each card's hook for the longest budget of the services that apply when the card starts, and a service trusted later gets no more than its card allows. The request carries the budget as `budgetMs`: +`pre-tool` is every shell and file-writing tool call (read-only file tools reach only base protection, never a service, and only while base protection was on when the agent started), before it runs and in every permission mode, YOLO included: Claude Code, Codex and Qwen Code through their `PreToolUse` hook, OpenCode through CanvasTTY's OpenCode plugin (`tool.execute.before`). `appliesTo` limits the agents; all four when omitted. The host sends `canvastty.decide` (host-only) and waits at most `timeoutMs` (1000 to 60000; 3000 when omitted). The agent's call waits as long, so ask for more only when the answer needs it (for example a local model reading the command); CanvasTTY sizes each card's hook for the longest budget of the services that apply when the card starts, and a service trusted later gets no more than its card allows. The request carries the budget as `budgetMs`: ```ts interface DecisionRequest { @@ -316,7 +317,7 @@ The full example is [`examples/plugins/deny-rm`](../examples/plugins/deny-rm): i ### Agent tools (`tools:agents`) -A service may offer up to 16 `tools` to agents. They appear in the `canvastty_agents` MCP server as `__` (dots in the plugin id become `_`, so `com.example.tools` + `lookup` is `com_example_tools__lookup`; Anthropic and OpenAI allow only letters, digits, `_` and `-` in tool names, at most 64 characters, and a longer name keeps the start of the id plus a short hash), next to CanvasTTY's own orchestration tools, and count as CanvasTTY's own tools: base protection does not check them (it checks shells and file writes only). +A service may offer up to 16 `tools` to agents. They appear in the `canvastty_agents` MCP server as `__` (dots in the plugin id become `_`, so `com.example.tools` + `lookup` is `com_example_tools__lookup`; Anthropic and OpenAI allow only letters, digits, `_` and `-` in tool names, at most 64 characters, and a longer name keeps the start of the id plus a short hash), next to CanvasTTY's own orchestration tools, and count as CanvasTTY's own tools: base protection does not check them (it checks shells and file writes, and read-only file tools for its private-data rule only). ```json "permissions": ["tools:agents"], @@ -406,7 +407,9 @@ A service may run another browser engine for agents' background tabs, for exampl Two safety parts are built in and need no plugin: -- **Base protection** (Settings → Agents, on by default; the person can turn it off) denies, through the same hook, sudo and other elevation, piping downloaded or generated text into a shell, download-and-run, disk and format commands, fork bombs, and writing or deleting outside the working folder: the home folder, other projects and `/tmp` included, and deleting the working folder itself. An agent's own plan and memory folders (`~/.claude/plans`, `~/.claude/projects//memory`, and the same inside the run's `CLAUDE_CONFIG_DIR`) are not "outside". It also denies any use of CanvasTTY's own private data (from the app's userData folder: the agent-control token and descriptor, the gateways' connection records and sockets, the secret stores, account homes; and the control/runtime socket folders under the temporary folder), by any program, interpreter one-liners and socket clients included; the reason points the model at an **Orchestrator** launch and the `canvastty_agents` tools. The bundled control CLI may name its descriptor. It only ever denies; each reason tells the model what to do instead (a write to `/tmp` suggests a scratch folder inside the project). +- **Base protection** (Settings → Agents, on by default; the person can turn it off) denies, through the same hook, sudo and other elevation, piping downloaded or generated text into a shell, download-and-run, disk and format commands, fork bombs, and writing or deleting outside the working folder: the home folder, other projects and `/tmp` included, and deleting the working folder itself. An agent's own plan and memory folders (`~/.claude/plans`, `~/.claude/projects//memory`, and the same inside the run's `CLAUDE_CONFIG_DIR`) are not "outside". It also denies use of CanvasTTY's own private data (from the app's userData folder: the agent-control token and descriptor, the gateways' connection records and sockets, the secret stores, account homes, plugins' data (`plugin-data`) except the account home this card's launch plugin selected for it; and the control/runtime socket folders under the temporary folder), even when the working folder contains it (a project at the home folder): commands that name it, including interpreter one-liners and socket clients it recognises, file writes, and read-only file tools (Claude Code's Read, NotebookRead, Grep, Glob and LS; Qwen Code's and OpenCode's equivalents, checked for nothing else and only while base protection was on when the agent started); the reason points the model at an **Orchestrator** launch and the `canvastty_agents` tools. The bundled control CLI may name its descriptor. It only ever denies; each reason tells the model what to do instead (a write to `/tmp` suggests a scratch folder inside the project). + + Base protection is best effort: it reads each tool call before it runs, so it does not catch a path the call builds at run time (a command substitution, a variable it cannot resolve, a script that walks folders itself such as Python's `os.walk`), and it does not refuse a recursive search started above the app's data folder (`grep -r token ~`, a Grep of the home folder). A call that names more paths, longer paths or more nested shells than it checks is refused rather than checked in part, and so is a shell, file-edit or read call whose check fails with an error. A project set to the home folder, or to any folder that contains CanvasTTY's data, is therefore not a safe configuration on its own. Run such agents with agent isolation (sandbox-exec on macOS, bubblewrap on Linux): it hides the private data from every program in the agent's process tree, walks included. - **Secret redaction**: every text CanvasTTY hands from one agent to another (`observe_agent`, `wait_for_agent`, `get_agent_result`, the control CLI's `screen`, `result` and failure details) is masked: provider keys CanvasTTY holds, launch `secretEnv` values, values a service registered with `redaction.register`, also when the terminal wrapped them over lines, plus common key shapes (`sk-…`, GitHub, Slack, AWS, Google, JWT, `Bearer …`, `"apiKey": "…"`, PEM private keys, long random runs). Plugin tool answers, `screen` in session events, card badges and card action messages are masked the same way. host.onStorageChange(listener) notifies every live contribution of the same plugin — canvases, HOME widgets, and separate windows — of writes made through host.storage.set, avoiding polling when a plugin coordinates several surfaces. diff --git a/docs/plugins.ru.md b/docs/plugins.ru.md index 5d2cfe9d1..558a4695d 100644 --- a/docs/plugins.ru.md +++ b/docs/plugins.ru.md @@ -203,6 +203,7 @@ host.service.onEvent(({ serviceId, event, data }) => { /* … */ }); - Claude Code применяет только последний `--settings`, поэтому встроенный JSON `--settings` плагина сливается с собственным JSON CanvasTTY (объекты вроде `env` по ключам, списки хуков дописываются); если он задаёт `permissions`, `hooks`, `disableAllHooks`, `sandbox`, `defaultMode` или `apiKeyHelper`, запуск отклоняется. Проверяются обе формы: `--settings ` и `--settings=`; файл настроек принимается только как один из собственных файлов запуска вклада (`{launchFiles}/…`) — CanvasTTY читает его, проверяет так же и передаёт встроенным JSON; любой другой путь к файлу отклоняется. `--bare`, `--safe-mode`, `--allowedTools`, `--permission-prompt-tool` и `--permission-prompts` тоже передаёт только CanvasTTY. - Нет ответа за 5 с, ошибка, неверный ответ, отсутствующий секрет или выключенный, удалённый либо переставший быть доверенным плагин отклоняют запуск, и причина видна в окне. Агент никогда не запускается без выбранного вклада. Восстановленное окно с недоступным плагином возвращается остановленным с этой причиной и сохраняет запись, пока плагин не вернётся или окно не закроют. - Обычный терминал параметров запуска не принимает. +- Изолированный агент не может читать данные плагинов (`/plugin-data`). Если значение `env` задаёт собственную переменную домашней папки CLI агента точным абсолютным путём к папке внутри папки данных этого плагина (не к самой папке данных; завершающий `/` не учитывается), для этого агента открывается только эта папка. Учитываются только эти переменные, каждая для своего CLI: `CODEX_HOME` (Codex), `CLAUDE_CONFIG_DIR` (Claude Code), `GROK_HOME` (Grok), `HERMES_HOME` (Hermes), `KIMI_HOME` (Kimi), `OPENCODE_CONFIG_DIR` (OpenCode) и `QWEN_HOME` (Qwen Code); никогда `HOME` или `XDG_*`. Это должна быть настоящая папка без символических ссылок на пути. Больше ничего из данных плагинов не открывается, и плагин не может назвать другие папки для открытия. **Профили запуска.** `profile` — это `normal` (по умолчанию), `yolo` или `auto`. `auto` есть только у агентов, чей CLI имеет собственный авторежим, проверенный по `--help` каждого CLI: Codex `--approve-for-me` (его собственная проверка в песочнице `workspace-write`), Claude Code `--permission-mode auto` с его песочницей (`{"sandbox":{"enabled":true,"autoAllowBashIfSandboxed":false}}` сливается в единственный `--settings`) и Grok `--permission-mode auto` (на песочницу Grok не полагаемся). Базовая защита и сервисы решений отвечают перед ним (Codex, Claude Code). Если хотя бы один вклад ответил `thirdPartyModel: true`, `auto` превращается в режим «только правки» того же CLI в той же песочнице (Codex `--sandbox workspace-write --ask-for-approval on-request`, Claude Code и Grok `--permission-mode acceptEdits`): собственной проверкой была бы та же модель, а классификатор слабой модели — не граница безопасности. @@ -385,6 +386,8 @@ interface PluginSessionEvent { Две части безопасности встроены и не требуют плагина: - **Базовая защита** (Настройки → Агенты, включена по умолчанию; человек может её выключить) через тот же хук запрещает sudo и другое повышение прав, передачу скачанного или сгенерированного текста в shell, скачивание с запуском, команды для дисков и форматирования, форк-бомбы, а также запись и удаление вне рабочей папки — включая домашнюю папку, другие проекты и `/tmp` — и удаление самой рабочей папки. Собственные папки планов и памяти агента (`~/.claude/plans`, `~/.claude/projects//memory` и то же внутри `CLAUDE_CONFIG_DIR` запуска) не считаются «вне». Она только запрещает; каждая причина говорит модели, что сделать вместо этого (для записи в `/tmp` — завести временную папку внутри проекта). + + Она также запрещает использование собственных закрытых данных CanvasTTY (токен и описание agent-control, записи подключений и сокеты шлюзов, хранилища секретов, домашние папки аккаунтов, папки сокетов во временной папке), даже если рабочая папка их содержит: команды, которые называют эти пути, запись файлов и инструменты только для чтения (Read, NotebookRead, Grep, Glob и LS у Claude Code; аналоги у Qwen Code и OpenCode, только если базовая защита была включена при запуске агента). Эта проверка — по возможности: она читает вызов инструмента до запуска, поэтому не замечает путь, собранный во время выполнения (подстановка команды, неизвестная переменная, скрипт, который сам обходит папки, например `os.walk` в Python), и не запрещает рекурсивный поиск, начатый выше папки данных приложения (`grep -r token ~`, Grep по домашней папке). Вызов, в котором больше путей, более длинные пути или больше вложенных shell, чем она проверяет, запрещается, а не проверяется частично; так же запрещается вызов shell, правки или чтения файлов, проверка которого завершилась ошибкой. Поэтому проект в домашней папке или в любой папке, содержащей данные CanvasTTY, сам по себе небезопасен. Запускайте такие агенты в изоляции (sandbox-exec на macOS, bubblewrap на Linux): она скрывает закрытые данные от всех программ процесса агента, включая обход папок. - **Скрытие секретов**: весь текст, который CanvasTTY передаёт от одного агента другому (`observe_agent`, `wait_for_agent`, `get_agent_result`, `screen`, `result` и детали ошибок в control CLI), маскируется: ключи провайдеров, которые хранит CanvasTTY, значения `secretEnv` запуска, значения, зарегистрированные сервисом через `redaction.register`, в том числе перенесённые терминалом на несколько строк, а также типичные формы ключей (`sk-…`, GitHub, Slack, AWS, Google, JWT, `Bearer …`, `"apiKey": "…"`, приватные ключи PEM, длинные случайные строки). Так же маскируются ответы инструментов плагинов, `screen` в событиях сессий, метки на окнах и сообщения действий. host.onStorageChange(listener) сообщает всем открытым поверхностям того же плагина — canvas cards, HOME widgets и отдельным окнам — об изменениях через host.storage.set, поэтому нескольким поверхностям не требуется постоянный polling. diff --git a/docs/plugins.zh-CN.md b/docs/plugins.zh-CN.md index c91ff2c60..2b9db7948 100644 --- a/docs/plugins.zh-CN.md +++ b/docs/plugins.zh-CN.md @@ -201,6 +201,7 @@ host.service.onEvent(({ serviceId, event, data }) => { /* … */ }); - Claude Code 只应用最后一个 `--settings`，因此插件的内联 `--settings` JSON 会合并进 CanvasTTY 自己的 JSON（`env` 等对象按键合并，hook 列表追加）；若其中设置了 `permissions`、`hooks`、`disableAllHooks`、`sandbox`、`defaultMode` 或 `apiKeyHelper`，启动会被拒绝。`--settings ` 与 `--settings=` 两种形式都会检查；设置文件只接受贡献自身的启动文件（`{launchFiles}/…`），CanvasTTY 读取并以同样方式检查后以内联 JSON 传递；其他文件路径一律拒绝。`--bare`、`--safe-mode`、`--allowedTools`、`--permission-prompt-tool` 和 `--permission-prompts` 也只由 CanvasTTY 传递。 - 5 秒内无应答、出错、应答无效、缺少机密，或插件被禁用、删除或不再受信任，都会拒绝启动并在卡片上显示原因。智能体绝不会在缺少用户所选贡献的情况下启动。插件不可用的恢复卡片以停止状态返回并显示该原因，记录保留到插件恢复或卡片被关闭。 - 普通终端不接受启动选项。 +- 隔离的智能体无法读取任何插件的数据文件夹（`/plugin-data`）。如果某个 `env` 值把该智能体 CLI 自己的主目录变量设为该插件自身数据文件夹内的一个精确绝对路径文件夹（不能是数据文件夹本身；末尾的 `/` 会被忽略），就只为这个智能体重新开放这一个文件夹。只有以下变量算数，且各自只对应自己的 CLI：`CODEX_HOME`（Codex）、`CLAUDE_CONFIG_DIR`（Claude Code）、`GROK_HOME`（Grok）、`HERMES_HOME`（Hermes）、`KIMI_HOME`（Kimi）、`OPENCODE_CONFIG_DIR`（OpenCode）和 `QWEN_HOME`（Qwen Code）；`HOME` 和 `XDG_*` 永远不算。它必须是真实的文件夹，路径上不能有符号链接。插件数据中的其他内容都不会被重新开放，插件也不能指定其他要开放的文件夹。 **启动配置档。** `profile` 为 `normal`（默认）、`yolo` 或 `auto`。`auto` 仅适用于 CLI 自带自动模式的智能体，均按各 CLI 的 `--help` 核实：Codex `--approve-for-me`（其自身审查，运行于 `workspace-write` 沙箱），Claude Code `--permission-mode auto` 及其沙箱（`{"sandbox":{"enabled":true,"autoAllowBashIfSandboxed":false}}` 合并进唯一的 `--settings`），以及 Grok `--permission-mode auto`（不依赖 Grok 沙箱）。基础保护和决策服务仍在其前面作答（Codex、Claude Code）。只要任一贡献者回答 `thirdPartyModel: true`，`auto` 就改为同一沙箱中 CLI 的“仅接受编辑”模式（Codex `--sandbox workspace-write --ask-for-approval on-request`，Claude Code 和 Grok `--permission-mode acceptEdits`）：原生审查者将是同一个模型，而弱模型自己的分类器不是安全边界。 @@ -382,7 +383,9 @@ interface PluginSessionEvent { 两项安全功能内置，无需插件： -- **基础保护**（设置 → Agents → Base protection，默认开启；用户可以关闭）通过同一个 hook 拒绝：sudo 及其他提权、把下载或生成的文本管道给 shell、下载后直接运行、磁盘和格式化命令、fork 炸弹，以及在工作文件夹之外写入或删除（包括主目录、其他项目和 `/tmp`），以及删除工作文件夹本身。agent 自己的计划和记忆文件夹（`~/.claude/plans`、`~/.claude/projects//memory`，以及本次运行 `CLAUDE_CONFIG_DIR` 中的相同位置）不算"外部"。它还会拒绝任何程序（包括解释器单行命令和套接字客户端）使用 CanvasTTY 自己的私有数据（来自应用的 userData 文件夹：agent-control 令牌与描述文件、各网关的连接记录与套接字、密钥存储、账户主目录；以及临时文件夹下的控制/运行时套接字文件夹）；拒绝原因会引导模型改用 **Orchestrator** 启动和 `canvastty_agents` 工具。内置控制 CLI 可以引用它的描述文件。它只会拒绝；每条原因都告诉模型应当改做什么（写入 `/tmp` 时建议在项目内建立临时文件夹）。 +- **基础保护**（设置 → Agents → Base protection，默认开启；用户可以关闭）通过同一个 hook 拒绝：sudo 及其他提权、把下载或生成的文本管道给 shell、下载后直接运行、磁盘和格式化命令、fork 炸弹，以及在工作文件夹之外写入或删除（包括主目录、其他项目和 `/tmp`），以及删除工作文件夹本身。agent 自己的计划和记忆文件夹（`~/.claude/plans`、`~/.claude/projects//memory`，以及本次运行 `CLAUDE_CONFIG_DIR` 中的相同位置）不算"外部"。它还会拒绝使用 CanvasTTY 自己的私有数据（来自应用的 userData 文件夹：agent-control 令牌与描述文件、各网关的连接记录与套接字、密钥存储、账户主目录；以及临时文件夹下的控制/运行时套接字文件夹），即使工作文件夹包含它们（例如项目就是主目录）：包括点名这些路径的命令（含它能识别的解释器单行命令和套接字客户端）、文件写入，以及只读文件工具（Claude Code 的 Read、NotebookRead、Grep、Glob、LS，以及 Qwen Code 和 OpenCode 的对应工具；只读工具不检查其他内容，且仅在 agent 启动时基础保护已开启的情况下才检查）；拒绝原因会引导模型改用 **Orchestrator** 启动和 `canvastty_agents` 工具。内置控制 CLI 可以引用它的描述文件。它只会拒绝；每条原因都告诉模型应当改做什么（写入 `/tmp` 时建议在项目内建立临时文件夹）。 + + 基础保护是尽力而为的检查：它在工具调用运行前读取调用内容，因此无法发现运行时才拼出的路径（命令替换、无法解析的变量、像 Python `os.walk` 这样自行遍历文件夹的脚本），也不会拒绝从应用数据文件夹上层开始的递归搜索（`grep -r token ~`、对主目录的 Grep）。如果一次调用包含的路径数量、路径长度或嵌套 shell 层数超过它的检查范围，该调用会被拒绝，而不是只检查一部分；检查过程出错的 shell、文件编辑或读取调用同样会被拒绝。因此，把项目设为主目录或任何包含 CanvasTTY 数据的文件夹，本身并不安全。请在 agent 隔离下运行这类 agent（macOS 上的 sandbox-exec，Linux 上的 bubblewrap）：它会对 agent 进程树中的所有程序隐藏私有数据，包括遍历。 - **密钥遮蔽**：CanvasTTY 从一个 agent 交给另一个 agent 的所有文本（`observe_agent`、`wait_for_agent`、`get_agent_result`，以及 control CLI 的 `screen`、`result` 和失败详情）都会被遮蔽：CanvasTTY 保存的服务商密钥、启动时的 `secretEnv` 值、服务通过 `redaction.register` 注册的值（包括被终端折行拆开的情况），以及常见密钥形式（`sk-…`、GitHub、Slack、AWS、Google、JWT、`Bearer …`、`"apiKey": "…"`、PEM 私钥、长随机串）。插件工具的回答、会话事件中的 `screen`、卡片标记和卡片动作消息也以同样方式遮蔽。 host.onStorageChange(listener) 会把 host.storage.set 的写入通知给同一插件的所有活动界面——画布卡片、HOME 小组件和独立窗口——从而避免轮询。 diff --git a/native/canvastty-helper/catalog.json b/native/canvastty-helper/catalog.json index d3ec39a27..ab71c4250 100644 --- a/native/canvastty-helper/catalog.json +++ b/native/canvastty-helper/catalog.json @@ -964,7 +964,7 @@ }, { "name": "spawn_agent", - "description": "Launch another provider's agent as a CanvasTTY subagent of this session and optionally deliver a first prompt. Returns the new session id. provider must be an id from list_providers (known ids: codex, claude, qwen, kimi, opencode, hermes, grok, omp, pi, cursor, minimax, devin, antigravity); call list_providers first to see which are installed and signed in. Give each subagent one self-contained part of the task and an absolute cwd. If the person names a model, pass it as model in the format list_providers gives for that provider (OpenCode: provider/model); effort sets the reasoning effort where that CLI has one (list_providers shows its efforts). An unsupported model or effort is refused with the reason. profile: without it the subagent gets this session's launch profile, or the next lower one its CLI has; a subagent never gets more than this session (plan < normal < acceptEdits < auto) and never YOLO. \"auto\" lets it work without asking the person for each edit or command inside the project; CanvasTTY's outer layers (agent isolation, base protection) still stop writes outside the project, secrets and system changes. The answer says the profile it got. cwd must be this project's folder or a folder inside it. launchOptions passes plugin launch options exactly as a plugin tool gives them (for example the account a plugin picked), for plugins that allow an orchestrator to choose them. Refusals say why (a limit the person set, a folder outside the project, a profile above this session's): change the request instead of retrying it. Then call wait_for_agent and get_agent_result.", + "description": "Launch another provider's agent as a CanvasTTY subagent of this session and optionally deliver a first prompt. Returns the new session id. provider must be an id from list_providers (known ids: codex, claude, qwen, kimi, opencode, hermes, grok, omp, pi, cursor, minimax, devin, antigravity); call list_providers first to see which are installed and signed in. Give each subagent one self-contained part of the task and an absolute cwd. If the person names a model, pass it as model in the format list_providers gives for that provider (OpenCode: provider/model); effort sets the reasoning effort where that CLI has one (list_providers shows its efforts). An unsupported model or effort is refused with the reason. profile: without it the subagent gets this session's launch profile, or the next lower one its CLI has; a subagent never gets more than this session (plan < normal < acceptEdits < auto) and never YOLO. \"auto\" lets it work without asking the person for each edit or command inside the project; CanvasTTY's outer layers (agent isolation, base protection) still stop writes outside the project, secrets and system changes. The answer says the profile it got. cwd must be this project's folder or a folder inside it. launchOptions passes plugin launch options exactly as a plugin tool gives them (for example the account a plugin picked), for plugins that allow an orchestrator to choose them. A plugin this call does not name keeps this session's own launch options (for example its model account), so the subagent never falls back silently to its CLI's default model; a subagent of another CLI is refused while this session's model comes from a plugin's options (spawn one of this CLI instead), and model next to inherited options that set the model is refused. The answer's servedBy says which CLI and requested model serve it and where each plugin's options came from. Refusals say why (a limit the person set, a folder outside the project, a profile above this session's): change the request instead of retrying it. Then call wait_for_agent and get_agent_result.", "inputSchema": { "type": "object", "properties": { @@ -1072,7 +1072,7 @@ }, { "name": "observe_agent", - "description": "Read the capped terminal tail and status of one of this session's subagents.", + "description": "Read the status of one of this session's subagents and the end of its screen as masked plain text (once its process exited, also exitLines).", "inputSchema": { "type": "object", "properties": { @@ -1095,7 +1095,7 @@ }, { "name": "wait_for_agent", - "description": "Wait until one of this session's subagents stops working, instead of polling observe_agent or get_agent_result; nothing is sent to it while it waits. Returns reason \"idle\" (the turn that answers your latest prompt ended and it waits for input; an idle before that turn started does not count), \"needs_approval\" (its card shows a prompt only the person may answer; never answer it yourself), \"done\" or \"failed\" (its process exited), \"quiet\" (it reports no status or no turn start and its screen stopped changing, so judge from output), \"closed\" (its card was closed) or \"timeout\" after timeoutSeconds (default 55, at most 600), with status, exitCode, waitedMs, the masked terminal tail as output, and answer (the final reply of the turn that ended, for Codex and OpenCode subagents); when the subagent's process exited (for example at once, on a model its CLI does not know), exitLines holds the last lines of its screen as plain text, which say why. After a timeout, call it again. Then read get_agent_result.", + "description": "Wait until one of this session's subagents stops working, instead of polling observe_agent or get_agent_result; nothing is sent to it while it waits. Returns reason \"idle\" (the turn that answers your latest prompt ended and it waits for input; an idle before that turn started does not count), \"needs_approval\" (its card shows a prompt only the person may answer; never answer it yourself), \"done\" or \"failed\" (its process exited), \"quiet\" (it reports no status or no turn start and its screen stopped changing, so judge from output), \"closed\" (its card was closed) or \"timeout\" after timeoutSeconds (default 50, at most 100; unless this session is Claude Code, a call returns after at most 50 seconds, before its MCP client would end the call), with status, exitCode, waitedMs, the end of its screen as masked plain text as output, and answer (the final reply of the turn that ended, for Codex and OpenCode subagents; when CanvasTTY status hooks are off it cannot be captured and answerUnavailable says so instead: then read output); when the subagent's process exited (for example at once, on a model its CLI does not know), exitLines holds the last lines of its screen as plain text, which say why. A timeout means it is still running: read output as its progress and call wait_for_agent again. Then read get_agent_result.", "inputSchema": { "type": "object", "properties": { @@ -1107,7 +1107,7 @@ "timeoutSeconds": { "type": "integer", "minimum": 1, - "maximum": 600 + "maximum": 100 } }, "required": [ @@ -1118,7 +1118,7 @@ }, { "name": "get_agent_result", - "description": "Get one of this session's subagents' result: answer (its last turn's final reply as the agent reported it, for Codex and OpenCode subagents; truncated:true when only the end was kept), status (idle once its turn ended), the exit state (running | done | failed; an interactive CLI stays running after a task) and the masked terminal tail as output. Prefer answer; the terminal tail is raw screen output.", + "description": "Get one of this session's subagents' result: answer (its last turn's final reply as the agent reported it, for Codex and OpenCode subagents; truncated:true when only the end was kept), status (idle once its turn ended), the exit state (running | done | failed; an interactive CLI stays running after a task) and the end of its screen as masked plain text as output (what the terminal shows, without control sequences). Prefer answer. answerUnavailable (reason status_hooks_off) means no answer can be captured because CanvasTTY status hooks are off: read output instead.", "inputSchema": { "type": "object", "properties": { @@ -1178,6 +1178,6 @@ ], "maxPluginToolNameLength": 64, "maxLaunchOptionsBytes": 16384, - "defaultAgentWaitSeconds": 55 + "defaultAgentWaitSeconds": 50 } } diff --git a/src/agent-browser/orchestration-catalog.d.mts b/src/agent-browser/orchestration-catalog.d.mts index 48685eb78..1af1e815c 100644 --- a/src/agent-browser/orchestration-catalog.d.mts +++ b/src/agent-browser/orchestration-catalog.d.mts @@ -3,6 +3,7 @@ export const MAX_ORCHESTRATION_PAYLOAD_BYTES: number; export const AGENT_PROVIDER_IDS: readonly string[]; export const MAX_AGENT_WAIT_SECONDS: number; export const DEFAULT_AGENT_WAIT_SECONDS: number; +export const CLIENT_SAFE_AGENT_WAIT_SECONDS: number; export const REASONING_EFFORT_IDS: readonly string[]; export function unknownProviderMessage(value: unknown): string; diff --git a/src/agent-browser/orchestration-catalog.mjs b/src/agent-browser/orchestration-catalog.mjs index ad7b53d22..3829b3f62 100644 --- a/src/agent-browser/orchestration-catalog.mjs +++ b/src/agent-browser/orchestration-catalog.mjs @@ -28,10 +28,13 @@ const prompt = string({ minLength: 1, maxLength: 65_536 }); export const AGENT_PROVIDER_IDS = Object.freeze([ "codex", "claude", "qwen", "kimi", "opencode", "hermes", "grok", "omp", "pi", "cursor", "minimax", "devin", "antigravity" ]); -/** wait_for_agent: the longest wait one call may ask for, and the wait without timeoutSeconds (some MCP clients - * end a tool call after 60 seconds). */ -export const MAX_AGENT_WAIT_SECONDS = 600; -export const DEFAULT_AGENT_WAIT_SECONDS = 55; +/** wait_for_agent: the longest wait one call may ask for, and the wait without timeoutSeconds. Most MCP clients end a + * tool call after 60 seconds (Codex's default tool_timeout_sec, OpenCode's client), so a call from any orchestrator + * but Claude Code (whose client waits far longer) returns after at most CLIENT_SAFE_AGENT_WAIT_SECONDS: a "timeout" + * that says the subagent still runs, never a call its client cut off. */ +export const MAX_AGENT_WAIT_SECONDS = 100; +export const DEFAULT_AGENT_WAIT_SECONDS = 50; +export const CLIENT_SAFE_AGENT_WAIT_SECONDS = 50; /** spawn_agent.effort: every level some CLI takes (src/shared/launchModel.ts REASONING_EFFORTS; a test keeps them equal). */ export const REASONING_EFFORT_IDS = Object.freeze(["minimal", "low", "medium", "high", "xhigh", "max"]); const provider = string({ minLength: 1, maxLength: 32, enum: [...AGENT_PROVIDER_IDS] }); @@ -58,7 +61,7 @@ export const ORCHESTRATION_TOOL_DEFINITIONS = Object.freeze([ ), tool( "spawn_agent", - `Launch another provider's agent as a CanvasTTY subagent of this session and optionally deliver a first prompt. Returns the new session id. provider must be an id from list_providers (known ids: ${AGENT_PROVIDER_IDS.join(", ")}); call list_providers first to see which are installed and signed in. Give each subagent one self-contained part of the task and an absolute cwd. If the person names a model, pass it as model in the format list_providers gives for that provider (OpenCode: provider/model); effort sets the reasoning effort where that CLI has one (list_providers shows its efforts). An unsupported model or effort is refused with the reason. profile: without it the subagent gets this session's launch profile, or the next lower one its CLI has; a subagent never gets more than this session (plan < normal < acceptEdits < auto) and never YOLO. "auto" lets it work without asking the person for each edit or command inside the project; CanvasTTY's outer layers (agent isolation, base protection) still stop writes outside the project, secrets and system changes. The answer says the profile it got. cwd must be this project's folder or a folder inside it. launchOptions passes plugin launch options exactly as a plugin tool gives them (for example the account a plugin picked), for plugins that allow an orchestrator to choose them. Refusals say why (a limit the person set, a folder outside the project, a profile above this session's): change the request instead of retrying it. Then call wait_for_agent and get_agent_result.`, + `Launch another provider's agent as a CanvasTTY subagent of this session and optionally deliver a first prompt. Returns the new session id. provider must be an id from list_providers (known ids: ${AGENT_PROVIDER_IDS.join(", ")}); call list_providers first to see which are installed and signed in. Give each subagent one self-contained part of the task and an absolute cwd. If the person names a model, pass it as model in the format list_providers gives for that provider (OpenCode: provider/model); effort sets the reasoning effort where that CLI has one (list_providers shows its efforts). An unsupported model or effort is refused with the reason. profile: without it the subagent gets this session's launch profile, or the next lower one its CLI has; a subagent never gets more than this session (plan < normal < acceptEdits < auto) and never YOLO. "auto" lets it work without asking the person for each edit or command inside the project; CanvasTTY's outer layers (agent isolation, base protection) still stop writes outside the project, secrets and system changes. The answer says the profile it got. cwd must be this project's folder or a folder inside it. launchOptions passes plugin launch options exactly as a plugin tool gives them (for example the account a plugin picked), for plugins that allow an orchestrator to choose them. A plugin this call does not name keeps this session's own launch options (for example its model account), so the subagent never falls back silently to its CLI's default model; a subagent of another CLI is refused while this session's model comes from a plugin's options (spawn one of this CLI instead), and model next to inherited options that set the model is refused. The answer's servedBy says which CLI and requested model serve it and where each plugin's options came from. Refusals say why (a limit the person set, a folder outside the project, a profile above this session's): change the request instead of retrying it. Then call wait_for_agent and get_agent_result.`, { provider, cwd: string({ minLength: 1, maxLength: 4_096 }), @@ -79,19 +82,19 @@ export const ORCHESTRATION_TOOL_DEFINITIONS = Object.freeze([ ), tool( "observe_agent", - "Read the capped terminal tail and status of one of this session's subagents.", + "Read the status of one of this session's subagents and the end of its screen as masked plain text (once its process exited, also exitLines).", { sessionId, maxChars: integer({ minimum: 256, maximum: 8_192 }) }, ["sessionId"] ), tool( "wait_for_agent", - `Wait until one of this session's subagents stops working, instead of polling observe_agent or get_agent_result; nothing is sent to it while it waits. Returns reason "idle" (the turn that answers your latest prompt ended and it waits for input; an idle before that turn started does not count), "needs_approval" (its card shows a prompt only the person may answer; never answer it yourself), "done" or "failed" (its process exited), "quiet" (it reports no status or no turn start and its screen stopped changing, so judge from output), "closed" (its card was closed) or "timeout" after timeoutSeconds (default ${DEFAULT_AGENT_WAIT_SECONDS}, at most ${MAX_AGENT_WAIT_SECONDS}), with status, exitCode, waitedMs, the masked terminal tail as output, and answer (the final reply of the turn that ended, for Codex and OpenCode subagents); when the subagent's process exited (for example at once, on a model its CLI does not know), exitLines holds the last lines of its screen as plain text, which say why. After a timeout, call it again. Then read get_agent_result.`, + `Wait until one of this session's subagents stops working, instead of polling observe_agent or get_agent_result; nothing is sent to it while it waits. Returns reason "idle" (the turn that answers your latest prompt ended and it waits for input; an idle before that turn started does not count), "needs_approval" (its card shows a prompt only the person may answer; never answer it yourself), "done" or "failed" (its process exited), "quiet" (it reports no status or no turn start and its screen stopped changing, so judge from output), "closed" (its card was closed) or "timeout" after timeoutSeconds (default ${DEFAULT_AGENT_WAIT_SECONDS}, at most ${MAX_AGENT_WAIT_SECONDS}; unless this session is Claude Code, a call returns after at most ${CLIENT_SAFE_AGENT_WAIT_SECONDS} seconds, before its MCP client would end the call), with status, exitCode, waitedMs, the end of its screen as masked plain text as output, and answer (the final reply of the turn that ended, for Codex and OpenCode subagents; when CanvasTTY status hooks are off it cannot be captured and answerUnavailable says so instead: then read output); when the subagent's process exited (for example at once, on a model its CLI does not know), exitLines holds the last lines of its screen as plain text, which say why. A timeout means it is still running: read output as its progress and call wait_for_agent again. Then read get_agent_result.`, { sessionId, timeoutSeconds: integer({ minimum: 1, maximum: MAX_AGENT_WAIT_SECONDS }) }, ["sessionId"] ), tool( "get_agent_result", - "Get one of this session's subagents' result: answer (its last turn's final reply as the agent reported it, for Codex and OpenCode subagents; truncated:true when only the end was kept), status (idle once its turn ended), the exit state (running | done | failed; an interactive CLI stays running after a task) and the masked terminal tail as output. Prefer answer; the terminal tail is raw screen output.", + "Get one of this session's subagents' result: answer (its last turn's final reply as the agent reported it, for Codex and OpenCode subagents; truncated:true when only the end was kept), status (idle once its turn ended), the exit state (running | done | failed; an interactive CLI stays running after a task) and the end of its screen as masked plain text as output (what the terminal shows, without control sequences). Prefer answer. answerUnavailable (reason status_hooks_off) means no answer can be captured because CanvasTTY status hooks are off: read output instead.", { sessionId }, ["sessionId"] ), diff --git a/src/agent-runtime/opencode-decisions.mjs b/src/agent-runtime/opencode-decisions.mjs index d9e9c4ca3..3915290c6 100644 --- a/src/agent-runtime/opencode-decisions.mjs +++ b/src/agent-runtime/opencode-decisions.mjs @@ -1,12 +1,13 @@ import { FAIL_CLOSED_MESSAGE, buildRequest, exchange, identityFrom } from "./permission-gate.mjs"; -import { OPENCODE_DECISIONS_ENV, helperDeadlineMs } from "./runtime-protocol.mjs"; +import { OPENCODE_DECISIONS_ENV, OPENCODE_DECISION_READS_ENV, helperDeadlineMs } from "./runtime-protocol.mjs"; /** * Decision hooks for OpenCode, used by opencode-plugin.mjs while `CANVASTTY_RUNTIME_DECISIONS` is "1". * * `guard` runs in `tool.execute.before` (before OpenCode's own permission check, in every mode, YOLO included) * and sends each shell and file-writing call (bash, write, edit, multiedit, apply_patch/patch) to CanvasTTY the way - * the Claude/Codex/Qwen PreToolUse hook does. A deny throws, which fails that tool call with the reason as its error + * the Claude/Codex/Qwen PreToolUse hook does; with `CANVASTTY_RUNTIME_DECISION_READS` "1" (base protection on at + * launch) also read, grep, glob and list, for base protection's private-data rule. A deny throws, which fails that tool call with the reason as its error * text for the model. An allow (only from plugins the person let allow) is remembered by call id: when OpenCode then * asks for that call (`permission.asked`), `permissionAsked` answers `once` through OpenCode's own API * (POST /permission/{requestID}/reply), never `always`. Anything else (ask, no verdict, an error) leaves OpenCode's @@ -23,6 +24,9 @@ const MAX_ALLOWED = 64; export function createOpenCodeDecisions({ client, env = process.env, send = exchange } = {}) { const identity = identityFrom(env); const enabled = env[OPENCODE_DECISIONS_ENV] === "1" && identity?.provider === "opencode"; + // Read-only file tools are sent only when the launch asked for it (base protection was on): otherwise a read never + // waits on, or fails closed because of, the gateway. + const reads = enabled && env[OPENCODE_DECISION_READS_ENV] === "1"; /** callIDs a plugin allowed; their permission request is answered `once`. */ const allowed = new Set(); @@ -30,7 +34,7 @@ export function createOpenCodeDecisions({ client, env = process.env, send = exch if (!enabled) return; let decision = null; try { - const call = guardedCall(stringOf(input?.tool), output && typeof output === "object" ? output.args : null); + const call = guardedCall(stringOf(input?.tool), output && typeof output === "object" ? output.args : null, reads); if (!call) return; const message = buildRequest({ tool_name: call.toolName, tool_input: call.toolInput }, identity); if (message) decision = await send(identity.address, message, helperDeadlineMs(env)); @@ -63,10 +67,11 @@ export function createOpenCodeDecisions({ client, env = process.env, send = exch /** * An OpenCode tool call in the shape the decision hook reads (the Claude/Codex hook tool names), or null for a tool - * that neither runs commands nor writes files. OpenCode's own names: bash {command, workdir}, write {filePath, - * content}, edit {filePath, oldString, newString}, multiedit {filePath, edits}, apply_patch {patchText}. + * that neither runs commands, writes files nor reads them. OpenCode's own names: bash {command, workdir}, write {filePath, + * content}, edit {filePath, oldString, newString}, multiedit {filePath, edits}, apply_patch {patchText}. Read-only + * tools (read, grep, glob, list) are mapped only when `reads` is set. */ -export function guardedCall(tool, args) { +export function guardedCall(tool, args, reads = false) { if (!tool || !args || typeof args !== "object") return null; if (tool === "bash") { if (typeof args.command !== "string" || args.command.length === 0) return null; @@ -82,6 +87,20 @@ export function guardedCall(tool, args) { : multieditText(args.edits); return { toolName: "edit", toolInput: { file_path: filePath, ...(content !== null ? { content } : {}) } }; } + // Read-only tools, only when the launch checks them (base protection on): only CanvasTTY's private data is checked + // (read {filePath}, grep {pattern, path, include}, glob {pattern, path}, list {path}). A grep without a path + // searches the current folder, so it is always sent, as `.`. + if (reads && (tool === "read" || tool === "grep" || tool === "glob" || tool === "list")) { + const text = (value) => typeof value === "string" && value.length > 0 ? value : null; + const toolInput = {}; + const filePath = text(args.filePath) ?? text(args.file_path); + if (filePath) toolInput.file_path = filePath; + const path = text(args.path) ?? (tool === "grep" ? "." : null); + if (path) toolInput.path = path; + if (tool === "glob" && text(args.pattern)) toolInput.pattern = args.pattern; + if (Object.keys(toolInput).length === 0) return null; + return { toolName: { read: "Read", grep: "Grep", glob: "Glob", list: "LS" }[tool], toolInput }; + } if (tool === "apply_patch" || tool === "patch") { const patch = typeof args.patchText === "string" ? args.patchText : typeof args.patch === "string" ? args.patch : null; return patch ? { toolName: "apply_patch", toolInput: { patch } } : null; diff --git a/src/agent-runtime/opencode-plugin.mjs b/src/agent-runtime/opencode-plugin.mjs index bd8151b86..69ae5f00c 100644 --- a/src/agent-runtime/opencode-plugin.mjs +++ b/src/agent-runtime/opencode-plugin.mjs @@ -3,10 +3,13 @@ import { CAPTURE_RESULT_ENV } from "./runtime-protocol.mjs"; import { finalAnswer } from "./opencode-final-answer.mjs"; import { createOpenCodeDecisions } from "./opencode-decisions.mjs"; import { spawn } from "node:child_process"; +import { randomUUID } from "node:crypto"; import { HOOK_TIMEOUT_MS, preparePluginHook } from "./plugin-hook-dispatch.mjs"; let rootSessionId = null; let rootWorking = false; +// One id per root turn (not the session id), so a late answer of an earlier turn is never taken for the current one. +let rootTurnId = null; const lifecycleEnabled = process.env.CANVASTTY_LIFECYCLE_HOOKS_ENABLED !== "0"; const pluginHookRegistry = process.env.CANVASTTY_PLUGIN_HOOK_REGISTRY ?? ""; const pluginHookRunnerCommand = process.env.CANVASTTY_PLUGIN_HOOK_RUNNER_COMMAND ?? ""; @@ -45,6 +48,7 @@ async function lifecycleEvent(event, decisions, client) { if (session.parentID || session.parentId) return; rootSessionId = stringField(session.id, sessionId); rootWorking = false; + rootTurnId = null; if (!rootSessionId) return; if (lifecycleEnabled) { await reportLifecycle({ state: "idle", event: event.type, threadId: rootSessionId }); @@ -63,44 +67,55 @@ async function lifecycleEvent(event, decisions, client) { ? statusValue.type : null; if (status === "busy" || status === "retry") { + const startsTurn = status === "busy" && !rootWorking; + if (startsTurn || !rootTurnId) rootTurnId = randomUUID(); + rootWorking = true; if (lifecycleEnabled) { - await reportLifecycle({ state: "working", event: `session.status:${status}`, turnId: rootSessionId }); + await reportLifecycle({ state: "working", event: `session.status:${status}`, turnId: rootTurnId }); } - if (status === "busy" && !rootWorking) { + if (startsTurn) { runPluginHooks("prompt-submit", `session.status:${status}`, event); } - rootWorking = true; } else if (status === "idle") { rootWorking = false; - if (lifecycleEnabled) await reportLifecycle({ state: "idle", event: "session.status:idle", turnId: rootSessionId }); + // A result-capturing turn ends with session.idle, once its final answer was read: an idle reported here would + // let a waiting orchestrator take the turn as finished before its answer arrived. + if (lifecycleEnabled && !captureResult) await reportLifecycle({ state: "idle", event: "session.status:idle", turnId: rootTurnId }); } return; } if (event.type === "session.idle") { rootWorking = false; - const result = captureResult ? await finalAnswer(client, rootSessionId) : undefined; - if (lifecycleEnabled || result) { - await reportLifecycle({ state: "idle", event: event.type, turnId: rootSessionId, ...(result ? { result } : {}) }); + const endingSessionId = rootSessionId, endingTurnId = rootTurnId; + // With CanvasTTY status hooks off nothing is reported, the answer included: it is never read. + const result = lifecycleEnabled && captureResult && endingTurnId ? await finalAnswer(client, endingSessionId) : undefined; + // The read is asynchronous: when the next turn started meanwhile, this answer and this idle are no longer current. + const superseded = endingSessionId !== rootSessionId || endingTurnId !== rootTurnId || rootWorking; + if (lifecycleEnabled && !superseded) { + await reportLifecycle({ state: "idle", event: event.type, turnId: endingTurnId, ...(result ? { result } : {}) }); } runPluginHooks("stop", event.type, event); } else if (event.type === "permission.asked") { - if (lifecycleEnabled) await reportLifecycle({ state: "needs_approval", event: event.type, turnId: rootSessionId }); + if (lifecycleEnabled) await reportLifecycle({ state: "needs_approval", event: event.type, turnId: rootTurnId }); runPluginHooks("permission-request", event.type, event); } else if (event.type === "permission.replied") { - if (lifecycleEnabled) await reportLifecycle({ state: "working", event: event.type, turnId: rootSessionId }); + rootWorking = true; + if (lifecycleEnabled) await reportLifecycle({ state: "working", event: event.type, turnId: rootTurnId }); runPluginHooks("permission-result", event.type, event); } else if (event.type === "question.asked") { - if (lifecycleEnabled) await reportLifecycle({ state: "needs_approval", event: event.type, turnId: rootSessionId }); + if (lifecycleEnabled) await reportLifecycle({ state: "needs_approval", event: event.type, turnId: rootTurnId }); } else if (event.type === "question.replied" || event.type === "question.rejected") { - if (lifecycleEnabled) await reportLifecycle({ state: "working", event: event.type, turnId: rootSessionId }); + rootWorking = true; + if (lifecycleEnabled) await reportLifecycle({ state: "working", event: event.type, turnId: rootTurnId }); } else if (event.type === "session.error") { rootWorking = false; - if (lifecycleEnabled) await reportLifecycle({ state: "idle", event: event.type, turnId: rootSessionId }); + if (lifecycleEnabled) await reportLifecycle({ state: "idle", event: event.type, turnId: rootTurnId }); runPluginHooks("stop", event.type, event); } else if (event.type === "session.deleted") { runPluginHooks("session-end", event.type, event); rootWorking = false; rootSessionId = null; + rootTurnId = null; } else if (event.type === "tool.execute.after") { runPluginHooks("after-tool", event.type, event); } diff --git a/src/agent-runtime/runtime-protocol.d.mts b/src/agent-runtime/runtime-protocol.d.mts index cdf16b466..833491282 100644 --- a/src/agent-runtime/runtime-protocol.d.mts +++ b/src/agent-runtime/runtime-protocol.d.mts @@ -25,6 +25,7 @@ export const PERMISSION_GATE: Readonly<{ messageChars: number; }>; export const OPENCODE_DECISIONS_ENV: "CANVASTTY_RUNTIME_DECISIONS"; +export const OPENCODE_DECISION_READS_ENV: "CANVASTTY_RUNTIME_DECISION_READS"; export const DECISION_BUDGET_ENV: "CANVASTTY_RUNTIME_DECISION_MS"; export const DECISION_FAIL_CLOSED_ENV: "CANVASTTY_RUNTIME_FAIL_CLOSED"; export const DEFAULT_DECIDE_TIMEOUT_MS: number; diff --git a/src/agent-runtime/runtime-protocol.mjs b/src/agent-runtime/runtime-protocol.mjs index 4aad87536..ff1beb369 100644 --- a/src/agent-runtime/runtime-protocol.mjs +++ b/src/agent-runtime/runtime-protocol.mjs @@ -59,6 +59,8 @@ export const PERMISSION_GATE = Object.freeze({ }); // Set to "1" for an OpenCode session launched with decision hooks: its plugin then checks each shell and file call. export const OPENCODE_DECISIONS_ENV = "CANVASTTY_RUNTIME_DECISIONS"; +// Set to "1" as well when base protection was on at launch: the plugin then also checks read-only file tools. +export const OPENCODE_DECISION_READS_ENV = "CANVASTTY_RUNTIME_DECISION_READS"; // A decision service may ask for more time than the default 3 s (`decide.timeoutMs`, 1-60 s), for example to ask a // local model. The session's hook, helper and gateway deadlines are set at launch from the longest such budget and // passed to the helper in this variable; without it the defaults above hold. diff --git a/src/main/index.ts b/src/main/index.ts index b9832dcee..f5d354a68 100644 --- a/src/main/index.ts +++ b/src/main/index.ts @@ -36,6 +36,7 @@ import { AgentChatHistoryService } from "./services/AgentChatHistoryService"; import { LimitsService, withoutAccountScope } from "./services/LimitsService"; import { UsageHistoryService } from "./services/UsageHistoryService"; import { LocalUsageCollector } from "./services/LocalUsageCollector"; +import { runShutdownSteps } from "./services/shutdownSteps"; import { createProviderCliRegistry, providerCliAvailability, @@ -91,6 +92,7 @@ import { RuntimeGateway } from "./services/agent-runtime"; import type { RuntimeHookHelperLaunch } from "./services/agent-runtime/ProviderRuntimeLaunch"; +import { applyRuntimeSignal } from "./services/agent-runtime/applyRuntimeSignal"; import { agentHelperLaunches } from "./services/agentHelpers"; import { recoverHermesConfigurationOnStartup, @@ -529,15 +531,7 @@ async function initializeServices(ipc: IpcRegistrar): Promise { runtimeDirectory: lifecycleRuntimeDirectory, windowsHostPath, onSignal: (terminalSessionId, signal) => { - terminalManager?.applyProviderSignal(terminalSessionId, { - kind: "lifecycle", - state: signal.state, - event: signal.event, - ...(signal.turnId ? { requestId: signal.turnId } : {}), - ...(signal.threadId ? { threadId: signal.threadId } : {}) - }); - // A subagent's final answer (Codex Stop hook, OpenCode plugin) for get_agent_result and wait_for_agent. - if (signal.result) terminalManager?.recordAnswer(terminalSessionId, signal.result); + if (terminalManager) applyRuntimeSignal(terminalManager, terminalSessionId, signal); agentControl?.onSignal(terminalSessionId, signal); if (signal.lastAssistantMessage !== undefined && signal.answerCaptureGrantExpiresAt !== undefined) { evenG2?.answer( @@ -573,6 +567,7 @@ async function initializeServices(ipc: IpcRegistrar): Promise { permissionGate: helpers.permissionGate, wantsDecisions: (provider) => decisionHooks.wanted(provider), decisionBudgetMs: (provider) => decisionHooks.budgetMs(provider), + decisionReads: () => decisionHooks.checksReads(), claudeHttpHooks: (facts) => claudeHttpHookPolicy.verdict(facts), pluginHooks: { runner: { @@ -695,7 +690,8 @@ async function initializeServices(ipc: IpcRegistrar): Promise { const managedTerminals = terminalManager; const agentControlService = new AgentControlService(managedTerminals, { limits: () => ({ maxDepth: settings.get().orchestrationMaxDepth, maxSubagents: settings.get().orchestrationMaxSubagents }), - containment: () => managedTerminals.containment() + containment: () => managedTerminals.containment(), + userDataPath }); orchestrationGateway = new OrchestrationGateway({ runtimeDirectory: join(userDataPath, "orchestration", "runtime"), @@ -1275,30 +1271,36 @@ void IPC.terminalData; async function shutdownServices(): Promise { diagnostics.record("info", "application", "shutdown.started"); - mascotManager?.dispose(); - agentChatHistory?.dispose(); - if (agentControl) await Promise.allSettled([agentControl.close()]); if (updateTimer) clearTimeout(updateTimer); if (updateInterval) clearInterval(updateInterval); for (const request of browserRequests.values()) { clearTimeout(request.timer); request.reject(new Error("App closing")); } browserRequests.clear(); - await evenG2?.close(); - if (terminalManager) await terminalManager.shutdown(); + // Each step runs even when an earlier one fails, so the terminals are always hung up and saved. + await runShutdownSteps([ + { name: "mascots", run: () => mascotManager?.dispose() }, + { name: "agent chat history", run: () => agentChatHistory?.dispose() }, + { name: "agent control", run: () => agentControl?.close() }, + { name: "Even G2", run: () => evenG2?.close() }, + { name: "terminal processes", run: async () => { if (terminalManager) await terminalManager.shutdown(); } } + ]); // The hung-up PTYs exit while the other services close; quitting waits for them (see waitForProcessExits). const ptyExits = terminalManager?.waitForProcessExits().then((left) => { if (left > 0) console.warn(`CanvasTTY quit with ${left} terminal process(es) that did not exit after SIGKILL.`); }); - // Stop the timer first, then abort in-flight limit requests so the final save is prompt. - const usageHistoryClosed = usageHistory?.dispose(); - limitsService?.dispose(); - await usageHistoryClosed; - if (agentGateway) await Promise.allSettled([agentGateway.close()]); - if (runtimeGateway) await Promise.allSettled([runtimeGateway.close()]); - if (browserService) await Promise.allSettled([browserService.dispose()]); - if (pluginServices) await Promise.allSettled([pluginServices.dispose()]); - if (pluginManager) await Promise.allSettled([pluginManager.dispose()]); - await handoffService?.dispose(); - if (materialService) await Promise.allSettled([materialService.dispose()]); + // Stop the usage-history timer first, then abort in-flight limit requests so its final save is prompt. + let usageHistoryClosed: Promise | undefined; + await runShutdownSteps([ + { name: "usage history", run: () => { usageHistoryClosed = usageHistory?.dispose(); } }, + { name: "limits", run: () => limitsService?.dispose() }, + { name: "usage history save", run: () => usageHistoryClosed }, + { name: "agent gateway", run: () => agentGateway?.close() }, + { name: "runtime gateway", run: () => runtimeGateway?.close() }, + { name: "browser", run: () => browserService?.dispose() }, + { name: "plugin services", run: () => pluginServices?.dispose() }, + { name: "plugins", run: () => pluginManager?.dispose() }, + { name: "handoff", run: () => handoffService?.dispose() }, + { name: "materials", run: () => materialService?.dispose() } + ]); await ptyExits; diagnostics.record("info", "application", "shutdown.completed"); await diagnostics.flush(); diff --git a/src/main/services/AgentControlService.ts b/src/main/services/AgentControlService.ts index 3b35fa78a..b4464b671 100644 --- a/src/main/services/AgentControlService.ts +++ b/src/main/services/AgentControlService.ts @@ -6,7 +6,7 @@ import type { SessionSnapshot } from "../../shared/contracts.ts"; import { realpathSync } from "node:fs"; -import { isAbsolute, resolve } from "node:path"; +import { isAbsolute, join, relative, resolve, sep } from "node:path"; import { PROVIDER_CAPABILITIES } from "../../shared/contracts.ts"; import { isLaunchProfile, PROFILE_RANK, profileAvailable, profileCeiling, type LaunchProfile } from "../../shared/autoMode.ts"; import { isPathInside } from "../../agent-runtime/path-inside.mjs"; @@ -14,7 +14,9 @@ import type { TerminalManager } from "./TerminalManager.ts"; import { onDiskPath, otherSpellings } from "./onDiskPath.ts"; import { LaunchRefusal } from "./launchRefusal.ts"; import { terminalFailureDetails } from "./terminalFailureDetails.ts"; +import { terminalScreenText } from "./terminalScreenText.ts"; import { RESULT_CAPTURE_PROVIDERS } from "./resultCapture.ts"; +import { PLUGIN_DATA_FOLDER, privateAppData } from "./isolation/isolationPaths.ts"; // Cards one parent may have in total, live or exited (the live limit is the person's setting). const MAX_CHILDREN_PER_PARENT = 16; @@ -29,8 +31,12 @@ export interface DelegationLimits { } const MAX_OBSERVE_CHARS = 8_192; // An exited agent's reason is in its last screen lines (OpenCode: "Error: Unexpected server error" for an unknown model). -const MAX_EXIT_WINDOW_CHARS = 16_384; const MAX_EXIT_LINES = 20; +/** + * How much of the scrollback (masked first) the plain-text output is rendered from: several full redraws of a + * full-screen TUI, and many screens of a line-based CLI, while each call still masks a bounded window. + */ +const MAX_SCREEN_SOURCE_CHARS = 65_536; const CHILD_POSITION_STEP = { x: 60, y: 60 }; export interface SpawnAgentRequest { @@ -51,7 +57,7 @@ export interface SpawnAgentRequest { export interface AgentObservation { sessionId: string; status: SessionSnapshot["status"]; - /** Raw terminal tail, capped; capabilities with result \"none\" see nothing. */ + /** The screen's text as a terminal shows it (no control sequences), masked, capped to its last characters. */ output: string; /** Once the process exited: its exit code and the last lines of its screen as plain text, masked. */ exitCode?: number | null; @@ -69,12 +75,14 @@ export interface AgentWaitResult { status?: SessionSnapshot["status"]; exitCode: number | null; waitedMs: number; - /** Masked terminal tail (masked before the cut). */ + /** The screen's text as a terminal shows it (no control sequences), masked, capped to its last characters. */ output: string; /** Once the process exited: the last lines of its screen as plain text, masked (why it stopped, e.g. a bad model). */ exitLines?: string; /** The final answer of the turn that ended (Codex, OpenCode subagents), masked. */ answer?: AgentAnswer; + /** Why a Codex or OpenCode subagent has no answer and never will in this launch (read its output instead). */ + answerUnavailable?: AgentAnswerUnavailable; } export interface AgentWaitTiming { @@ -93,6 +101,8 @@ export interface AgentControlOptions { limits?: () => DelegationLimits; /** CanvasTTY's isolation layer can contain an agent on this computer now (a "contained" auto needs it). */ containment?: () => boolean; + /** CanvasTTY's userData folder: a subagent never works in its private data (tokens, account homes, plugin data). */ + userDataPath?: string; } /** The longest wait one call may ask for (wait_for_agent's timeoutSeconds maximum). */ @@ -105,13 +115,26 @@ export interface AgentResult { /** The session's status (idle once its turn ended; state stays "running" while the CLI is open). */ status: SessionSnapshot["status"]; exitCode: number | null; + /** The screen's text as a terminal shows it (no control sequences), masked, capped to its last characters. */ output: string; /** Once the process exited: the last lines of its screen as plain text, masked. */ exitLines?: string; /** The last turn's final answer as the agent reported it (Codex, OpenCode subagents), masked; absent otherwise. */ answer?: AgentAnswer; + /** Why a Codex or OpenCode subagent has no answer and never will in this launch (read its output instead). */ + answerUnavailable?: AgentAnswerUnavailable; +} + +export interface AgentAnswerUnavailable { + /** "status_hooks_off": CanvasTTY status hooks were off for its launch, so it has no hooks to report an answer. */ + reason: "status_hooks_off"; + message: string; } +const STATUS_HOOKS_OFF_MESSAGE = "This subagent's final answer cannot be captured because CanvasTTY status hooks are off " + + "(Settings → Agents; turning them on applies to agents started or restarted afterwards). Read its output instead: " + + "the output field here, or observe_agent."; + export interface AgentAnswer { text: string; /** Only the end of a longer answer was kept. */ @@ -141,9 +164,10 @@ export class AgentControlService { /** * Creates the subagent card. With an initial prompt it resolves only once the prompt reached the agent's PTY * (after an asynchronous launch has started) and rejects with PromptNotDeliveredError when that launch did not - * start; the card stays, so the caller can inspect or cancel it. + * start; the card stays, so the caller can inspect or cancel it. `promptWaitMs` bounds that wait (the launch and + * the CLI's readiness together; LAUNCH_INPUT_WAIT_MS without it). */ - spawn(request: SpawnAgentRequest, signal?: AbortSignal): Promise { + spawn(request: SpawnAgentRequest, signal?: AbortSignal, promptWaitMs?: number): Promise { if (!request || typeof request.parentSessionId !== "string") { throw new Error("A parent session id is required."); } @@ -151,6 +175,8 @@ export class AgentControlService { const capabilities = PROVIDER_CAPABILITIES[request.provider]; if (!capabilities) throw new Error("Unknown agent provider."); if (!capabilities.send) throw new Error(`${request.provider} cannot receive prompts.`); + const inheritance = this.subagentLaunchOptions(parent.id, request.provider, request.launchOptions, request.model); + if (inheritance.launchOptions !== request.launchOptions) request = { ...request, launchOptions: inheritance.launchOptions }; const children = this.children(parent.id); if (children.length >= MAX_CHILDREN_PER_PARENT) { @@ -159,7 +185,7 @@ export class AgentControlService { const lineage = this.lineage(parent.id); const root = lineage.at(-1)!; // What the request asks for first (its folder, its profile), then the person's limits. - const cwd = subagentFolder(root.cwd, parent.cwd, request.cwd); + const cwd = subagentFolder(root.cwd, parent.cwd, request.cwd, this.options.userDataPath); if ("error" in cwd) throw new DelegationRefusal(cwd.error); const profile = subagentProfile(parent.profile, request.provider, request.profile, this.containment()); if ("error" in profile) throw new DelegationRefusal(profile.error); @@ -191,7 +217,7 @@ export class AgentControlService { ...(request.effort !== undefined ? { effort: request.effort } : {}) }, { ...(RESULT_CAPTURE_PROVIDERS.has(request.provider) ? { captureResult: true } : {}), origin: "subagent" }); if (request.initialPrompt === undefined || request.initialPrompt.length === 0) return Promise.resolve(created); - return this.deliver(created.id, `${request.initialPrompt}\r`, "prompt", signal) + return this.deliver(created.id, `${request.initialPrompt}\r`, "prompt", signal, promptWaitMs) .then(() => this.terminals.getMetadata(created.id) ?? created, (error: unknown) => { // Cancelled while it started: nobody receives its id, so the card is closed instead of left running. if (!signal?.aborted) throw error; @@ -200,20 +226,71 @@ export class AgentControlService { }); } + /** + * The plugin launch options a subagent of this parent starts with. A plugin the spawn does not name keeps the + * orchestrator's own values, which the person chose: a subagent never falls back silently to its CLI's default model + * and sign-in (for OpenCode that is OpenCode Zen, a third-party service) while its orchestrator runs on a model the + * person picked through a plugin. Naming a plugin is an explicit choice (checked by the launch: another value needs + * the plugin's delegable declaration). Another CLI cannot inherit a plugin choice that set the orchestrator's model, + * and is refused unless the spawn names that plugin. An explicit model next to an inherited choice that sets the model is refused. + */ + subagentLaunchOptions( + parentSessionId: string, + provider: AgentProviderId, + requested: SpawnAgentRequest["launchOptions"], + model?: string + ): { launchOptions: SpawnAgentRequest["launchOptions"]; sources: Record } { + const parent = this.requireSession(parentSessionId); + const sources: Record = {}; + for (const pluginId of Object.keys(requested ?? {})) sources[pluginId] = "explicit"; + const parentOptions = this.terminals.launchOptionsOf(parent.id); + const unnamed = Object.keys(parentOptions ?? {}).filter((pluginId) => !Object.hasOwn(sources, pluginId)); + if (!parentOptions || unnamed.length === 0) return { launchOptions: requested, sources }; + const modelSetBy = new Set(this.terminals.modelSetByOf(parent.id)); + const modelPlugins = unnamed.filter((pluginId) => modelSetBy.has(pluginId)); + if (parent.provider !== provider) { + if (modelPlugins.length === 0) return { launchOptions: requested, sources }; + const named = modelPlugins.map((pluginId) => pluginId.slice(0, 80)).join(", "); + // The person chose those options for this CLI; another CLI would run on its own default model and sign-in. + throw new DelegationRefusal(`This orchestrator's model is set by its ${named} launch options, which the person chose for ${parent.provider}; ` + + `a ${provider} subagent would not get that model. Spawn a ${parent.provider} subagent instead.`); + } + if (model !== undefined && modelPlugins.length > 0) { + const named = modelPlugins.map((pluginId) => pluginId.slice(0, 80)).join(", "); + throw new DelegationRefusal(`This subagent inherits its orchestrator's ${named} launch options, which set its model; it cannot also take model ${model.slice(0, 80)}. ` + + `Omit model, or pass launchOptions for ${named} that choose the model you want.`); + } + // A plugin turned off since the orchestrator started is left out, unless it set the model: then the subagent would + // run on its CLI's own default model and sign-in, so it is refused instead. + const gone = new Set(this.terminals.unavailableLaunchPlugins(Object.fromEntries(unnamed.map((pluginId) => [pluginId, parentOptions[pluginId]])))); + const goneModel = modelPlugins.filter((pluginId) => gone.has(pluginId)); + if (goneModel.length > 0) { + const named = goneModel.map((pluginId) => pluginId.slice(0, 80)).join(", "); + throw new DelegationRefusal(`This orchestrator's model is set by its ${named} launch options, and ${named} is now disabled, removed or not trusted; ` + + "a subagent would run on its CLI's own default model instead. The person can turn the plugin back on, or start the orchestrator again."); + } + const launchOptions: NonNullable = { ...(requested ?? {}) }; + for (const pluginId of unnamed.filter((id) => !gone.has(id))) { + launchOptions[pluginId] = parentOptions[pluginId]!; + sources[pluginId] = "inherited"; + } + return { launchOptions, sources }; + } + /** The profile a subagent of this parent gets for this request (what spawn will use), or why it gets none. */ profileFor(parentSessionId: string, provider: AgentProviderId, requested?: unknown): { profile: LaunchProfile; inherited: boolean } | { error: string } { return subagentProfile(this.requireSession(parentSessionId).profile, provider, requested, this.containment()); } /** Validates at once (throws); resolves once the text reached the agent, and rejects when it did not. */ - send(sessionId: string, text: string, submit = true, signal?: AbortSignal): Promise { + send(sessionId: string, text: string, submit = true, signal?: AbortSignal, waitMs?: number): Promise { const session = this.requireSession(sessionId); if (session.provider === "terminal") throw new Error("Plain terminals are not agents."); const capabilities = PROVIDER_CAPABILITIES[session.provider as AgentProviderId]; if (!capabilities.send) throw new Error(`${session.provider} cannot receive prompts.`); if (typeof text !== "string" || text.length === 0) throw new Error("Prompt text is required."); if (session.exitCode !== null) throw new Error("Agent session has already exited."); - return this.deliver(sessionId, submit ? `${text}\r` : text, "text", signal); + return this.deliver(sessionId, submit ? `${text}\r` : text, "text", signal, waitMs); } status(sessionId: string): SessionMetadata { @@ -284,25 +361,28 @@ export class AgentControlService { return false; } - observe(sessionId: string, maxChars = MAX_OBSERVE_CHARS): AgentObservation { - const session = this.requireSession(sessionId); - if (session.provider === "terminal") throw new Error("Plain terminals are not agents."); - const capabilities = PROVIDER_CAPABILITIES[session.provider as AgentProviderId]; - if (!capabilities.observe) throw new Error(`${session.provider} cannot be observed.`); + async observe(sessionId: string, maxChars = MAX_OBSERVE_CHARS): Promise { + const session = this.observableSession(sessionId); const buffer = this.terminals.readBuffer(sessionId).buffer; - const exitLines = session.exitCode === null ? null : this.exitLines(buffer); + const exitLines = session.exitCode === null ? null : await this.screenExitLines(sessionId, buffer); return { sessionId: session.id, status: session.status, - // Masked before the cut (a cut inside a secret would leave a tail no pattern recognizes), over a window - // wider than any match rather than the whole scrollback. - output: this.redactTail(buffer, maxChars), + output: await this.screenText(sessionId, buffer, maxChars), ...(session.exitCode === null ? {} : { exitCode: session.exitCode }), ...(exitLines ? { exitLines } : {}) }; } - result(sessionId: string): AgentResult { + private observableSession(sessionId: string): SessionMetadata { + const session = this.requireSession(sessionId); + if (session.provider === "terminal") throw new Error("Plain terminals are not agents."); + const capabilities = PROVIDER_CAPABILITIES[session.provider as AgentProviderId]; + if (!capabilities.observe) throw new Error(`${session.provider} cannot be observed.`); + return session; + } + + async result(sessionId: string): Promise { const session = this.requireSession(sessionId); if (session.provider === "terminal") throw new Error("Plain terminals are not agents."); const capabilities = PROVIDER_CAPABILITIES[session.provider as AgentProviderId]; @@ -313,6 +393,7 @@ export class AgentControlService { const buffer = capabilities.result === "terminal" ? this.terminals.readBuffer(sessionId).buffer : ""; + const exitLines = session.exitCode === null ? null : await this.screenExitLines(sessionId, buffer); return { sessionId: session.id, state: session.exitCode === null @@ -320,9 +401,9 @@ export class AgentControlService { : session.exitCode === 0 ? "done" : "failed", status: session.status, exitCode: session.exitCode, - output: this.redactTail(buffer, MAX_OBSERVE_CHARS), - ...(session.exitCode !== null && this.exitLines(buffer) ? { exitLines: this.exitLines(buffer)! } : {}), - ...(answer ? { answer } : {}) + output: buffer ? await this.screenText(sessionId, buffer, MAX_OBSERVE_CHARS) : "", + ...(exitLines ? { exitLines } : {}), + ...(answer ? { answer } : this.answerUnavailable(sessionId)) }; } @@ -340,14 +421,23 @@ export class AgentControlService { const timing = this.options.waitTiming ?? AGENT_WAIT_TIMING; const timeoutMs = Math.min(MAX_AGENT_WAIT_MS, Math.max(0, request.timeoutMs)); const started = Date.now(); - const answer = (session: SessionMetadata | null, reason: AgentWaitReason): AgentWaitResult => { + const answer = async (session: SessionMetadata | null, reason: AgentWaitReason): Promise => { const waitedMs = Date.now() - started; if (!session) return { sessionId, reason, exitCode: null, waitedMs, output: "" }; - let observation: AgentObservation | null = null; - try { observation = this.observe(sessionId); } catch { observation = null; } + let observable = false; + try { observable = Boolean(this.observableSession(sessionId)); } catch { observable = false; } + let output = ""; + let exitLines: string | null = null; + if (observable) { + try { + const buffer = this.terminals.readBuffer(sessionId).buffer; + output = await this.screenText(sessionId, buffer, MAX_OBSERVE_CHARS); + if (session.exitCode !== null) exitLines = await this.screenExitLines(sessionId, buffer); + } catch { output = ""; } + } const answer = reason === "timeout" || reason === "needs_approval" ? null : this.answer(sessionId); - return { sessionId, reason, status: session.status, exitCode: session.exitCode, waitedMs, output: observation?.output ?? "", - ...(observation?.exitLines ? { exitLines: observation.exitLines } : {}), ...(answer ? { answer } : {}) }; + return { sessionId, reason, status: session.status, exitCode: session.exitCode, waitedMs, output, + ...(exitLines ? { exitLines } : {}), ...(answer ? { answer } : this.answerUnavailable(sessionId)) }; }; let offset = this.outputOffset(sessionId); let changedAt = started; @@ -381,20 +471,51 @@ export class AgentControlService { } /** Through the terminal manager's one delivery rule: exactly once, into the launch that is starting now. */ - private async deliver(sessionId: string, data: string, what: "prompt" | "text", signal?: AbortSignal): Promise { - const delivery = await this.terminals.deliverInput(sessionId, data, undefined, signal); + private async deliver(sessionId: string, data: string, what: "prompt" | "text", signal?: AbortSignal, waitMs?: number): Promise { + const delivery = await this.terminals.deliverInput(sessionId, data, waitMs, signal); if (!delivery.delivered) { throw new PromptNotDeliveredError(sessionId, `The ${what} for agent ${sessionId} was not delivered: ${delivery.reason}`); } } - /** The last lines of an exited agent's screen as plain text: masked over the whole tail window before it is cut. */ - private exitLines(buffer: string): string | null { - const text = terminalFailureDetails(this.redactTail(buffer, MAX_EXIT_WINDOW_CHARS)); + /** + * The last lines of the screen, rendered like `output` (a TUI that redraws only the cells that changed would + * otherwise read as its redraws joined together) from the same window as `output` (a line drawn early and never + * redrawn, such as the model's status line, is still on the screen), masked before it is cut and again after. + */ + private async screenExitLines(sessionId: string, buffer: string): Promise { + const source = fromSequenceBoundary(this.redactTail(buffer, MAX_SCREEN_SOURCE_CHARS), buffer.length > MAX_SCREEN_SOURCE_CHARS); + let screen: string; + try { + screen = this.redactTail(await terminalScreenText(source, this.geometry(sessionId)), MAX_SCREEN_SOURCE_CHARS); + } catch { + screen = source; + } + const text = terminalFailureDetails(screen); if (!text) return null; return text.split("\n").slice(-MAX_EXIT_LINES).join("\n"); } + /** + * The text the agent's terminal shows, as plain text: a TUI's redraws (cursor moves, erases, colors) rendered by a + * headless terminal instead of returned as escape sequences. Masked twice: the scrollback before it is rendered + * (a held value the TUI wrote whole is gone before any redraw can split it), and the rendered text before its + * last `maxChars` are cut (a value the screen shows across wrapped rows or box sides). Should the rendering fail, + * the masked raw tail is returned as before. + */ + private async screenText(sessionId: string, buffer: string, maxChars: number): Promise { + const source = fromSequenceBoundary(this.redactTail(buffer, MAX_SCREEN_SOURCE_CHARS), buffer.length > MAX_SCREEN_SOURCE_CHARS); + try { + return this.redactTail(await terminalScreenText(source, this.geometry(sessionId)), maxChars); + } catch { + return this.redactTail(buffer, maxChars); + } + } + + private geometry(sessionId: string): { cols: number; rows: number } | null { + try { return typeof this.terminals.geometry === "function" ? this.terminals.geometry(sessionId) : null; } catch { return null; } + } + /** Plugin launch secrets never reach another agent through observed output: the tail as masking the whole text leaves it. */ private redactTail(text: string, maxChars: number): string { if (typeof this.terminals.redactSecretsTail === "function") return this.terminals.redactSecretsTail(text, maxChars); @@ -405,6 +526,14 @@ export class AgentControlService { try { return typeof this.terminals.turnProgress === "function" ? this.terminals.turnProgress(sessionId) : null; } catch { return null; } } + /** Says so when no answer can be captured for this agent, so an orchestrator reads its output instead of waiting for one. */ + private answerUnavailable(sessionId: string): { answerUnavailable?: AgentAnswerUnavailable } { + try { + const reason = typeof this.terminals.answerUnavailable === "function" ? this.terminals.answerUnavailable(sessionId) : null; + return reason === "status_hooks_off" ? { answerUnavailable: { reason, message: STATUS_HOOKS_OFF_MESSAGE } } : {}; + } catch { return {}; } + } + private answer(sessionId: string): AgentAnswer | null { try { const answer = typeof this.terminals.answer === "function" ? this.terminals.answer(sessionId) : null; @@ -477,9 +606,11 @@ export function subagentProfile( /** * Where a subagent may work: its orchestrator's project folder (the folder the person chose for the agent it * descends from) or a folder inside it, compared as real paths in the spelling the disk uses (NFC and NFD name the - * same folder on macOS). A relative folder is taken from the orchestrator's own folder. + * same folder on macOS). A relative folder is taken from the orchestrator's own folder. Never CanvasTTY's private + * data under `userDataPath` (a project such as the home folder contains it): a folder inside one plugin's data only + * when the project itself is a folder inside that same plugin's data (an environment's worktree). */ -export function subagentFolder(projectRoot: string, parentCwd: string, requested: unknown): { cwd: string } | { error: string } { +export function subagentFolder(projectRoot: string, parentCwd: string, requested: unknown, userDataPath?: string): { cwd: string } | { error: string } { if (typeof requested !== "string" || requested.trim().length === 0) return { error: "cwd is required: a folder inside this project." }; const wanted = onDiskPath(isAbsolute(requested) ? requested : resolve(parentCwd, requested)); const real = (path: string): string | null => { @@ -494,9 +625,32 @@ export function subagentFolder(projectRoot: string, parentCwd: string, requested if (!root || !(isPathInside(root, folder) || isPathInside(root.normalize("NFC"), folder.normalize("NFC")))) { return { error: `A subagent works only inside this project's folder (${projectRoot}); ${requested} is outside it. Only the person can start an agent in another folder.` }; } + if (userDataPath && insidePrivateData(root, folder, userDataPath, real)) { + return { error: `A subagent cannot work in ${requested}: it is CanvasTTY's private data (tokens, account homes, plugins' data). Only the person can start an agent there.` }; + } return { cwd: folder }; } +/** + * `folder` (a real path) is inside CanvasTTY's private data, and not inside a project that is itself a folder within + * one plugin's data (an environment's worktree; then that plugin folder is the project's own). + */ +function insidePrivateData(root: string, folder: string, userDataPath: string, real: (path: string) => string | null): boolean { + const inside = (base: string, path: string): boolean => isPathInside(base, path) || isPathInside(base.normalize("NFC"), path.normalize("NFC")); + const data = real(resolve(userDataPath)) ?? resolve(userDataPath); + for (const hidden of privateAppData(data)) { + const base = real(hidden) ?? hidden; + if (!inside(base, folder)) continue; + if (hidden !== join(data, PLUGIN_DATA_FOLDER)) return true; + // //…: the project must be strictly inside that same plugin's folder. + const plugin = relative(base.normalize("NFC"), folder.normalize("NFC")).split(sep)[0]; + if (!plugin) return true; + const pluginFolder = join(base, plugin); + if (!inside(pluginFolder, root) || inside(root, pluginFolder)) return true; + } + return false; +} + /** Sleeps, or rejects with an AbortError as soon as `signal` aborts. */ function pause(ms: number, signal?: AbortSignal): Promise { return new Promise((resolve, reject) => { @@ -521,3 +675,13 @@ function tail(text: string, maxChars: number): string { function spawnCanceled(): Error { return new DOMException("The spawn was canceled before its agent started.", "AbortError"); } + +/** + * A window cut from the end of a terminal stream may begin inside an escape sequence, whose rest (`18;9H`, `m`) a + * terminal would print as text. A cut window therefore starts at its first escape or line break instead. + */ +function fromSequenceBoundary(window: string, cut: boolean): string { + if (!cut) return window; + const starts = [window.indexOf("\u001b"), window.indexOf("\n")].filter((index) => index >= 0); + return starts.length > 0 ? window.slice(Math.min(...starts)) : window; +} diff --git a/src/main/services/DecisionHooks.ts b/src/main/services/DecisionHooks.ts index a4c910731..b768fbb0d 100644 --- a/src/main/services/DecisionHooks.ts +++ b/src/main/services/DecisionHooks.ts @@ -25,7 +25,10 @@ export interface DecisionSession { role: SessionRole; /** The session's working folder: "outside" is measured from here. */ cwd: string; - /** The agent's own config folders (CLAUDE_CONFIG_DIR of this run), whose plans and memory are not "outside". */ + /** + * The agent's own config folders (CLAUDE_CONFIG_DIR of this run, and the account home a launch contributor selected + * for it): their plans and memory are not "outside", and CanvasTTY's private data does not include them. + */ configDirs: string[]; /** The profile the card runs in. */ profile?: LaunchProfileId; @@ -85,7 +88,8 @@ const MAX_SERVICES = 8; * first and its deny is final. Then every trusted decision service that applies answers deny, ask, allow or * nothing, in parallel, within 3 s. Any deny wins; else any ask (a timeout, an error or an unreadable answer is an * ask); else an allow counts only from a plugin the person separately let allow; else no verdict and the agent goes - * on as it would without CanvasTTY. Nothing here ever turns a failure into an allow. + * on as it would without CanvasTTY. Nothing here ever turns a failure into an allow. A read-only file tool call is + * checked by base protection's private-data rule only; decision services are not asked about it. */ export class DecisionHooks { private readonly deps: DecisionHooksDependencies; @@ -101,6 +105,11 @@ export class DecisionHooks { return services === null || services.length > 0; } + /** Whether a launch's hook should also see read-only file tools: only base protection checks them. */ + checksReads(): boolean { + return this.protects(); + } + /** The longest wait a decision service of this agent asked for: the session's gate is sized for it at launch. */ budgetMs(provider: AgentProviderId): number { return Math.max(DECIDE_TIMEOUT_MS, ...(this.applicable(provider) ?? []).map((service) => this.timeoutFor(service))); @@ -129,15 +138,18 @@ export class DecisionHooks { // Cut input: the rules could not see the command, or where a file tool writes. Unchecked is not allowed. if (request.truncated) { const cut = actionFromHook(request.toolName, request.toolInput, request.toolInputPreview); - if (cut.kind === "shell" || cut.kind === "edit" && cut.paths.length === 0) return { behavior: "deny", message: TOO_LARGE_MESSAGE }; + // A cut read-only call (a read_many_files list, a huge Grep pattern) may hide the path it reads. + if (cut.kind === "shell" || cut.kind === "read" || cut.kind === "edit" && cut.paths.length === 0) return { behavior: "deny", message: TOO_LARGE_MESSAGE }; } } + // Read-only file tools reach the hook only for base protection's private-data rule: plugins are not asked. + const action = actionFromHook(request.toolName, request.toolInput, request.toolInputPreview); + if (action.kind === "read") return { behavior: "none" }; const services = this.applicable(session.provider); // The plugin list itself failed: that is the gateway's failure (the person is asked, or a fail-closed gate // denies), never "no plugin has an opinion". if (services === null) throw new Error("decision services unavailable"); if (services.length === 0) return { behavior: "none" }; - const action = actionFromHook(request.toolName, request.toolInput, request.toolInputPreview); const params: Omit = { event: "pre-tool", sessionId, diff --git a/src/main/services/FileAccessService.ts b/src/main/services/FileAccessService.ts index 07c735a0e..6005c761b 100644 --- a/src/main/services/FileAccessService.ts +++ b/src/main/services/FileAccessService.ts @@ -204,9 +204,13 @@ export class FileAccessService { if (!isContainedPath(root.path, resolvedAfterOpen)) { return { kind: "unsupported", reason: "not-permitted" }; } + // Bind the descriptor to that contained path: the file it names now must be the very file that + // was opened. Where no descriptor path is exposed (macOS, Windows) this is what catches an open + // redirected outside the root by a swapped intermediate directory. Inode numbers are compared as + // bigints: 64-bit file ids (NTFS, some network file systems) lose precision as numbers. try { - const current = await stat(resolvedAfterOpen); - if (current.dev !== metadata.dev || current.ino !== metadata.ino) { + const [current, opened] = [await stat(resolvedAfterOpen, { bigint: true }), await handle.stat({ bigint: true })]; + if (current.dev !== opened.dev || current.ino !== opened.ino) { return { kind: "unsupported", reason: "not-permitted" }; } } catch { diff --git a/src/main/services/LaunchPipeline.ts b/src/main/services/LaunchPipeline.ts index 439370a56..7f6c00564 100644 --- a/src/main/services/LaunchPipeline.ts +++ b/src/main/services/LaunchPipeline.ts @@ -1,5 +1,5 @@ import { randomUUID } from "node:crypto"; -import { mkdir, rm, writeFile } from "node:fs/promises"; +import { mkdir, rm, rmdir, writeFile } from "node:fs/promises"; import { dirname, join, resolve } from "node:path"; import type { LaunchProfileId, @@ -13,6 +13,7 @@ import { claudeCoreSettingsKey, coreOwnedLaunchArgument, parseInlineSettings } f import { MAX_PLUGIN_SLOT_BYTES } from "./TerminalSessionStore.ts"; import { LaunchRefusal } from "./launchRefusal.ts"; import { MAX_INSPECTED_CONFIG_BYTES, parseJsonc, readInspectedFile } from "./inspectedConfig.ts"; +import { contributedAccountHome } from "./isolation/accountHomeIsolation.ts"; /** A trusted plugin service that declared launch options (PluginManager.launchContributors). */ export interface LaunchContributor { @@ -20,6 +21,8 @@ export interface LaunchContributor { pluginName: string; serviceId: string; launch: PluginServiceLaunch; + /** The plugin's private data folder, set by the host; never sent to the plugin or taken from its answer. */ + dataDir?: string; /** The plugin holds the `secrets` permission, so `secretEnv` may name its secrets. */ secrets: boolean; } @@ -73,6 +76,16 @@ export type PreparedLaunch = envSources: Record; /** A contributor runs the agent on another model than its vendor's: "auto" becomes accept-edits. */ thirdPartyModel: boolean; + /** Plugin ids whose contribution decides the model: it passes the CLI's model flag or runs a third-party model. */ + modelSetBy: string[]; + /** This run's own folder below the card's folder; `cleanup` removes it. */ + runDirectory: string; + /** + * The CLI home a chosen contributor started this launch with inside its own data folder (an account it selected): + * the one folder of plugin data the isolation layer reopens for this agent. Derived by the host, never a grant a + * plugin can name. + */ + accountHome?: string; cleanup(): Promise; } | { ok: false; reason: string }; @@ -118,9 +131,25 @@ export class LaunchPipeline { return rm(this.dependencies.runsRoot, { recursive: true, force: true }); } - /** Removes a closed session's file folders. */ - forgetSession(sessionId: string): Promise { - return rm(join(this.dependencies.runsRoot, safeSegment(sessionId)), { recursive: true, force: true }); + /** + * Removes a closed card's run folders that this pipeline prepared for it, then the card's folder if nothing else + * is left in it. A card brought back under the same id may already be writing a new run folder there; it is kept. + */ + async forgetSession(sessionId: string, runDirectories: Iterable): Promise { + const parent = join(this.dependencies.runsRoot, safeSegment(sessionId)); + const errors: unknown[] = []; + for (const runDirectory of new Set(runDirectories)) { + // Only a run folder directly below this card's folder is ever removed here. + if (dirname(resolve(runDirectory)) !== resolve(parent)) continue; + try { await removeRunFolder(runDirectory); } catch (error) { errors.push(error); } + } + try { + await rmdir(parent); + } catch (error) { + // Still in use by a newer run of the same card, or already gone; the startup sweep removes what is left. + if (!PARENT_KEPT_CODES.has((error as NodeJS.ErrnoException).code ?? "")) errors.push(error); + } + if (errors.length) throw new AggregateError(errors, "Launch run folders could not be removed."); } /** Launch policies that apply to this agent: every launch of it asks them, chosen or not. */ @@ -145,45 +174,45 @@ export class LaunchPipeline { * Checks launcher values against each plugin's declared fields and fills defaults. * Throws with a person-readable reason; returns undefined when nothing was chosen. */ - normalizeOptions(provider: ProviderId, candidate: unknown, context?: { delegated?: boolean }): Record | undefined { + normalizeOptions(provider: ProviderId, candidate: unknown, context?: { + delegated?: boolean; + /** The orchestrator's own options: repeating them for its subagent is the person's choice, not the orchestrator's. */ + inherited?: Readonly>>; + }): Record | undefined { if (candidate === undefined) return undefined; if (!isRecord(candidate) || Object.keys(candidate).length > MAX_OPTION_PLUGINS) throw new Error("Launch options are invalid."); if (Object.keys(candidate).length === 0) return undefined; if (provider === "terminal") throw new Error("A plain terminal takes no launch options."); const contributors = new Map(this.dependencies.contributors().map((contributor) => [contributor.pluginId, contributor])); const options: Record = {}; - for (const [pluginId, raw] of Object.entries(candidate)) { + for (const [pluginId, requested] of Object.entries(candidate)) { + let raw = requested; const contributor = contributors.get(pluginId); if (!contributor) throw new Error(unavailableReason(pluginId)); // An orchestrator picks options for its subagents only where the plugin said that is safe. + const notDelegable = `${contributor.pluginName} has not declared its launch options safe for an orchestrator to choose; only the person chooses them, in the launcher.`; + // Without the declaration a subagent may only repeat the orchestrator's own values: the keys it leaves out are + // the inherited ones, never the plugin's defaults, and the result is checked against them below. + const inherited = context?.delegated && contributor.launch.delegable !== true ? context.inherited?.[pluginId] : undefined; if (context?.delegated && contributor.launch.delegable !== true) { - throw new LaunchRefusal(`${contributor.pluginName} has not declared its launch options safe for an orchestrator to choose; only the person chooses them, in the launcher.`); + if (!inherited || !sameAsInherited(raw, inherited)) throw new LaunchRefusal(notDelegable); + raw = { ...inherited, ...(raw as Record) }; } - const { appliesTo, fields } = contributor.launch; - if (appliesTo && !appliesTo.includes(provider as never)) { - throw new Error(`${contributor.pluginName} launch options do not apply to ${provider}.`); + if (!inherited) { + options[pluginId] = launchValues(contributor, provider, raw); + continue; } - if (!isRecord(raw)) throw new Error(`${contributor.pluginName} launch options are invalid.`); - const known = new Set(fields.map((field) => field.key)); - const unknown = Object.keys(raw).find((key) => !known.has(key)); - if (unknown) throw new Error(`${contributor.pluginName} has no launch option ${unknown.slice(0, 40)}.`); - const values: PluginLaunchValues = {}; - for (const field of fields) { - const value = raw[field.key] ?? field.default - ?? (field.kind === "boolean" ? false : field.kind === "select" ? field.options?.[0]?.value ?? "" : ""); - const plainText = (limit: number): boolean => typeof value === "string" && value.length <= limit && !/[\u0000-\u001f\u007f]/.test(value); - // A service-provided choice may have changed since the launcher showed it: the service checks it when it prepares. - const valid = field.kind === "boolean" - ? typeof value === "boolean" - : field.kind === "select" - ? typeof value === "string" && (Boolean(field.options?.some((option) => option.value === value)) || (field.optionsFrom === "service" && plainText(200))) - : plainText(field.maxLength ?? 200); - if (!valid) throw new Error(`${contributor.pluginName} launch option ${field.label} is invalid.`); - values[field.key] = value as boolean | string; - } - if (Buffer.byteLength(JSON.stringify(values), "utf8") > MAX_PLUGIN_SLOT_BYTES) { - throw new Error(`${contributor.pluginName} launch options exceed 4 KB.`); + // The person's earlier choice no longer fits the plugin (it changed its fields or what it applies to): a subagent + // never gets other values than its orchestrator's, so it is refused with that reason. + const changed = `${contributor.pluginName} launch options changed since this orchestrator started, so its subagents cannot repeat the person's choice. ` + + "The person can start the orchestrator again with the new options."; + let values: PluginLaunchValues; + try { + values = launchValues(contributor, provider, raw); + } catch { + throw new LaunchRefusal(changed); } + if (!sameValues(values, inherited)) throw new LaunchRefusal(changed); options[pluginId] = values; } return options; @@ -239,8 +268,10 @@ export class LaunchPipeline { * refuses the whole launch with a reason; nothing is ever launched without a contribution the person selected, * or past a policy that did not answer. */ - async prepare(context: LaunchSessionContext): Promise { + async prepare(context: LaunchSessionContext, launch: { model?: string } = {}): Promise { const contributors = new Map(this.dependencies.contributors().map((contributor) => [contributor.pluginId, contributor])); + // Only a subagent's launch is refused for a second model; it was chosen for it, not by the person. + const delegated = context.role === "subagent"; const selected = Object.keys(context.options).sort(); for (const pluginId of selected) { if (!contributors.has(pluginId)) return { ok: false, reason: unavailableReason(pluginId) }; @@ -251,7 +282,7 @@ export class LaunchPipeline { ...policies.map((contributor) => this.ask(contributor, context, false)) ]); const runDirectory = join(this.dependencies.runsRoot, safeSegment(context.sessionId), randomUUID()); - const cleanup = (): Promise => rm(runDirectory, { recursive: true, force: true }); + const cleanup = (): Promise => removeRunFolder(runDirectory); const refuse = async (reason: string): Promise => { await cleanup().catch(() => undefined); return { ok: false, reason }; @@ -267,6 +298,8 @@ export class LaunchPipeline { const args: string[] = []; const secrets: string[] = []; let thirdPartyModel = false; + const modelSetBy: string[] = []; + let accountHome: string | undefined; for (const answer of answers) { if ("refuse" in answer) return refuse(answer.refuse); } @@ -285,6 +318,19 @@ export class LaunchPipeline { } const forbidden = contributedArgs.find((argument) => coreOwnedLaunchArgument(context.provider, argument)); if (forbidden) return refuse(`${name} added ${forbidden.slice(0, 60)}, which only CanvasTTY may pass.`); + const modelFlag = modelSettingArgument(context.provider, contributedArgs); + if (modelFlag !== undefined || contribution.thirdPartyModel === true) { + // A subagent's CLI gets one model: a second model flag next to the launch's own made OpenCode crash at start. + // A person's own launches (and their restores) start as they always did; the person chose both. + if (delegated && modelFlag !== undefined && launch.model !== undefined) { + return refuse(`${name} sets the model of this launch, and the launch also asks for model ${launch.model.slice(0, 80)}. Omit the model, or choose ${name} options for the model you want.`); + } + if (delegated && modelSetBy.length > 0) { + const first = contributors.get(modelSetBy[0]!)?.pluginName ?? modelSetBy[0]!; + return refuse(`${first} and ${name} both set the model of this launch; choose launch options from only one of them.`); + } + modelSetBy.push(contributor.pluginId); + } // A configuration a CLI reads from the environment or an argument may not decide its approvals either. const widening = permissionConfigProblem(context.provider, contribution, context.cwd); if (widening) return refuse(`${name} ${widening}`); @@ -313,6 +359,7 @@ export class LaunchPipeline { const conflict = claim(key); if (conflict) return refuse(conflict); env[key] = expand(value); + accountHome = contributedAccountHome(context.provider, contributor.dataDir, key, env[key]) ?? accountHome; } for (const [key, secretKey] of Object.entries(contribution.secretEnv)) { const conflict = claim(key); @@ -327,7 +374,7 @@ export class LaunchPipeline { args.push(...contributedArgs.map(expand)); thirdPartyModel ||= contribution.thirdPartyModel === true; } - return { ok: true, env, args, secrets, envSources, thirdPartyModel, cleanup }; + return { ok: true, env, args, secrets, envSources, thirdPartyModel, modelSetBy, runDirectory, ...(accountHome ? { accountHome } : {}), cleanup }; } private async ask( @@ -562,6 +609,17 @@ function unavailableReason(pluginId: string): string { return `Needs plugin ${pluginId} for its launch options; it is disabled, removed, or its native code is not trusted.`; } +/** Errors that leave a card's folder in place on purpose: another run still uses it, or it is gone. */ +const PARENT_KEPT_CODES = new Set(["ENOENT", "ENOTEMPTY", "EEXIST", "EBUSY", "EPERM"]); + +/** + * Removes one run folder. On Windows the agent may still hold a file there for a moment after its PTY exits, so + * busy/permission errors are retried briefly before giving up. + */ +function removeRunFolder(runDirectory: string): Promise { + return rm(runDirectory, { recursive: true, force: true, maxRetries: 5, retryDelay: 100 }); +} + function safeSegment(value: string): string { return value.replace(/[^A-Za-z0-9._-]/g, "_").replace(/^\.+/, "_").slice(0, 128) || "_"; } @@ -575,3 +633,92 @@ export function errorText(error: unknown): string { export function isRecord(value: unknown): value is Record { return Boolean(value && typeof value === "object" && !Array.isArray(value)); } + +/** A plugin's launch values checked against its declared fields, with defaults filled in. Throws with a person-readable reason. */ +function launchValues(contributor: LaunchContributor, provider: ProviderId, raw: unknown): PluginLaunchValues { + const { appliesTo, fields } = contributor.launch; + if (appliesTo && !appliesTo.includes(provider as never)) { + throw new Error(`${contributor.pluginName} launch options do not apply to ${provider}.`); + } + if (!isRecord(raw)) throw new Error(`${contributor.pluginName} launch options are invalid.`); + const known = new Set(fields.map((field) => field.key)); + const unknown = Object.keys(raw).find((key) => !known.has(key)); + if (unknown) throw new Error(`${contributor.pluginName} has no launch option ${unknown.slice(0, 40)}.`); + const values: PluginLaunchValues = {}; + for (const field of fields) { + const value = raw[field.key] ?? field.default + ?? (field.kind === "boolean" ? false : field.kind === "select" ? field.options?.[0]?.value ?? "" : ""); + const plainText = (limit: number): boolean => typeof value === "string" && value.length <= limit && !/[\u0000-\u001f\u007f]/.test(value); + // A service-provided choice may have changed since the launcher showed it: the service checks it when it prepares. + const valid = field.kind === "boolean" + ? typeof value === "boolean" + : field.kind === "select" + ? typeof value === "string" && (Boolean(field.options?.some((option) => option.value === value)) || (field.optionsFrom === "service" && plainText(200))) + : plainText(field.maxLength ?? 200); + if (!valid) throw new Error(`${contributor.pluginName} launch option ${field.label} is invalid.`); + values[field.key] = value as boolean | string; + } + if (Buffer.byteLength(JSON.stringify(values), "utf8") > MAX_PLUGIN_SLOT_BYTES) { + throw new Error(`${contributor.pluginName} launch options exceed 4 KB.`); + } + return values; +} + +/** The orchestrator's own choice repeated for its subagent: only keys it chose, each with the same value. */ +function sameAsInherited(raw: unknown, inherited: Readonly | undefined): boolean { + if (!inherited || !isRecord(raw)) return false; + const keys = Object.keys(raw); + // An empty choice repeats only an empty one (a plugin without fields, chosen in the launcher). + if (keys.length === 0) return Object.keys(inherited).length === 0; + return keys.every((key) => Object.hasOwn(inherited, key) && inherited[key] === raw[key]); +} + +/** Exactly the same keys and values. */ +function sameValues(values: Readonly, inherited: Readonly): boolean { + const keys = Object.keys(values); + return keys.length === Object.keys(inherited).length + && keys.every((key) => Object.hasOwn(inherited, key) && inherited[key] === values[key]); +} + +/** + * How each CLI's arguments choose its model. Every CLI here takes `--model`; `short` are its short model flags, + * `profile` flags that select a configuration profile (which may name a model), and `config` flags that set a + * configuration key, with the keys that decide the model. + */ +const MODEL_ARGUMENTS: Readonly>> = { + codex: { short: ["-m"], profile: ["--profile", "-p"], config: { flags: ["-c", "--config"], keys: ["model", "model_provider", "profile"] } }, + opencode: { short: ["-m"] }, + qwen: { short: ["-m"] }, + kimi: { short: ["-m"] }, + grok: { short: ["-m"] } +}; + +/** + * The first argument that chooses the CLI's model, in any form: `--model x`, `--model=x`, a short `-m x` or `-mx`, + * and for Codex `--profile`/`-p` and `-c`/`--config` setting `model`, `model_provider` or `profile`. Undefined when none does. + */ +export function modelSettingArgument(provider: ProviderId, args: readonly string[]): string | undefined { + const table = MODEL_ARGUMENTS[provider] ?? {}; + const valueFlags = ["--model", ...(table.short ?? []), ...(table.profile ?? [])]; + const configKeys = new Set(table.config?.keys ?? []); + const configSetsModel = (value: string | undefined): boolean => { + if (value === undefined) return false; + const key = value.split("=", 1)[0]!.trim().replace(/^["']|["']$/gu, ""); + return value.includes("=") && configKeys.has(key); + }; + for (let index = 0; index < args.length; index++) { + const argument = args[index]!; + if (valueFlags.some((flag) => argument === flag || attachedValue(argument, flag) !== undefined)) return argument; + for (const flag of table.config?.flags ?? []) { + const value = argument === flag ? args[index + 1] : attachedValue(argument, flag); + if (configSetsModel(value)) return argument === flag ? `${argument} ${value}` : argument; + } + } + return undefined; +} + +/** The value written into the flag's own argument: `--flag=value`, or `-fvalue` for a short flag. */ +function attachedValue(argument: string, flag: string): string | undefined { + if (flag.startsWith("--")) return argument.startsWith(`${flag}=`) ? argument.slice(flag.length + 1) : undefined; + return argument.length > flag.length && argument.startsWith(flag) ? argument.slice(flag.length) : undefined; +} diff --git a/src/main/services/LocalUsageCollector.ts b/src/main/services/LocalUsageCollector.ts index 2ab948d52..14331003c 100644 --- a/src/main/services/LocalUsageCollector.ts +++ b/src/main/services/LocalUsageCollector.ts @@ -149,6 +149,28 @@ export function sanitizeCollectorState(input: unknown): Clean { async function sideFile(path: string): Promise { try { return (await stat(path)).size; } catch { return null; } } +/** Attempts of a rollback-journal read while a writer holds the database, each after the CLI's busy timeout. */ +const BUSY_ATTEMPTS = 3; +const BUSY_RETRY_MS = 250; +const isBusy = (error: unknown): boolean => /database (?:table )?is locked|SQLITE_BUSY|\bbusy\b/i + .test(String((error as { stderr?: unknown }).stderr ?? (error as Error)?.message ?? '')); +/** + * Whether the database header marks WAL mode: file format write and read versions (bytes 18 and 19) are both 2. + * A rollback-journal database has 1 there; an unreadable or foreign header counts as not WAL. + */ +async function walHeader(path: string): Promise { + let handle: FileHandle | undefined; + try { + handle = await open(path, 'r'); + const header = Buffer.alloc(100); + const { bytesRead } = await handle.read(header, 0, header.length, 0); + return bytesRead === header.length && header.toString('latin1', 0, 16) === 'SQLite format 3\0' && header[18] === 2 && header[19] === 2; + } catch { + return false; + } finally { + await handle?.close().catch(() => undefined); + } +} function sqliteProblem(error: unknown): string { if (error instanceof SourceProblem) return error.message; const e = error as NodeJS.ErrnoException & { stderr?: unknown; killed?: boolean }; @@ -396,11 +418,16 @@ export class LocalUsageCollector { } /** - * A plain read-only connection creates a missing -shm file; readonly_shm never does. When no WAL - * frames can exist (no or empty -wal) an immutable read takes no locks and creates nothing. + * A rollback-journal database (header not WAL) is read through a plain read-only connection: it takes a shared + * lock, so a writer's half-written pages are never read, and it creates no side files. A busy database is + * retried a few times, then skipped for this poll. For a WAL database a plain read-only connection creates a + * missing -shm file; readonly_shm never does. When no WAL frames can exist (no or empty -wal) an immutable read + * takes no locks and creates nothing; only a WAL database may be read that way, since its writer always keeps + * -wal and -shm beside it while it runs. */ private async sqlite(database: string, sql: string): Promise[]> { const target = (query: string): string => `${pathToFileURL(database).href}?${query}`; + if (!await walHeader(database)) return await this.lockedQuery(target('mode=ro'), sql); const [wal, shm] = await Promise.all([sideFile(`${database}-wal`), sideFile(`${database}-shm`)]); if (shm === null && wal) throw new SourceProblem('WAL without shared-memory index (writer stopped uncleanly); skipped to avoid a stale read'); // A read-only WAL connection can still create side files beside a stopped writer. @@ -414,6 +441,17 @@ export class LocalUsageCollector { } } + private async lockedQuery(target: string, sql: string): Promise[]> { + for (let attempt = 1; ; attempt++) { + try { + return await this.query(target, sql); + } catch (error) { + if (attempt >= BUSY_ATTEMPTS || !isBusy(error)) throw error; + await new Promise(done => setTimeout(done, BUSY_RETRY_MS * attempt)); + } + } + } + private async query(target: string, sql: string): Promise[]> { const executable = this.sqlitePath ?? await discoverSqlite(this.sqliteEnvironment); if (!executable) throw new SourceProblem('sqlite3 is unavailable (not found on PATH)'); diff --git a/src/main/services/PluginManager.ts b/src/main/services/PluginManager.ts index 0e86e7192..bfedc269c 100644 --- a/src/main/services/PluginManager.ts +++ b/src/main/services/PluginManager.ts @@ -654,6 +654,7 @@ export class PluginManager { pluginName: manifest.name, serviceId: service.id, launch: structuredClone(service.launch), + dataDir: join(this.dataRoot, plugin.manifest.id), secrets: manifest.permissions.includes("secrets") }); } diff --git a/src/main/services/TerminalManager.ts b/src/main/services/TerminalManager.ts index 2b70689de..3afd86bf6 100644 --- a/src/main/services/TerminalManager.ts +++ b/src/main/services/TerminalManager.ts @@ -1,4 +1,5 @@ import { randomUUID } from "node:crypto"; +import { constants as osConstants } from "node:os"; import { realpathSync, statSync } from "node:fs"; import { basename, dirname } from "node:path"; import { isPathInside } from "../../agent-runtime/path-inside.mjs"; @@ -8,6 +9,7 @@ import type { CreateSessionRequest, GitRiskReport, Point, + PluginLaunchValues, ProviderId, SessionBounds, SessionEnvironmentChoice, @@ -64,8 +66,9 @@ import { terminalFailureDetails } from "./terminalFailureDetails.ts"; import { canResumeThreadById, resolveTerminalLaunch } from "./terminalLaunch.ts"; import { isLaunchProfile, PROFILE_RANK, profileAvailable, profileCeiling, type LaunchProfile } from "../../shared/autoMode.ts"; import type { AgentIsolation, IsolationDecision } from "./isolation/AgentIsolation.ts"; -import { controlGrantFolder } from "./isolation/AgentIsolation.ts"; +import { controlGrantFolder, gatewaySocketAddresses } from "./isolation/AgentIsolation.ts"; import { auditRepositories, neutralizeRepositories, type GitRiskRepository } from "./isolation/gitAudit.ts"; +import { providerHomeVariables, type PlacedProject } from "./isolation/isolationPaths.ts"; import { LaunchRefusal } from "./launchRefusal.ts"; import { configuredMode } from "./configuredMode.ts"; import { envKey, RESERVED_ENV, type LaunchPipeline, type PreparedLaunch } from "./LaunchPipeline.ts"; @@ -126,6 +129,20 @@ interface ManagedSession { promptTurnMark?: number; /** The last turn's final answer its hook or plugin reported (captureResult only); cleared when a turn starts. */ answer?: { text: string; truncated: boolean; at: number }; + /** + * Status hooks were off when this launch was planned, or were turned off since: it has no lifecycle hooks, so its + * card stays unavailable and no answer is captured until a restart with status hooks on. + */ + statusHooksOff?: boolean; + /** Something other than Enter was typed since the last submitted input (an Enter alone then submits nothing). */ + typedSinceSubmit?: boolean; + /** Bumped by every submitted input (except an approval answer) and every launch: the turn an answer must belong to. */ + inputGeneration?: number; + /** inputGeneration when the current turn started. */ + answerTurnGeneration?: number; + /** OpenCode's current turn id and the inputGeneration each recent turn id started in. */ + providerTurnId?: string; + providerTurnGenerations?: Map; /** * Plugin options, environment ref (or, until the plugin has prepared it, the launcher's environment choice) * and owning plugin carried into the saved record. @@ -135,10 +152,21 @@ interface ManagedSession { launchToken: number; /** Removes the current run's plugin files; called when the process exits. */ launchCleanup: (() => Promise) | null; + /** Cleanup already started by an earlier exit; closing the card must drain it before deleting its parent. */ + launchCleanupPending?: Promise; + /** Whole contributed launches, including asynchronous preparation/wrapping and abandoned-run cleanup. */ + launchTasks?: Set>; + /** + * The pipelines that prepared launches for this card, each with the run folders it created for this card object, + * retained across host reconfiguration/restart. Closing removes only these, never a later card's under the same id. + */ + launchRunFolders?: Map, Set>; /** A restored grok card waits for its grid before launching; plugins still learn it is a restore. */ restoringLaunch: boolean; /** The environment was prepared or resumed in this run of the app, so it can be wrapped now. */ environmentReady: boolean; + /** Plugins whose launch contribution decided this run's model (LaunchPipeline.prepare), once prepared. */ + modelSetBy?: string[]; /** Bumped by every launch the person or the app asks for (create, restart, restore); input waits for one. */ launchEpoch: number; /** Input waiting for this launch to start (deliverInput): woken whenever the launch moves on. */ @@ -151,6 +179,17 @@ interface ManagedSession { hookSignals?: number; /** Set after the person answered a hooked Claude prompt; see settleAnsweredPrompt. */ answeredPromptTimer?: ReturnType; + /** The last deliverInput of this card: the next one waits for it, so sends arrive in order. */ + inputQueue?: Promise; + /** This launch's CLI turned its text cursor on in its full-screen view: its input field is up (watchInputReadiness). */ + inputReady?: boolean; + /** While watching for inputReady: this launch's CLI is on its alternate (full-screen) buffer. */ + alternateScreen?: boolean; + /** While watching for inputReady: an escape sequence cut off at the end of the last output chunk. */ + readinessCarry?: string; + /** When this launch's first lifecycle hook stops waiting for inputReady (INPUT_READY_HOOK_GRACE_MS after it). */ + hookReadyAt?: number; + hookReadyTimer?: ReturnType; } type EnvironmentService = Pick = new Set(["opencode"]); + +/** + * Whether input for this CLI waits for its input field. Only on macOS and Linux: Windows' pseudo console (ConPTY, + * winpty) may not pass the alternate-screen and cursor sequences through, so there the gate could never open, and + * input is typed as soon as the PTY exists, as before. + */ +function waitsForInputReady(provider: ProviderId): boolean { + return (process.platform === "darwin" || process.platform === "linux") && WAITS_FOR_INPUT_READY.has(provider); +} + +/** + * A lifecycle hook alone is not proof that the input field is up: OpenCode's plugin reports (for a resumed session, + * idle on session.updated) about 0.6 s before typed keys reach its prompt. The cursor signal decides; a hook counts + * only this long after the launch's first one, for a CLI whose cursor sequence never arrives. + */ +export const INPUT_READY_HOOK_GRACE_MS = 2_000; + /** What happened to input handed to deliverInput. */ export type InputDelivery = { delivered: true } | { delivered: false; reason: string }; @@ -246,7 +311,11 @@ export class TerminalManager { // Every text an agent reads from another card passes through it (EP-8). private redaction = new SecretRedactionRegistry(); // Where each running card was actually started (an environment may move it) and its agent config folder. - private readonly launchContexts = new Map(); + /** + * Where each running card was started: its folder, and its own config folders (CLAUDE_CONFIG_DIR, and the account + * home a launch contributor selected inside its plugin data), which base protection treats as the agent's own. + */ + private readonly launchContexts = new Map(); private modelCheck: (provider: ProviderId, model: string) => string | null = () => null; // Whether base protection is on (Settings → Agents); OpenCode's auto profile lets shell commands run without asking // only then. Unknown counts as off. @@ -362,7 +431,7 @@ export class TerminalManager { provider: session.metadata.provider, role: session.metadata.role ?? "agent", cwd: launched?.cwd ?? session.metadata.cwd, - configDirs: launched?.configDir ? [launched.configDir] : [], + configDirs: launched ? [...launched.configDirs] : [], profile: session.metadata.profile }; } @@ -654,8 +723,15 @@ export class TerminalManager { } const delegated = origin === "subagent" || control.ownerPluginId !== undefined; + // The orchestrator's own options are the person's choice: its subagent may repeat them even where the plugin did + // not declare its options delegable. Any other choice needs the plugin's delegable declaration. + // Only a subagent of the same CLI: the person chose them for that CLI (as AgentControlService.subagentLaunchOptions). + const parent = origin === "subagent" && request.parentSessionId !== undefined ? this.sessions.get(request.parentSessionId) : undefined; + const parentOptions = parent && parent.metadata.provider === request.provider + ? parent.extras.options as Record | undefined : undefined; const launchOptions = this.launchPipeline - ? this.launchPipeline.normalizeOptions(request.provider, request.launchOptions, origin === "subagent" ? { delegated: true } : undefined) + ? this.launchPipeline.normalizeOptions(request.provider, request.launchOptions, origin === "subagent" + ? { delegated: true, ...(parentOptions ? { inherited: parentOptions } : {}) } : undefined) : request.launchOptions === undefined ? undefined : failWith("Plugin launch options are not available."); // The isolation layer decides before anything starts: a subagent without it runs in normal, a launch that // needs it and cannot have it is refused. @@ -896,29 +972,59 @@ export class TerminalManager { * The one delivery rule for text another agent, a plugin or a controller sends to a card (spawn_agent's first * prompt, send_to_agent, plugin sessions.send). A running card gets it at once. A card whose launch plugins are * still preparing (launch options, a launch policy, an environment) or that waits for its grid gets it exactly - * once, when that launch has started. A launch that is refused, fails, is cancelled or superseded (closed, - * restarted), or does not start within LAUNCH_INPUT_WAIT_MS delivers nothing, says why, and drops the text: - * it never reaches a later launch of the card. An aborted `signal` (the sender cancelled) delivers nothing either. + * once, when that launch has started. A CLI that drops keys typed while it starts (WAITS_FOR_INPUT_READY) gets it + * only once its input field is up. A launch that is refused, fails, is cancelled or superseded (closed, restarted), + * or does not start or become ready within `waitMs` delivers nothing, says why, and drops the text: it never reaches + * a later launch of the card. An aborted `signal` (the sender cancelled) delivers nothing either. Sends to one card + * are delivered in the order they were made; the wait of each counts from its own call. */ - async deliverInput(id: string, data: string, waitMs = LAUNCH_INPUT_WAIT_MS, signal?: AbortSignal): Promise { - const canceled: InputDelivery = { delivered: false, reason: "The delivery was cancelled." }; - if (signal?.aborted) return canceled; + deliverInput(id: string, data: string, waitMs = LAUNCH_INPUT_WAIT_MS, signal?: AbortSignal): Promise { const session = this.sessions.get(id); - if (!session) return { delivered: false, reason: "The session does not exist." }; + if (!session) return Promise.resolve({ delivered: false, reason: "The session does not exist." }); + // The launch this call was made for: a send queued behind another never reaches a later launch either. const epoch = session.launchEpoch; const deadline = Date.now() + waitMs; - const waiting = (): boolean => this.sessions.get(id) === session && session.launchEpoch === epoch - && session.metadata.exitCode === null && !session.process; - while (waiting() && !signal?.aborted) { + const previous = session.inputQueue; + const run = (): Promise => this.deliverQueuedInput(id, session, epoch, data, waitMs, deadline, signal); + const pending = previous ? previous.then(run, run) : run(); + session.inputQueue = pending; + void pending.finally(() => { if (session.inputQueue === pending) delete session.inputQueue; }).catch(() => undefined); + if (!signal || !previous) return pending; + // A send cancelled while it waits behind another settles at once; its turn in the queue then types nothing. + const canceled: InputDelivery = { delivered: false, reason: "The delivery was cancelled." }; + if (signal.aborted) return Promise.resolve(canceled); + return new Promise((resolve) => { + const abort = (): void => resolve(canceled); + signal.addEventListener("abort", abort, { once: true }); + void pending.then(resolve, () => resolve(canceled)).finally(() => signal.removeEventListener("abort", abort)); + }); + } + + private async deliverQueuedInput( + id: string, session: ManagedSession, epoch: number, data: string, waitMs: number, deadline: number, signal?: AbortSignal + ): Promise { + const canceled: InputDelivery = { delivered: false, reason: "The delivery was cancelled." }; + if (signal?.aborted) return canceled; + // What the text still waits for: the launch to start, or (once its PTY exists) the CLI's input field. + const awaiting = (): "launch" | "input" | null => { + if (this.sessions.get(id) !== session || session.launchEpoch !== epoch || session.metadata.exitCode !== null) return null; + if (!session.process) return "launch"; + return cliAcceptsInput(session) ? null : "input"; + }; + for (let stage = awaiting(); stage && !signal?.aborted; stage = awaiting()) { const remaining = deadline - Date.now(); if (remaining <= 0) { - return { delivered: false, reason: `The session did not start within ${Math.round(waitMs / 1000)} s.` }; + // A caller that does not wait (0) is told the state, not a wait it never made. + const within = waitMs <= 0 ? "yet" : waitMs >= 1_000 ? `within ${Math.round(waitMs / 1_000)} s` : `within ${waitMs} ms`; + return { delivered: false, reason: stage === "launch" + ? `The session did not start ${within}.` + : `The agent's CLI did not show its input field ${within}; nothing was typed. Look at the card, then send the prompt again.` }; } await new Promise((resolve) => { const timer = setTimeout(wake, remaining); function wake(): void { clearTimeout(timer); - session!.launchWaiters.delete(wake); + session.launchWaiters.delete(wake); signal?.removeEventListener("abort", wake); resolve(); } @@ -941,6 +1047,23 @@ export class TerminalManager { return { delivered: true }; } + /** + * Submitted text starts a new task (an approval answer continues the current turn), so an earlier answer is not its. + * An Enter with nothing typed before it in an idle card submits nothing: the stored answer stays. + */ + private noteSubmittedInput(session: ManagedSession, data: string): void { + const typed = /[^\r\n]/u.test(data); + if (!data.endsWith("\r")) { + if (typed) session.typedSinceSubmit = true; + return; + } + const empty = !typed && !session.typedSinceSubmit; + session.typedSinceSubmit = false; + if (empty && session.metadata.status === "idle") return; + if (session.metadata.status !== "needs_approval") session.inputGeneration = (session.inputGeneration ?? 0) + 1; + delete session.answer; + } + private wakeLaunchWaiters(session: ManagedSession): void { for (const wake of [...session.launchWaiters]) wake(); } @@ -949,8 +1072,18 @@ export class TerminalManager { if (typeof data !== "string" || data.length === 0) return false; const session = this.sessions.get(id); if (!session || session.metadata.exitCode !== null || !session.process) return false; - const process = session.process; + return this.writeInput(id, session, session.process, data); + } + + /** + * The one write of person or agent input to a card's CLI, and what it means for turns and answers. Everything that + * types into a card goes through here: input(), deliverInput, and any writer that holds input and replays it later, + * so a replayed Enter is a submitted prompt like a typed one. Nothing is written to a process the card no longer runs. + */ + private writeInput(id: string, session: ManagedSession, process: IPty, data: string): boolean { + if (this.sessions.get(id) !== session || session.process !== process || session.metadata.exitCode !== null) return false; const written = tryPtyOperation(() => process.write(data)); + if (written) this.noteSubmittedInput(session, data); if (written && ANSWERS_PROMPT.test(data)) this.settleAnsweredPrompt(id, session); return written; } @@ -1021,11 +1154,53 @@ export class TerminalManager { return publicSessionMetadata(session); } - /** `source` "hook" is the agent's own lifecycle hook (through the runtime gateway); "title" is its terminal title. */ - applyProviderSignal(id: string, signal: ProviderLifecycleSignal, source: "hook" | "title" = "hook"): void { + /** + * `source` "hook" is the agent's own lifecycle hook (through the runtime gateway); "title" is its terminal title. + * Returns whether the signal belongs to the current turn and the last submitted input, so the answer it carries may + * be kept (applyRuntimeSignal). With status hooks off for this launch nothing applies and no answer is kept. + */ + applyProviderSignal(id: string, signal: ProviderLifecycleSignal, source: "hook" | "title" = "hook"): boolean { const session = this.sessions.get(id); - if (!this.lifecycleHooksEnabled || !session || session.metadata.status === "done" || session.metadata.status === "failed") return; - if (source === "hook") session.hookSignals = (session.hookSignals ?? 0) + 1; + if (!this.lifecycleHooksEnabled || !session || session.statusHooksOff) return false; + // A result-capturing OpenCode names each turn. A signal of an earlier turn is stale and changes nothing; a signal + // of the current turn that started before the last submitted input still moves the status (so a wait ends), but + // is not that input's turn: it neither counts as a turn start nor brings an answer. + const exited = session.metadata.status === "done" || session.metadata.status === "failed"; + let current = true; + let startsProviderTurn = false; + if (source === "hook" && session.captureResult && session.metadata.provider === "opencode" && signal.requestId) { + const turns = session.providerTurnGenerations ??= new Map(); + const generation = session.inputGeneration ?? 0; + const known = turns.get(signal.requestId); + if (known === undefined) { + // A turn id not seen yet opens the next turn, also when its start was lost (as the gateway counts it); after + // the exit no turn opens. + if (exited) return false; + startsProviderTurn = true; + turns.set(signal.requestId, generation); + session.providerTurnId = signal.requestId; + while (turns.size > 64) turns.delete(turns.keys().next().value!); + } else if (session.providerTurnId !== signal.requestId) { + return false; + } else { + current = known === generation; + } + } + // The process already exited: its status stays, but the answer its last turn reported as it ended still counts. + if (exited) return current; + if (source === "hook") { + session.hookSignals = (session.hookSignals ?? 0) + 1; + // The CLI runs, but its input field may not be up yet: input waiting for it (deliverInput) goes on the cursor + // signal, or once the grace after this first hook has passed. + if (session.hookSignals === 1 && waitsForInputReady(session.metadata.provider) && !session.inputReady) { + session.hookReadyAt = Date.now() + INPUT_READY_HOOK_GRACE_MS; + session.hookReadyTimer = setTimeout(() => { + session.hookReadyTimer = undefined; + this.wakeLaunchWaiters(session); + }, INPUT_READY_HOOK_GRACE_MS); + session.hookReadyTimer.unref?.(); + } + } const threadId = normalizeThreadId(session.metadata.provider, signal.threadId); const threadChanged = Boolean(threadId && threadId !== session.threadId); @@ -1036,24 +1211,44 @@ export class TerminalManager { const nextStatus = signal.state; // A turn starts when the agent moves to working; wait_for_agent compares it with the last delivered prompt. - if (nextStatus === "working" && session.metadata.status !== "working") session.turnStarts = (session.turnStarts ?? 0) + 1; - // A new turn: the previous answer is no longer this turn's. - if (nextStatus === "working") session.answer = undefined; + if (current && nextStatus === "working") { + if (startsProviderTurn || session.metadata.status !== "working") session.turnStarts = (session.turnStarts ?? 0) + 1; + // A new turn: the previous answer is no longer this turn's. + session.answer = undefined; + session.answerTurnGeneration = session.inputGeneration ?? 0; + } const completed = nextStatus === "idle" && ["Stop", "StopFailure", "StopCancelled"].includes(signal.event ?? ""); const nextTurnCompleted = nextStatus === "working" ? false : completed || Boolean(session.metadata.turnCompleted); - if (!threadChanged && session.metadata.status === nextStatus && Boolean(session.metadata.turnCompleted) === nextTurnCompleted) return; + if (!threadChanged && session.metadata.status === nextStatus && Boolean(session.metadata.turnCompleted) === nextTurnCompleted) return current; session.metadata.status = nextStatus; session.metadata.turnCompleted = nextTurnCompleted; this.emitSession(session.metadata); + return current; + } + + /** + * Why a result-capturing session's final answer cannot be captured, or null when it can (or the session does not + * capture one). With CanvasTTY status hooks off for its launch it gets no hooks, so nothing reports the answer. + */ + answerUnavailable(id: string): "status_hooks_off" | null { + const session = this.sessions.get(id); + if (!session?.captureResult) return null; + return !this.lifecycleHooksEnabled || session.statusHooksOff ? "status_hooks_off" : null; } /** * Keeps the final answer a result-capturing session reported with its turn's end (the Codex Stop hook, the OpenCode - * plugin's session.idle); read back masked by answer(). In memory only, never saved. + * plugin's session.idle), only while it belongs to the current turn and the last submitted input; read back masked + * by answer(). OpenCode passes its turn id. In memory only, never saved. */ - recordAnswer(id: string, result: { text: string; truncated: boolean }): void { + recordAnswer(id: string, result: { text: string; truncated: boolean }, correlation?: { turnId?: string | null }): void { const session = this.sessions.get(id); if (!session?.captureResult || typeof result?.text !== "string") return; + const generation = session.inputGeneration ?? 0; + if (session.metadata.provider === "opencode" && correlation && "turnId" in correlation) { + if (!correlation.turnId || correlation.turnId !== session.providerTurnId + || session.providerTurnGenerations?.get(correlation.turnId) !== generation) return; + } else if (session.promptTurnMark !== undefined && session.answerTurnGeneration !== generation) return; session.answer = { text: result.text, truncated: result.truncated === true, at: Date.now() }; } @@ -1071,6 +1266,7 @@ export class TerminalManager { if (next) return; for (const session of this.sessions.values()) { session.lifecycle = null; + session.statusHooksOff = true; if ( session.metadata.provider === "terminal" || session.metadata.status === "done" @@ -1130,6 +1326,37 @@ export class TerminalManager { } } + /** Serializes this card's run cleanup across exit/restart; callbacks retain their specific launch ownership. */ + private cleanupLaunchFiles(session: ManagedSession): Promise { + const cleanup = session.launchCleanup; + session.launchCleanup = null; + const previous = session.launchCleanupPending; + if (!cleanup) return previous ?? Promise.resolve(); + const pending = (async (): Promise => { + if (previous) await previous.catch(() => undefined); + await cleanup(); + })(); + session.launchCleanupPending = pending; + void pending.finally(() => { + if (session.launchCleanupPending === pending) delete session.launchCleanupPending; + }).catch(() => undefined); + return pending; + } + + /** Closing a card waits for producers of run files, not just the current process's adopted cleanup. */ + private async forgetLaunchFiles(session: ManagedSession): Promise { + while (session.launchTasks?.size) await Promise.allSettled([...session.launchTasks]); + try { await this.cleanupLaunchFiles(session); } + finally { + const errors: unknown[] = []; + // Different pipeline instances may share runsRoot; even their parent deletions must not overlap. + for (const [pipeline, runDirectories] of session.launchRunFolders ?? []) { + try { await pipeline.forgetSession(session.metadata.id, runDirectories); } catch (error) { errors.push(error); } + } + if (errors.length) throw new AggregateError(errors, "Launch file folders could not be removed."); + } + } + /** * Closes a card. A card in a plugin environment releases it: `keepEnvironmentData` is the person's * answer to "Keep environment data?" (kept unless they said no). Quitting releases nothing. @@ -1163,9 +1390,11 @@ export class TerminalManager { this.redaction.clear(`session:${id}`); this.releaseIsolation(id); session.launchToken += 1; - void session.launchCleanup?.().catch(() => undefined); - session.launchCleanup = null; - if (session.extras.options) void this.launchPipeline?.forgetSession(id).catch(() => undefined); + // launchToken and launch waiters were invalidated above: pending preparation/wrapping can now unwind before + // parent removal. Its original pipeline remains the owner even if the host configuration has since changed. + void this.forgetLaunchFiles(session).catch((error: unknown) => { + console.warn(`CanvasTTY could not remove the launch files of card ${id}.`, error); + }); session.agentBrowser?.cleanup(); session.agentRuntime?.cleanup(); session.agentOrchestration?.cleanup(); @@ -1569,7 +1798,10 @@ export class TerminalManager { let spawn: { command: string; args: string[] | string; env: Record } = planned; if (decision.apply) { try { - spawn = this.wrapIsolated(id, provider, profile, planned); + spawn = this.wrapIsolated(id, provider, profile, planned, planned, { + ...(contribution?.accountHome ? { accountHome: contribution.accountHome } : {}), + placedProjects: this.placedProjects(id, parentSessionId, planned.cwd, undefined) + }); } catch (error) { planned.cleanup(); return { ...none, failure: { diagnostic: `Launch refused: ${error instanceof Error ? error.message : String(error)}`, exitCode: 1 } }; @@ -1579,7 +1811,7 @@ export class TerminalManager { const process = this.spawnPty(spawn.command, spawn.args, { name: "xterm-256color", cols, rows, cwd: planned.cwd, env: spawn.env }); - this.launchContexts.set(id, { cwd: planned.cwd, configDir: planned.env.CLAUDE_CONFIG_DIR ?? null }); + this.launchContexts.set(id, launchContext(planned.cwd, provider, planned.env, contribution?.accountHome)); return { process, agentBrowser: planned.agentBrowser, @@ -1633,10 +1865,13 @@ export class TerminalManager { } /** Wraps a planned launch in the isolation layer (throws when the layer cannot start: the launch is refused). */ - private wrapIsolated(id: string, provider: ProviderId, profile: LaunchProfile, planned: { command: string; args: string[] | string; cwd: string; env: Record }): { command: string; args: string[]; env: Record } { + private wrapIsolated(id: string, provider: ProviderId, profile: LaunchProfile, planned: { command: string; args: string[] | string; cwd: string; env: Record }, + gateways: Pick, + placement: { accountHome?: string; placedProjects: readonly PlacedProject[] }): { command: string; args: string[]; env: Record } { if (!this.isolation) throw new LaunchRefusal("agent isolation is not configured; the agent was not started without it."); if (typeof planned.args === "string") throw new LaunchRefusal("a Windows batch launcher cannot run inside agent isolation."); - const grant = controlGrantFolder(planned.env); + // Only the grant CanvasTTY handed this launch: a CANVASTTY_CONTROL_CONNECTION a plugin set reopens nothing. + const grant = gateways.controlGrant; const wrapped = this.isolation.wrap({ sessionId: id, provider, @@ -1645,7 +1880,12 @@ export class TerminalManager { args: planned.args, env: planned.env, profile, - ...(grant ? { grantedPrivate: [grant] } : {}) + ...(placement.accountHome ? { accountHome: placement.accountHome } : {}), + placedProjects: placement.placedProjects, + ...(grant ? { grantedPrivate: [grant] } : {}), + // Only the sockets CanvasTTY's own gateways prepared; a CANVASTTY_*_ADDRESS from elsewhere is never bound in. + gatewaySockets: gatewaySocketAddresses(gateways.agentBrowser?.environment, gateways.agentRuntime?.environment, + gateways.agentOrchestration?.environment) }); this.releaseIsolation(id); this.isolationCleanups.set(id, wrapped.cleanup); @@ -1740,13 +1980,14 @@ export class TerminalManager { const browserEnvironment = agentBrowser?.environment ?? {}; const runtimeEnvironment = agentRuntime?.environment ?? {}; const orchestrationEnvironment = agentOrchestration?.environment ?? {}; + const hostControlEnvironment = controlEnvironment(role, this.controlConnection, id); const providerEnvironment: Record = { ...(provider === "opencode" ? mergeOpenCodeLaunchEnvironment(browserEnvironment, runtimeEnvironment) : { ...browserEnvironment, ...runtimeEnvironment }), ...orchestrationEnvironment, // Orchestrators alone learn where the control descriptor and CLI are. - ...controlEnvironment(role, this.controlConnection, id) + ...hostControlEnvironment }; // OpenCode: the project folder in its other Unicode spelling is still this folder, not an external one. if (provider === "opencode") Object.assign(providerEnvironment, openCodeProjectFolderEnvironment({ ...baseEnvironment, ...providerEnvironment }, cwd)); @@ -1794,6 +2035,7 @@ export class TerminalManager { agentBrowser, agentRuntime, agentOrchestration, + controlGrant: controlGrantFolder(hostControlEnvironment), cleanup }; } catch (error) { @@ -1802,6 +2044,20 @@ export class TerminalManager { } } + /** + * The plugin launch options this card was started with (a copy), or undefined without any. A subagent inherits them + * unless its spawn names the plugin (AgentControlService.subagentLaunchOptions). + */ + launchOptionsOf(id: string): Record | undefined { + const options = this.sessions.get(id)?.extras.options; + return options ? structuredClone(options) as Record : undefined; + } + + /** Plugins whose launch contribution decided this card's model in its current run; empty until it is prepared. */ + modelSetByOf(id: string): string[] { + return [...(this.sessions.get(id)?.modelSetBy ?? [])]; + } + /** The model and effort this card's launches ask the CLI for. */ private launchModelOf(id: string): LaunchModelChoice { const starting = this.startingModels.get(id); @@ -1834,6 +2090,30 @@ export class TerminalManager { } } + /** + * The folders CanvasTTY itself placed a launch in, the only ones the isolation layer opens inside plugin data: the + * folder an environment prepared for this card (inside that plugin's own data), or, for a card the person started + * and its subagents, the top card's folder (the person's choice, or its environment's). A subagent's own folder is + * never one by itself: an agent asked for it. A plugin's own card is placed by nothing but an environment. + */ + private placedProjects(id: string, parentSessionId: string | undefined, cwd: string, + environment: { pluginId: string } | undefined): PlacedProject[] { + if (environment) return [{ folder: cwd, pluginId: environment.pluginId }]; + if (parentSessionId === undefined) { + return (this.sessions.get(id)?.extras.ownerPluginId ?? this.startingOwners.get(id)) === undefined ? [{ folder: cwd }] : []; + } + let root: ManagedSession | undefined; + let next: string | undefined = parentSessionId; + for (let depth = 0; next !== undefined && depth < 64; depth++) { + root = this.sessions.get(next); + if (!root) return []; + next = root.metadata.parentSessionId; + } + if (!root || next !== undefined) return []; + if (root.extras.environment) return [{ folder: root.metadata.cwd, pluginId: root.extras.environment.pluginId }]; + return root.extras.ownerPluginId === undefined ? [{ folder: root.metadata.cwd }] : []; + } + private contributed(session: ManagedSession): boolean { return Boolean(session.extras.options) || Boolean(session.extras.environment) || Boolean(session.extras.environmentChoice) || this.policyApplies(session.metadata.provider); @@ -1852,7 +2132,8 @@ export class TerminalManager { return this.environments?.available(environment) ?? false; } - private unavailableLaunchPlugins(options: Record | undefined): string[] { + /** The plugins among these launch options that are disabled, removed or not trusted now. */ + unavailableLaunchPlugins(options: Record | undefined): string[] { if (!options) return []; return this.launchPipeline ? this.launchPipeline.unavailable(options) : Object.keys(options).sort(); } @@ -1872,7 +2153,11 @@ export class TerminalManager { ): void { const token = ++session.launchToken; const { metadata } = session; - void this.runContributedLaunch(id, session, token, resume, restoring, answerCaptureGrantExpiresAt) + // Each launch decides its model again; until it is prepared no plugin has. + session.modelSetBy = []; + const tasks = session.launchTasks ??= new Set(); + // Register before invoking plugin code, which may synchronously close or restart its card. + const task = Promise.resolve().then(() => this.runContributedLaunch(id, session, token, resume, restoring, answerCaptureGrantExpiresAt)) .catch((error: unknown): LaunchOutcome => { metadata.failureDetails = this.redactSecrets(`Launch refused: ${error instanceof Error ? error.message : String(error)}`); return "failed"; @@ -1886,7 +2171,9 @@ export class TerminalManager { this.emitSession(metadata, outcome === "failed" ? failureOrigin : null); this.schedulePersistence(); }) - .finally(() => this.wakeLaunchWaiters(session)); + .finally(() => { tasks.delete(task); this.wakeLaunchWaiters(session); }); + tasks.add(task); + void task.catch(() => undefined); } private async runContributedLaunch( @@ -1905,6 +2192,8 @@ export class TerminalManager { return "failed"; }; const environments = this.environments; + // The launch starts a microtask after it was registered; a card closed or restarted in that same tick is done. + if (!live()) return "superseded"; // 1. Place a new session where the person chose. The choice stays saved with the card until the plugin has // prepared it, so a card whose preparation was cut short (quit, crash) or failed never restores locally. @@ -1946,142 +2235,148 @@ export class TerminalManager { // 3. Chosen launch contributors, and the launch policies that apply. let contribution: LaunchContribution | null = null; - const trustedFolder = session.extras.environment ? undefined : this.personTrustedFolder(metadata.parentSessionId, metadata.cwd); - if (session.extras.options || this.policyApplies(metadata.provider)) { - const pipeline = this.launchPipeline; - if (!pipeline) return refuse(missingLaunchPlugins(Object.keys(session.extras.options ?? {}))); - const placedIn = session.extras.environment; - const prepared = await pipeline.prepare({ - sessionId: id, - provider: metadata.provider, - profile: metadata.profile, - role: metadata.role, - cwd: metadata.cwd, - ...(metadata.parentSessionId !== undefined ? { parentSessionId: metadata.parentSessionId } : {}), - restoring, - resume: resume !== null, - options: structuredClone(session.extras.options ?? {}) as Record>, - environment: placedIn ? { pluginId: placedIn.pluginId, kind: placedIn.kind } : null, - ...(trustedFolder ? { trustedFolder } : {}) - }); - if (!live()) { - if (prepared.ok) void prepared.cleanup().catch(() => undefined); - return "superseded"; + let adopted = false; + let plannedCleanup: (() => void) | undefined; + try { + const trustedFolder = session.extras.environment ? undefined : this.personTrustedFolder(metadata.parentSessionId, metadata.cwd); + if (session.extras.options || this.policyApplies(metadata.provider)) { + const pipeline = this.launchPipeline; + if (!pipeline) return refuse(missingLaunchPlugins(Object.keys(session.extras.options ?? {}))); + // This pipeline may write run files for the card, so closing it later removes its folder there. + const runFolders = session.launchRunFolders ??= new Map(); + const ownFolders = runFolders.get(pipeline) ?? new Set(); + runFolders.set(pipeline, ownFolders); + const placedIn = session.extras.environment; + const prepared = await pipeline.prepare({ + sessionId: id, + provider: metadata.provider, + profile: metadata.profile, + role: metadata.role, + cwd: metadata.cwd, + ...(metadata.parentSessionId !== undefined ? { parentSessionId: metadata.parentSessionId } : {}), + restoring, + resume: resume !== null, + options: structuredClone(session.extras.options ?? {}) as Record>, + environment: placedIn ? { pluginId: placedIn.pluginId, kind: placedIn.kind } : null, + ...(trustedFolder ? { trustedFolder } : {}) + }, { ...(this.launchModelOf(id).model !== undefined ? { model: this.launchModelOf(id).model } : {}) }); + if (prepared.ok) { + contribution = prepared; + ownFolders.add(prepared.runDirectory); + } + if (!live()) return "superseded"; + if (!prepared.ok) return refuse(prepared.reason); + session.modelSetBy = prepared.modelSetBy; + this.addLaunchSecrets(session, prepared.secrets); } - if (!prepared.ok) return refuse(prepared.reason); - contribution = prepared; - this.addLaunchSecrets(session, prepared.secrets); - } - const dropContribution = (): void => { - void contribution?.cleanup().catch(() => undefined); - }; - // 4. What the environment keeps of CanvasTTY's protection, and the isolation layer for this launch. - const keeps = environment ? this.environments?.keeps?.(environment) ?? {} : {}; - if (environment && keeps.launch !== true && metadata.profile !== "normal") { - dropContribution(); - return refuse(`${environment.label} does not pass the launch on unchanged (the plugin does not declare it), so the ${metadata.profile} profile's settings and CanvasTTY's hooks would not reach the agent there. Launch it in normal, or use an environment that keeps them.`); - } - const decision = this.launchIsolation(id, metadata.provider, metadata.profile, metadata.role, - environment ? { isolated: keeps.isolated === true, label: environment.label } : null); - if (decision.refuse) { - dropContribution(); - return refuse(decision.refuse); - } - if (environment && keeps.launch !== true) { - metadata.isolation = { ...(metadata.isolation ?? { state: "environment" }), - reason: `${metadata.isolation?.reason ? `${metadata.isolation.reason} ` : ""}Base protection and CanvasTTY's hooks do not reach the agent in ${environment.label}.` }; - } + // 4. What the environment keeps of CanvasTTY's protection, and the isolation layer for this launch. + const keeps = environment ? this.environments?.keeps?.(environment) ?? {} : {}; + if (environment && keeps.launch !== true && metadata.profile !== "normal") { + return refuse(`${environment.label} does not pass the launch on unchanged (the plugin does not declare it), so the ${metadata.profile} profile's settings and CanvasTTY's hooks would not reach the agent there. Launch it in normal, or use an environment that keeps them.`); + } + const decision = this.launchIsolation(id, metadata.provider, metadata.profile, metadata.role, + environment ? { isolated: keeps.isolated === true, label: environment.label } : null); + if (decision.refuse) { + return refuse(decision.refuse); + } + if (environment && keeps.launch !== true) { + metadata.isolation = { ...(metadata.isolation ?? { state: "environment" }), + reason: `${metadata.isolation?.reason ? `${metadata.isolation.reason} ` : ""}Base protection and CanvasTTY's hooks do not reach the agent in ${environment.label}.` }; + } - // 5. The host spawns the PTY; an environment only rewrites what is spawned, and the isolation layer wraps that. - let planned: PlannedSpawn | { failure: LaunchFailure }; - try { - planned = this.planSpawn(id, metadata.provider, decision.profile, metadata.cwd, resume, - session.captureResult, metadata.role, answerCaptureGrantExpiresAt, contribution, trustedFolder, Boolean(environment), decision.apply); - } catch (error) { - dropContribution(); - metadata.failureDetails = this.redactSecrets(error instanceof Error ? error.message : String(error)); - return "failed"; - } - if ("failure" in planned) { - dropContribution(); - applyLaunchFailure(metadata, planned.failure); - return "failed"; - } - const abandon = (): void => { - planned.cleanup(); - dropContribution(); - }; - this.noteConfiguredMode(id, metadata.provider, decision.profile, planned.env, planned.cwd); - let spawn: { command: string; args: string[] | string; cwd: string; env: Record } = planned; - if (environment && environments) { - if (typeof planned.args === "string") { - abandon(); - return refuse("this provider's Windows batch launcher cannot run in a plugin environment."); + // 5. The host spawns the PTY; an environment only rewrites what is spawned, and the isolation layer wraps that. + let planned: PlannedSpawn | { failure: LaunchFailure }; + // Whether this launch gets lifecycle hooks is decided as it is planned; the wrap below may take a while. + const statusHooks = this.lifecycleHooksEnabled; + try { + planned = this.planSpawn(id, metadata.provider, decision.profile, metadata.cwd, resume, + session.captureResult, metadata.role, answerCaptureGrantExpiresAt, contribution, trustedFolder, Boolean(environment), decision.apply); + } catch (error) { + metadata.failureDetails = this.redactSecrets(error instanceof Error ? error.message : String(error)); + return "failed"; } - const secretValues = new Set(contribution?.secrets ?? []); - const secretEnvNames = Object.keys(contribution?.env ?? {}).filter((key) => secretValues.has(contribution!.env[key]!)); - // The environment sees the launch's own variables, never CanvasTTY's reserved ones or secret values. - const visible = Object.fromEntries(Object.entries(planned.launchEnvironment) - .filter(([key]) => !RESERVED_ENV.test(key) && !secretEnvNames.includes(key))); - const wrapped = await environments.wrap(environment, { - sessionId: id, - provider: metadata.provider, - launch: { command: planned.command, args: planned.args, env: visible, cwd: planned.cwd }, - secretEnvNames, - takenEnv: new Set(Object.keys(planned.launchEnvironment)), - path: launchSearchPath(planned.env) - }); - if (!live()) { - abandon(); - return "superseded"; + if ("failure" in planned) { + applyLaunchFailure(metadata, planned.failure); + return "failed"; } - if (!wrapped.ok) { - abandon(); - return refuse(wrapped.reason); + plannedCleanup = planned.cleanup; + this.noteConfiguredMode(id, metadata.provider, decision.profile, planned.env, planned.cwd); + let spawn: { command: string; args: string[] | string; cwd: string; env: Record } = planned; + if (environment && environments) { + if (typeof planned.args === "string") { + return refuse("this provider's Windows batch launcher cannot run in a plugin environment."); + } + const secretValues = new Set(contribution?.secrets ?? []); + const secretEnvNames = Object.keys(contribution?.env ?? {}).filter((key) => secretValues.has(contribution!.env[key]!)); + // The environment sees the launch's own variables, never CanvasTTY's reserved ones or secret values. + const visible = Object.fromEntries(Object.entries(planned.launchEnvironment) + .filter(([key]) => !RESERVED_ENV.test(key) && !secretEnvNames.includes(key))); + const wrapped = await environments.wrap(environment, { + sessionId: id, + provider: metadata.provider, + launch: { command: planned.command, args: planned.args, env: visible, cwd: planned.cwd }, + secretEnvNames, + takenEnv: new Set(Object.keys(planned.launchEnvironment)), + path: launchSearchPath(planned.env) + }); + if (!live()) { + return "superseded"; + } + if (!wrapped.ok) { + return refuse(wrapped.reason); + } + this.addLaunchSecrets(session, wrapped.secrets); + spawn = { command: wrapped.command, args: wrapped.args, cwd: wrapped.cwd, + env: { ...planned.env, ...(wrapped.cwd !== planned.cwd ? { PWD: wrapped.cwd } : {}), ...wrapped.env } }; } - this.addLaunchSecrets(session, wrapped.secrets); - spawn = { command: wrapped.command, args: wrapped.args, cwd: wrapped.cwd, - env: { ...planned.env, ...(wrapped.cwd !== planned.cwd ? { PWD: wrapped.cwd } : {}), ...wrapped.env } }; - } - if (decision.apply) { + if (decision.apply) { + try { + spawn = { ...this.wrapIsolated(id, metadata.provider, decision.profile, spawn, planned, { + ...(contribution?.accountHome ? { accountHome: contribution.accountHome } : {}), + placedProjects: this.placedProjects(id, metadata.parentSessionId, spawn.cwd, environment) + }), cwd: spawn.cwd }; + } catch (error) { + return refuse(error instanceof Error ? error.message : String(error)); + } + } + let process: IPty; try { - spawn = { ...this.wrapIsolated(id, metadata.provider, decision.profile, spawn), cwd: spawn.cwd }; + process = this.spawnPty(spawn.command, spawn.args, { + name: "xterm-256color", cols: session.cols, rows: session.rows, cwd: spawn.cwd, env: spawn.env + }); } catch (error) { - abandon(); - return refuse(error instanceof Error ? error.message : String(error)); + this.releaseIsolation(id); + metadata.failureDetails = this.redactSecrets(error instanceof Error ? error.message : String(error)); + return "failed"; + } + session.process = process; + if (planned.launchEnvironment.CANVASTTY_CODEX_KEYBOARD) { + metadata.nativeEditor = JSON.parse(planned.launchEnvironment.CANVASTTY_CODEX_KEYBOARD); + } else { + delete metadata.nativeEditor; + } + this.launchContexts.set(id, launchContext(spawn.cwd, metadata.provider, spawn.env, contribution?.accountHome)); + session.agentBrowser = planned.agentBrowser; + session.agentRuntime = planned.agentRuntime; + session.agentOrchestration = planned.agentOrchestration; + session.launchCleanup = contribution?.cleanup ?? null; + adopted = true; + metadata.status = initialSessionStatus(metadata.provider); + metadata.exitCode = null; + metadata.failureDetails = null; + this.bindProcess(id, session, process, statusHooks); + const runtimeStatus = this.agentRuntime?.currentStatus(id); + if (runtimeStatus) metadata.status = runtimeStatus; + if (environment) this.describeEnvironment(id, session, environment); + return "launched"; + } finally { + // Unadopted resources remain owned by this tracked launch, including throws during plugin wrapping. + if (!adopted) { + try { plannedCleanup?.(); } + finally { await contribution?.cleanup().catch(() => undefined); } } } - let process: IPty; - try { - process = this.spawnPty(spawn.command, spawn.args, { - name: "xterm-256color", cols: session.cols, rows: session.rows, cwd: spawn.cwd, env: spawn.env - }); - } catch (error) { - abandon(); - this.releaseIsolation(id); - metadata.failureDetails = this.redactSecrets(error instanceof Error ? error.message : String(error)); - return "failed"; - } - session.process = process; - if (planned.launchEnvironment.CANVASTTY_CODEX_KEYBOARD) { - metadata.nativeEditor = JSON.parse(planned.launchEnvironment.CANVASTTY_CODEX_KEYBOARD); - } else { - delete metadata.nativeEditor; - } - this.launchContexts.set(id, { cwd: spawn.cwd, configDir: spawn.env.CLAUDE_CONFIG_DIR ?? null }); - session.agentBrowser = planned.agentBrowser; - session.agentRuntime = planned.agentRuntime; - session.agentOrchestration = planned.agentOrchestration; - session.launchCleanup = contribution?.cleanup ?? null; - metadata.status = initialSessionStatus(metadata.provider); - metadata.exitCode = null; - metadata.failureDetails = null; - this.bindProcess(id, session, process); - const runtimeStatus = this.agentRuntime?.currentStatus(id); - if (runtimeStatus) metadata.status = runtimeStatus; - if (environment) this.describeEnvironment(id, session, environment); - return "launched"; } private addLaunchSecrets(session: ManagedSession, secrets: readonly string[]): void { @@ -2099,12 +2394,21 @@ export class TerminalManager { }).catch(() => undefined); } - private bindProcess(id: string, session: ManagedSession, process: IPty): void { + /** + * `statusHooks`: status hooks were on when this launch was planned (and so got its lifecycle hooks). A launch planned + * and spawned in one step passes nothing; a contributed launch passes what held before its asynchronous steps. + */ + private bindProcess(id: string, session: ManagedSession, process: IPty, statusHooks = this.lifecycleHooksEnabled): void { + if (statusHooks && this.lifecycleHooksEnabled) delete session.statusHooksOff; + else session.statusHooksOff = true; session.agentBrowser?.retainUntilExit?.(); process.onData((data) => { const current = this.sessions.get(id); if (!current || current !== session || current.process !== process) return; + if (!current.inputReady && waitsForInputReady(current.metadata.provider) && watchInputReadiness(current, data)) { + this.wakeLaunchWaiters(current); + } const lifecycleState = current.lifecycle?.push(data); if (lifecycleState) current.titleState = lifecycleState; if (lifecycleState && !titleDefersToHooks(current, lifecycleState)) { @@ -2118,38 +2422,46 @@ export class TerminalManager { let exited!: () => void; this.liveProcesses.set(process, new Promise((resolve) => { exited = resolve; })); - process.onExit(({ exitCode }) => { + process.onExit(({ exitCode, signal }) => { this.liveProcesses.delete(process); exited(); const current = this.sessions.get(id); if (!current || current !== session || current.process !== process) return; // node-pty calls this from a native callback that aborts the whole app when JavaScript throws in it. try { - this.recordExit(id, current, exitCode); + this.recordExit(id, current, exitCode, signal); } catch (error) { console.warn(`PTY ${id} exit could not be recorded.`, error); } }); } - private recordExit(id: string, current: ManagedSession, exitCode: number): void { + private recordExit(id: string, current: ManagedSession, reportedExitCode: number, signal?: number): void { this.flushOutput(id, current); this.releaseIsolation(id); if (current.extras.gitAuditSince !== undefined) void this.auditGit(id, current); + // node-pty reports a process killed by a signal as exitCode 0 plus the signal. That is a crash, not a clean finish: + // record it the way a shell does, as 128+signal, so the card, wait_for_agent and the saved session all say failed. + const killedBy = typeof signal === "number" && signal > 0 ? signal : 0; + const exitCode = killedBy ? 128 + killedBy : reportedExitCode; current.metadata.exitCode = exitCode; current.metadata.status = exitCode === 0 ? "done" : "failed"; - current.metadata.failureDetails = exitCode === 0 + const details = exitCode === 0 ? null : terminalFailureDetails(this.redactSecrets(current.bufferChunks.slice(current.bufferStart).join(""))); + current.metadata.failureDetails = killedBy + ? `${signalExitText(killedBy)}${details ? `\n${details}` : ""}`.slice(0, 8_000) + : details; current.agentBrowser?.cleanup(); current.agentBrowser = null; current.agentRuntime?.cleanup(); current.agentRuntime = null; current.agentOrchestration?.cleanup(); current.agentOrchestration = null; - void current.launchCleanup?.().catch(() => undefined); - current.launchCleanup = null; + void this.cleanupLaunchFiles(current).catch(() => undefined); this.emitSession(current.metadata); + // Input waiting for this CLI to become ready learns at once that it never will. + this.wakeLaunchWaiters(current); // Recorded at the moment of exit, so a finished agent is never relaunched. this.schedulePersistence(); } @@ -2250,12 +2562,64 @@ function resetLaunchSignals(session: ManagedSession): void { delete session.answer; delete session.turnStarts; delete session.promptTurnMark; + // A late answer of the ended launch never belongs to the new one. + session.inputGeneration = (session.inputGeneration ?? 0) + 1; + delete session.statusHooksOff; + delete session.typedSinceSubmit; + delete session.answerTurnGeneration; + delete session.providerTurnId; + delete session.providerTurnGenerations; delete session.hookSignals; delete session.titleState; + delete session.inputReady; + delete session.alternateScreen; + delete session.readinessCarry; + delete session.hookReadyAt; + if (session.hookReadyTimer) clearTimeout(session.hookReadyTimer); + session.hookReadyTimer = undefined; if (session.answeredPromptTimer) clearTimeout(session.answeredPromptTimer); session.answeredPromptTimer = undefined; } +/** Whether typed text reaches this launch's CLI now (see WAITS_FOR_INPUT_READY). */ +function cliAcceptsInput(session: ManagedSession): boolean { + return !waitsForInputReady(session.metadata.provider) || session.inputReady === true + || (session.hookReadyAt !== undefined && Date.now() >= session.hookReadyAt); +} + +/** DEC private modes set or reset (CSI ? Pm h / l): terminal control sequences, not text the CLI shows. */ +const DEC_PRIVATE_MODES = /\x1b\[\?([\d;]+)([hl])/gu; +/** The longest DEC private mode sequence kept from the end of one chunk for the next. */ +const READINESS_CARRY_CHARS = 32; + +/** + * Reads only terminal control sequences of one output chunk: the alternate screen (modes 1049, 1047, 47) and the text + * cursor (DECTCEM, mode 25). A full-screen CLI hides its cursor while it draws its start-up and shows it once its + * input field has focus; that moment is when typed keys start to reach that field. True once that happened; a + * sequence cut at the chunk's end is completed by the next chunk. + */ +function watchInputReadiness(session: ManagedSession, data: string): boolean { + const text = (session.readinessCarry ?? "") + data; + let consumed = 0; + for (const match of text.matchAll(DEC_PRIVATE_MODES)) { + consumed = match.index + match[0].length; + const set = match[2] === "h"; + for (const mode of match[1]!.split(";")) { + if (mode === "1049" || mode === "1047" || mode === "47") session.alternateScreen = set; + else if (mode === "25" && set && session.alternateScreen) session.inputReady = true; + } + if (session.inputReady) break; + } + if (session.inputReady) { + delete session.readinessCarry; + delete session.alternateScreen; + return true; + } + const escape = text.lastIndexOf("\x1b"); + session.readinessCarry = escape >= consumed && text.length - escape < READINESS_CARRY_CHARS ? text.slice(escape) : ""; + return false; +} + function titleDefersToHooks(session: ManagedSession, state: "idle" | "working" | "needs_approval"): boolean { return state !== "working" && session.metadata.provider === "claude" && session.hookSignals !== undefined; } @@ -2474,3 +2838,21 @@ function appendScrollback(session: ManagedSession, data: string): void { session.bufferStart = 0; } } + +/** "The process was killed by signal SIGKILL (9)." from node-pty's signal number. */ +function signalExitText(signal: number): string { + const name = Object.entries(osConstants.signals).find(([, value]) => value === signal)?.[0]; + return `The process was killed by signal ${name ? `${name} (${signal})` : signal}.`; +} + +/** + * A started card's folder and own config folders. The selected account home counts only while the CLI is still + * started with it (an environment may have replaced the variable): then it is that card's, and only that one's. + */ +function launchContext(cwd: string, provider: ProviderId, env: Readonly>, accountHome: string | undefined): { cwd: string; configDirs: string[] } { + const configDirs = env.CLAUDE_CONFIG_DIR ? [env.CLAUDE_CONFIG_DIR] : []; + const started = accountHome !== undefined && providerHomeVariables(provider) + .some((name) => typeof env[name] === "string" && env[name]!.replace(/[\\/]+$/u, "") === accountHome); + if (started && !configDirs.includes(accountHome)) configDirs.push(accountHome); + return { cwd, configDirs }; +} diff --git a/src/main/services/UsageHistoryService.ts b/src/main/services/UsageHistoryService.ts index c57fc16d4..0b0859127 100644 --- a/src/main/services/UsageHistoryService.ts +++ b/src/main/services/UsageHistoryService.ts @@ -1,6 +1,6 @@ import { randomUUID } from "node:crypto"; import { constants } from "node:fs"; -import { copyFile, lstat, open, rename, unlink, type FileHandle } from "node:fs/promises"; +import { lstat, open, rename, unlink, writeFile, type FileHandle } from "node:fs/promises"; import { join } from "node:path"; import { isDeepStrictEqual } from "node:util"; import type { LimitProviderId, LimitsSnapshot } from "../../shared/contracts.ts"; @@ -114,19 +114,21 @@ export class UsageHistoryService { async load(): Promise { const now = this.now(); this.startedAt = now; - let raw: string; + let bytes: Buffer; let handle: FileHandle | undefined; try { - // Windows lacks O_NOFOLLOW; check the path before and after opening too. - if ((await lstat(this.filePath)).isSymbolicLink()) throw Object.assign(new Error("symlink"), { code: "ELOOP" }); + // Windows lacks O_NOFOLLOW; check the path before and after opening too. File ids are compared as + // bigints: 64-bit NTFS ids lose precision as numbers. + const checked = await lstat(this.filePath, { bigint: true }); + if (checked.isSymbolicLink()) throw Object.assign(new Error("symlink"), { code: "ELOOP" }); handle = await open(this.filePath, constants.O_RDONLY | (constants.O_NOFOLLOW ?? 0)); - const opened = await handle.stat(); - const current = await lstat(this.filePath); - if (current.isSymbolicLink() || opened.dev !== current.dev || opened.ino !== current.ino) { + const opened = await handle.stat({ bigint: true }); + const current = await lstat(this.filePath, { bigint: true }); + if (current.isSymbolicLink() || [checked, current].some((path) => path.dev !== opened.dev || path.ino !== opened.ino)) { throw Object.assign(new Error("history path changed"), { code: "ELOOP" }); } if (!opened.isFile()) throw Object.assign(new Error("not a file"), { code: "ENOTFILE" }); - raw = await handle.readFile("utf8"); + bytes = await handle.readFile(); } catch (error) { if (errorCode(error) === "ENOENT") return; this.block(`Usage history file could not be read (${errorCode(error)}); it was left untouched and new history is not being saved.`); @@ -137,7 +139,7 @@ export class UsageHistoryService { let parsed: unknown; try { - parsed = JSON.parse(raw); + parsed = JSON.parse(bytes.toString("utf8")); } catch { await this.quarantine("invalid JSON", now); return; @@ -160,7 +162,8 @@ export class UsageHistoryService { // The cleaned history will replace the file, so the original is copied aside first. const name = preservedName(now); try { - await copyFile(this.filePath, join(this.directory, name), constants.COPYFILE_EXCL); + // The bytes read through the checked handle, not a second read by path that a swap could redirect. + await writeFile(join(this.directory, name), bytes, { flag: "wx", mode: 0o600 }); } catch (error) { this.block(`Usage history contains ${restored.invalid} invalid records and could not be preserved (${errorCode(error)}); it was left untouched and new history is not being saved.`); return; diff --git a/src/main/services/agent-browser/OrchestrationTools.ts b/src/main/services/agent-browser/OrchestrationTools.ts index ba6afebf3..28566b785 100644 --- a/src/main/services/agent-browser/OrchestrationTools.ts +++ b/src/main/services/agent-browser/OrchestrationTools.ts @@ -6,6 +6,7 @@ import { PromptNotDeliveredError, type AgentControlService, type SpawnAgentReque import { LaunchRefusal } from "../launchRefusal.ts"; import type { PluginAgentTools } from "../PluginAgentTools.ts"; import { + CLIENT_SAFE_AGENT_WAIT_SECONDS, DEFAULT_AGENT_WAIT_SECONDS, MAX_AGENT_WAIT_SECONDS, ORCHESTRATION_TOOL_DEFINITIONS, @@ -22,6 +23,13 @@ import type { McpToolDefinition } from "../../../agent-browser/orchestration-cat * a foreign session id is a protocol error, never a filtered result, so an * orchestrator cannot probe sessions it does not own. */ +/** + * How long spawn_agent (its checks, the launch and the CLI's readiness together) and send_to_agent wait for a prompt + * to reach the agent: under the about 60 s after which most MCP clients end a tool call, so the orchestrator gets + * the reason (and the session id) instead of a client timeout. + */ +export const AGENT_TOOL_DELIVERY_MS = 45_000; + export class ScopedOrchestrationHandler implements OrchestrationCommandHandler { private readonly control: AgentControlService; private readonly plugins: Pick | null; @@ -65,9 +73,9 @@ export class ScopedOrchestrationHandler implements OrchestrationCommandHandler { case "send_to_agent": return await this.send(sessionId, request.arguments, signal); case "observe_agent": - return this.observe(sessionId, request.arguments); + return await this.observe(sessionId, request.arguments); case "get_agent_result": - return this.result(sessionId, request.arguments); + return await this.result(sessionId, request.arguments); case "cancel_agent": return this.cancel(sessionId, request.arguments); case "list_agents": @@ -118,12 +126,15 @@ export class ScopedOrchestrationHandler implements OrchestrationCommandHandler { } this.requireOwned(orchestratorId, target); const seconds = typeof args.timeoutSeconds === "number" ? args.timeoutSeconds : DEFAULT_AGENT_WAIT_SECONDS; + // Shorter than the orchestrator's MCP client lets a tool call run, so it gets "timeout" and progress, not an error. + const limit = this.control.status(orchestratorId).provider === "claude" ? MAX_AGENT_WAIT_SECONDS : CLIENT_SAFE_AGENT_WAIT_SECONDS; try { const result = await this.control.waitFor(target, { - timeoutMs: Math.min(MAX_AGENT_WAIT_SECONDS, Math.max(1, seconds)) * 1_000, + timeoutMs: Math.min(limit, Math.max(1, seconds)) * 1_000, ...(signal ? { signal } : {}) }); - return { ...result }; + return { ...result, ...(result.reason === "timeout" + ? { message: "Still running. Read output for its progress and call wait_for_agent again." } : {}) }; } catch (error) { if (signal?.aborted || (error instanceof Error && error.name === "AbortError")) throw canceledError(); throw error; @@ -131,6 +142,7 @@ export class ScopedOrchestrationHandler implements OrchestrationCommandHandler { } private async spawn(orchestratorId: string, args: Record, signal?: AbortSignal): Promise> { + const deadline = Date.now() + AGENT_TOOL_DELIVERY_MS; if (!(AGENT_PROVIDERS as readonly unknown[]).includes(args.provider)) { throw orchestrationBridgeError("INVALID_REQUEST", unknownProviderMessage(args.provider), false); } @@ -146,17 +158,26 @@ export class ScopedOrchestrationHandler implements OrchestrationCommandHandler { if (unknown) throw orchestrationBridgeError("INVALID_REQUEST", unknown, false); } if (signal?.aborted) throw canceledError(); + // A plugin the spawn does not name keeps this session's own options (never silently the CLI's default model). + const requested = args.launchOptions as SpawnAgentRequest["launchOptions"]; + let inheritance: ReturnType; + try { + inheritance = this.control.subagentLaunchOptions(orchestratorId, provider, requested, args.model as string | undefined); + } catch (error) { + throw orchestrationBridgeError("INVALID_REQUEST", error instanceof Error ? error.message : String(error), false); + } + const launchOptions = inheritance.launchOptions; const created = await this.control.spawn({ parentSessionId: orchestratorId, provider: args.provider as never, cwd: args.cwd as string, ...(args.title !== undefined ? { title: args.title as string } : {}), ...(args.prompt !== undefined ? { initialPrompt: args.prompt as string } : {}), - ...(args.launchOptions !== undefined ? { launchOptions: args.launchOptions as SpawnAgentRequest["launchOptions"] } : {}), + ...(launchOptions !== undefined ? { launchOptions } : {}), ...(args.model !== undefined ? { model: args.model as string } : {}), ...(args.effort !== undefined ? { effort: args.effort as SpawnAgentRequest["effort"] } : {}), profile: profile.profile - }, signal).catch((error: unknown) => { + }, signal, Math.max(0, deadline - Date.now())).catch((error: unknown) => { if (signal?.aborted) throw canceledError(); throw error; }); @@ -177,7 +198,10 @@ export class ScopedOrchestrationHandler implements OrchestrationCommandHandler { profile: created.profile, ...(profile.inherited ? { profileInherited: true } : {}), ...(created.model !== undefined ? { model: created.model } : {}), - ...(created.effort !== undefined ? { effort: created.effort } : {}) + ...(created.effort !== undefined ? { effort: created.effort } : {}), + // Which CLI serves it, the model it was asked for (null: its launch options or the CLI's own default decide), + // and where each plugin's launch options came from. + servedBy: { provider: created.provider, model: created.model ?? null, launchOptions: inheritance.sources } }; } @@ -188,7 +212,8 @@ export class ScopedOrchestrationHandler implements OrchestrationCommandHandler { args.sessionId as string, args.prompt as string, args.submit === undefined ? true : Boolean(args.submit), - signal + signal, + AGENT_TOOL_DELIVERY_MS ); } catch (error) { if (signal?.aborted) throw canceledError(); @@ -198,9 +223,9 @@ export class ScopedOrchestrationHandler implements OrchestrationCommandHandler { return { sessionId: args.sessionId as string, sent: true }; } - private observe(orchestratorId: string, args: Record): Record { + private async observe(orchestratorId: string, args: Record): Promise> { this.requireOwned(orchestratorId, args.sessionId as string); - const observation = this.control.observe( + const observation = await this.control.observe( args.sessionId as string, args.maxChars as number | undefined ); @@ -213,9 +238,9 @@ export class ScopedOrchestrationHandler implements OrchestrationCommandHandler { }; } - private result(orchestratorId: string, args: Record): Record { + private async result(orchestratorId: string, args: Record): Promise> { this.requireOwned(orchestratorId, args.sessionId as string); - const result = this.control.result(args.sessionId as string); + const result = await this.control.result(args.sessionId as string); return { sessionId: result.sessionId, state: result.state, @@ -223,7 +248,8 @@ export class ScopedOrchestrationHandler implements OrchestrationCommandHandler { exitCode: result.exitCode, output: result.output, ...(result.exitLines ? { exitLines: result.exitLines } : {}), - ...(result.answer ? { answer: result.answer } : {}) + ...(result.answer ? { answer: result.answer } : {}), + ...(result.answerUnavailable ? { answerUnavailable: result.answerUnavailable } : {}) }; } diff --git a/src/main/services/agent-runtime/AgentRuntimeBridge.ts b/src/main/services/agent-runtime/AgentRuntimeBridge.ts index 18b6ef2d2..739382f89 100644 --- a/src/main/services/agent-runtime/AgentRuntimeBridge.ts +++ b/src/main/services/agent-runtime/AgentRuntimeBridge.ts @@ -47,6 +47,11 @@ export interface AgentRuntimeBridgeOptions extends ProviderRuntimeLaunchOptions wantsDecisions?(provider: Exclude): boolean; /** The longest decision budget for this agent (ms); the session's gate deadlines are sized from it. */ decisionBudgetMs?(provider: Exclude): number; + /** + * Whether the decision hook also sees read-only file tools (base protection on): only then does a read wait on the + * gateway, and fail closed without it. + */ + decisionReads?(): boolean; /** Whether a Claude launch may use HTTP lifecycle hooks; without it every launch uses the command helper. */ claudeHttpHooks?(facts: ClaudeHttpLaunchFacts): ClaudeHttpVerdict; } @@ -60,12 +65,14 @@ export class AgentRuntimeBridge implements AgentRuntimeLaunchCoordinator { private coreHooksEnabled: boolean; private readonly wantsDecisions: AgentRuntimeBridgeOptions["wantsDecisions"]; private readonly decisionBudgetMs: AgentRuntimeBridgeOptions["decisionBudgetMs"]; + private readonly decisionReads: AgentRuntimeBridgeOptions["decisionReads"]; private readonly claudeHttpHooks: AgentRuntimeBridgeOptions["claudeHttpHooks"]; constructor(gateway: RuntimeGateway, options: AgentRuntimeBridgeOptions) { this.gateway = gateway; this.wantsDecisions = options.wantsDecisions; this.decisionBudgetMs = options.decisionBudgetMs; + this.decisionReads = options.decisionReads; this.claudeHttpHooks = options.claudeHttpHooks; this.providers = new ProviderRuntimeLaunchAdapters(options); this.coreHooksEnabled = options.coreHooksEnabled !== false; @@ -78,6 +85,8 @@ export class AgentRuntimeBridge implements AgentRuntimeLaunchCoordinator { && this.providers.decisionsSupported(input.provider); let budgetMs: number | undefined; try { budgetMs = decisions ? this.decisionBudgetMs?.(input.provider) : undefined; } catch { budgetMs = undefined; } + let reads = false; + try { reads = decisions && this.decisionReads?.() === true; } catch { reads = false; } const capability = this.coreHooksEnabled || decisions ? this.gateway.registerSession( input.terminalSessionId, @@ -92,7 +101,7 @@ export class AgentRuntimeBridge implements AgentRuntimeLaunchCoordinator { let prepared; try { prepared = this.providers.prepare(input.provider, input.terminalSessionId, this.coreHooksEnabled, decisions, budgetMs, - httpHookBase ?? undefined); + httpHookBase ?? undefined, reads); } catch (error) { if (capability) this.gateway.revokeTerminalSession(input.terminalSessionId, capability.capabilityToken); throw error; diff --git a/src/main/services/agent-runtime/ProviderRuntimeLaunch.ts b/src/main/services/agent-runtime/ProviderRuntimeLaunch.ts index a945496fe..c9fca1edd 100644 --- a/src/main/services/agent-runtime/ProviderRuntimeLaunch.ts +++ b/src/main/services/agent-runtime/ProviderRuntimeLaunch.ts @@ -26,6 +26,7 @@ import { CLAUDE_HTTP_HOOK, DECISION_BUDGET_ENV, OPENCODE_DECISIONS_ENV, + OPENCODE_DECISION_READS_ENV, permissionGateTimings } from "../../../agent-runtime/runtime-protocol.mjs"; @@ -175,6 +176,7 @@ export class ProviderRuntimeLaunchAdapters { * `decisionBudgetMs` (a decision service's `decide.timeoutMs`) lengthens the hook's deadlines to fit it. * `claudeHttpHookBase` (Claude Code only) sends its lifecycle events, except SessionStart, as HTTP hooks to the * gateway's loopback listener instead of running the helper; the decision hook stays a command. + * `decisionReads` (base protection on at launch) lets the decision hook see read-only file tools as well. */ prepare( provider: AgentProvider, @@ -182,13 +184,14 @@ export class ProviderRuntimeLaunchAdapters { coreHooksEnabled = true, decisions = false, decisionBudgetMs?: number, - claudeHttpHookBase?: string + claudeHttpHookBase?: string, + decisionReads = false ): PreparedProviderRuntimeLaunch { const pluginRegistrations = this.options.pluginHooks?.list(provider) ?? []; const gate = decisions && this.decisionsSupported(provider); const pluginCommands = [ ...this.pluginHookCommands(provider, pluginRegistrations), - ...(gate && provider !== "opencode" ? decisionHookCommands(provider as DecisionHookProvider, this.options.permissionGate!, this.platform, decisionBudgetMs) : []) + ...(gate && provider !== "opencode" ? decisionHookCommands(provider as DecisionHookProvider, this.options.permissionGate!, this.platform, decisionBudgetMs, decisionReads) : []) ]; const openCodeDecisions = gate && provider === "opencode"; // Only providers with a hook adapter get lifecycle configuration. Cursor, @@ -232,7 +235,11 @@ export class ProviderRuntimeLaunchAdapters { return prepared([], { ...environment, ...pluginEnvironment, - ...(openCodeDecisions ? { [OPENCODE_DECISIONS_ENV]: "1", ...budgetEnvironment(decisionBudgetMs) } : {}), + ...(openCodeDecisions ? { + [OPENCODE_DECISIONS_ENV]: "1", + ...(decisionReads ? { [OPENCODE_DECISION_READS_ENV]: "1" } : {}), + ...budgetEnvironment(decisionBudgetMs) + } : {}), [OPENCODE_CONFIG_CONTENT]: openCodeLifecycleConfig( this.environment[OPENCODE_CONFIG_CONTENT], this.options.openCodePluginPath @@ -406,6 +413,8 @@ const CLAUDE_HOOKS: readonly HookMapping[] = [ { event: "UserPromptSubmit", state: "working" }, { event: "PermissionRequest", state: "needs_approval" }, { event: "PostToolUse", state: "working" }, + // An approved tool that fails runs PostToolUseFailure, not PostToolUse: the turn goes on after the approval. + { event: "PostToolUseFailure", state: "working" }, { event: "Stop", state: "idle" }, { event: "StopFailure", state: "idle" }, { event: "SessionEnd", state: "idle" }, @@ -547,17 +556,28 @@ const DECISION_TOOL_MATCHERS: Readonly> = { qwen: "^(run_shell_command|write_file|edit|replace)$" }; +/** + * With base protection on at launch the hook also matches read-only file tools, for its private-data rule only. + * Codex has no read-only file tool in its hooks: it reads through its shell. + */ +const DECISION_READ_TOOL_MATCHERS: Readonly> = { + claude: "Bash|Write|Edit|MultiEdit|NotebookEdit|Read|NotebookRead|Grep|Glob|LS", + codex: "Bash|apply_patch|Edit|Write", + qwen: "^(run_shell_command|write_file|edit|replace|read_file|read_many_files|grep_search|search_file_content|glob|list_directory)$" +}; + function decisionHookCommands( provider: DecisionHookProvider, gate: RuntimeHookHelperLaunch, platform: NodeJS.Platform, - decisionBudgetMs?: number + decisionBudgetMs?: number, + reads = false ): ProviderHookCommand[] { validateHelper(gate); const { hookSeconds } = permissionGateTimings(decisionBudgetMs); return [{ event: "PreToolUse", - matcher: DECISION_TOOL_MATCHERS[provider], + matcher: (reads ? DECISION_READ_TOOL_MATCHERS : DECISION_TOOL_MATCHERS)[provider], // The hook is installed only when something decides for this session, so it always fails closed. command: commandWithEnvironment( [gate.command, ...gate.args, "pretool"], diff --git a/src/main/services/agent-runtime/RuntimeGateway.ts b/src/main/services/agent-runtime/RuntimeGateway.ts index 7ad4f9d48..42a27aad8 100644 --- a/src/main/services/agent-runtime/RuntimeGateway.ts +++ b/src/main/services/agent-runtime/RuntimeGateway.ts @@ -79,6 +79,8 @@ interface RuntimeLease { provider: Exclude; tokenDigest: Buffer; activeTurnId: string | null; + /** A turn started and has not reported idle yet; between turns an unseen turn id may open the next one. */ + turnOpen: boolean; /** Turn ids this session already reported (bounded, oldest dropped): a late start of one of them is stale. */ seenTurnIds: Set; latest: RuntimeLifecycleSignal | null; @@ -276,6 +278,7 @@ export class RuntimeGateway { provider, tokenDigest: tokenDigest(capabilityToken), activeTurnId: null, + turnOpen: false, seenTurnIds: new Set(), latest: null, captureResult: captureResultOrGrantExpiresAt === true, @@ -469,21 +472,27 @@ export class RuntimeGateway { throw new Error("Answer capture is not authorized for this session."); } - if (message.turnId && isTurnStart(message.event)) { + const startsTurn = isTurnStart(message.event); + if (message.turnId && startsTurn) { // Each hook runs on its own connection, so a start of an earlier turn can arrive after the next turn began; // it must not make that newer turn's events look stale. if (message.turnId !== lease.activeTurnId && lease.seenTurnIds.has(message.turnId)) return null; lease.activeTurnId = message.turnId; + lease.turnOpen = true; + rememberTurn(lease, message.turnId); + } else if (message.turnId && message.turnId !== lease.activeTurnId) { + // A turn whose start hook was lost (it timed out, or the CLI skipped it) still counts: after an idle, an + // unseen id opens the next turn. During an open turn, or for an id already seen, the event is out of order. + if (lease.activeTurnId && (lease.turnOpen || lease.seenTurnIds.has(message.turnId))) return null; + lease.activeTurnId = message.turnId; + lease.turnOpen = true; rememberTurn(lease, message.turnId); - } else if ( - message.turnId - && lease.activeTurnId - && message.turnId !== lease.activeTurnId - ) { - return null; } else if (message.turnId) { rememberTurn(lease, message.turnId); + } else if (startsTurn) { + lease.turnOpen = true; } + if (message.state === "idle") lease.turnOpen = false; const signal: RuntimeLifecycleSignal = { state: message.state, event: message.event, @@ -877,7 +886,8 @@ function isTurnStart(event: string): boolean { return event === "UserPromptSubmit" || event === "TurnStarted" || event === "pre_llm_call" - || event === "session.status:busy"; + || event === "session.status:busy" + || event === "session.status:retry"; } function isRecord(value: unknown): value is Record { diff --git a/src/main/services/agent-runtime/applyRuntimeSignal.ts b/src/main/services/agent-runtime/applyRuntimeSignal.ts new file mode 100644 index 000000000..47b800df6 --- /dev/null +++ b/src/main/services/agent-runtime/applyRuntimeSignal.ts @@ -0,0 +1,22 @@ +import type { RuntimeLifecycleSignal } from "./RuntimeGateway.ts"; +import type { TerminalManager } from "../TerminalManager.ts"; + +/** + * What main does with one runtime gateway signal: the card's status first, then the final answer it carries (Codex + * Stop hook, OpenCode plugin) for get_agent_result and wait_for_agent, kept only when the signal belongs to the + * current turn and the last submitted input. + */ +export function applyRuntimeSignal( + terminals: Pick, + terminalSessionId: string, + signal: RuntimeLifecycleSignal +): void { + const current = terminals.applyProviderSignal(terminalSessionId, { + kind: "lifecycle", + state: signal.state, + event: signal.event, + ...(signal.turnId ? { requestId: signal.turnId } : {}), + ...(signal.threadId ? { threadId: signal.threadId } : {}) + }); + if (current && signal.result) terminals.recordAnswer(terminalSessionId, signal.result, { turnId: signal.turnId }); +} diff --git a/src/main/services/isolation/AgentIsolation.ts b/src/main/services/isolation/AgentIsolation.ts index 5d648c290..58ee5b333 100644 --- a/src/main/services/isolation/AgentIsolation.ts +++ b/src/main/services/isolation/AgentIsolation.ts @@ -1,13 +1,14 @@ import { spawnSync } from "node:child_process"; import { chmodSync, existsSync, mkdirSync, mkdtempSync, readdirSync, realpathSync, rmdirSync, rmSync, statSync, writeFileSync } from "node:fs"; import { tmpdir } from "node:os"; -import { delimiter, dirname, join } from "node:path"; +import { delimiter, dirname, isAbsolute, join } from "node:path"; import type { ProviderId, SessionIsolation } from "../../../shared/contracts.ts"; import { autoKind, PROFILE_RANK, type LaunchProfile } from "../../../shared/autoMode.ts"; import { LaunchRefusal } from "../launchRefusal.ts"; -import { isolationPaths } from "./isolationPaths.ts"; +import { isolationPaths, type PlacedProject } from "./isolationPaths.ts"; +import { validateAccountHome } from "./accountHomeIsolation.ts"; import { seatbeltProfile } from "./seatbelt.ts"; -import { bubblewrapArguments, projectHooks } from "./bubblewrap.ts"; +import { bubblewrapArguments, projectHooks, type PathKind } from "./bubblewrap.ts"; import { LinuxHostPaths } from "./linuxHostPaths.ts"; export const SANDBOX_EXEC = "/usr/bin/sandbox-exec"; @@ -75,6 +76,12 @@ export interface IsolationLaunch { env: Record; /** Folders under CanvasTTY's private data this launch was handed (its control grant, its account home). */ grantedPrivate?: readonly string[]; + /** The CLI home a launch contributor selected inside its plugin data (LaunchPipeline); checked again here. */ + accountHome?: string; + /** CanvasTTY's own gateway sockets for this launch (gatewaySocketAddresses), never read from `env`. */ + gatewaySockets?: readonly string[]; + /** Folders CanvasTTY itself placed this launch in (an environment's worktree, the person's chosen folder). */ + placedProjects?: readonly PlacedProject[]; /** The launch profile: in plan the project is not writable. */ profile?: LaunchProfile; } @@ -184,6 +191,10 @@ export class AgentIsolation { const temp = join(folder, "tmp"); mkdirSync(temp, { mode: 0o700 }); const cwd = realpathSync(launch.cwd); + const accountHome = launch.accountHome === undefined + ? undefined + : validateAccountHome(this.options.userDataPath, launch.provider, launch.accountHome, launch.env); + const granted = [...(launch.grantedPrivate ?? []), ...(accountHome ? [accountHome] : [])]; const paths = isolationPaths({ provider: launch.provider, cwd, @@ -192,14 +203,19 @@ export class AgentIsolation { hostEnvironment: this.hostEnvironment, userDataPath: this.options.userDataPath, sessionId: launch.sessionId, - ...(launch.grantedPrivate ? { grantedPrivate: launch.grantedPrivate } : {}), + ...(granted.length > 0 ? { grantedPrivate: granted } : {}), + ...(launch.gatewaySockets ? { gatewaySockets: launch.gatewaySockets } : {}), + ...(launch.placedProjects ? { placedProjects: launch.placedProjects } : {}), ...(launch.profile === "plan" ? { readOnlyProject: true } : {}) }); // `git init` and `git clone` copy git's template, sample hooks included: an empty one writes no hooks. const gitTemplate = join(folder, "git-template"); mkdirSync(gitTemplate, { mode: 0o500 }); // An agent that meets "Operation not permitted" can read why here, instead of trying other ways around it. - const env = { ...launch.env, TMPDIR: `${temp}/`, TMP: temp, TEMP: temp, GIT_TEMPLATE_DIR: gitTemplate, [ISOLATION_ENV]: ISOLATION_NOTE }; + const env: Record = { ...launch.env, TMPDIR: `${temp}/`, TMP: temp, TEMP: temp, GIT_TEMPLATE_DIR: gitTemplate, [ISOLATION_ENV]: ISOLATION_NOTE }; + // Claude Code ignores TMPDIR for its Bash tool and editor files: `${CLAUDE_CODE_TMPDIR:-/tmp}/claude-`, which + // all of the person's Claude sessions share and the layer keeps read-only. Its own folder is the launch's temp. + if (launch.provider === "claude") env.CLAUDE_CODE_TMPDIR = temp; if (available.layer === "seatbelt") { const profilePath = join(folder, "profile.sb"); writeFileSync(profilePath, seatbeltProfile(paths), { mode: 0o600, flag: "wx" }); @@ -208,8 +224,8 @@ export class AgentIsolation { // bubblewrap mounts only what exists: the CLI's own missing folders are created and a missing protected file // gets a placeholder first (LinuxHostPaths), both undone by cleanup(). releaseHostPaths = this.linuxHostPaths.prepare(paths); - const kind = (path: string): "file" | "directory" | null => { - try { const stat = statSync(path); return stat.isDirectory() ? "directory" : stat.isFile() ? "file" : null; } catch { return null; } + const kind = (path: string): PathKind => { + try { const stat = statSync(path); return stat.isDirectory() ? "directory" : stat.isFile() ? "file" : stat.isSocket() ? "socket" : null; } catch { return null; } }; const args = bubblewrapArguments(paths, { command: launch.command, args: launch.args, cwd, ...(launch.env.XDG_RUNTIME_DIR ? { runtimeDir: launch.env.XDG_RUNTIME_DIR } : {}) }, kind); const hooks = projectHooks(cwd); @@ -285,7 +301,20 @@ function findOnPath(name: string, exists: (path: string) => boolean): string | n return null; } -/** The folder of a session's control grant, from the launch environment (it is readable inside the layer). */ +/** + * The Unix socket addresses in environments CanvasTTY's own gateways prepared for a launch (browser, lifecycle, + * orchestration). Pass only those: the launch's full environment also holds the person's variables and plugins'. + */ +export function gatewaySocketAddresses(...environments: ReadonlyArray> | undefined>): string[] { + return [...new Set(environments.flatMap((environment) => Object.entries(environment ?? {}) + .filter(([name, value]) => /^CANVASTTY_.*_ADDRESS$/u.test(name) && isAbsolute(value)) + .map(([, value]) => value)))]; +} + +/** + * The folder of a session's control grant, from the control environment CanvasTTY itself prepared for the launch + * (`controlEnvironment`), never from the launch's full environment: a plugin can set the variable there too. + */ export function controlGrantFolder(env: Readonly>): string | null { const connection = env.CANVASTTY_CONTROL_CONNECTION; return connection ? dirname(connection) : null; diff --git a/src/main/services/isolation/accountHomeIsolation.ts b/src/main/services/isolation/accountHomeIsolation.ts new file mode 100644 index 000000000..9f3eee6b3 --- /dev/null +++ b/src/main/services/isolation/accountHomeIsolation.ts @@ -0,0 +1,78 @@ +import { lstatSync, realpathSync } from "node:fs"; +import { isAbsolute, join, relative, resolve, sep } from "node:path"; +import type { ProviderId } from "../../../shared/contracts.ts"; +import { PLUGIN_DATA_FOLDER, providerHomeVariables } from "./isolationPaths.ts"; + +/** The path segments of `path` below `root`, or null when it is not strictly inside. */ +function segmentsBelow(root: string, path: string): string[] | null { + const rest = relative(root, path); + if (rest === "" || rest.startsWith("..") || isAbsolute(rest)) return null; + return rest.split(sep); +} + +/** A folder path without trailing separators (`/a/b/` names the same folder as `/a/b`); the root stays as it is. */ +function withoutTrailingSeparator(path: string): string { + const trimmed = path.replace(/[\\/]+$/u, ""); + return trimmed === "" || /^[A-Za-z]:$/u.test(trimmed) ? path : trimmed; +} + +/** + * The CLI home a launch contributor handed this launch inside its own data folder: the value of the provider's own + * home variable (providerHomeVariables: CODEX_HOME, CLAUDE_CONFIG_DIR, GROK_HOME, HERMES_HOME, KIMI_HOME, + * OPENCODE_CONFIG_DIR, QWEN_HOME; never HOME or XDG_*) when it is an exact absolute path strictly below the + * contributor's data folder (a trailing separator is dropped: `homes/a/` is `homes/a`). `dataDir` comes from the + * host's contributor record, never from the plugin's answer, so a plugin can only point at a folder of its own data, + * and only the one the CLI is actually started with. + */ +export function contributedAccountHome( + provider: ProviderId, + dataDir: string | undefined, + key: string, + value: string +): string | undefined { + if (!dataDir || !isAbsolute(dataDir) || !providerHomeVariables(provider).includes(key)) return undefined; + const folder = withoutTrailingSeparator(value); + if (!isAbsolute(folder) || resolve(folder) !== folder) return undefined; + return segmentsBelow(resolve(dataDir), folder) ? folder : undefined; +} + +/** + * Checks a contributed account home again when the isolation layer is built. Nothing is reopened when the launch no + * longer starts the CLI with it (an environment replaced the variable). Otherwise it must be a real folder strictly + * inside one plugin's data folder, reached without a symbolic link at any step from `/plugin-data`: the + * path to reopen is returned, and anything else throws (the launch is refused). + */ +export function validateAccountHome( + userDataPath: string, + provider: ProviderId, + candidate: string, + env: Readonly> +): string | undefined { + const started = (value: string | undefined): boolean => typeof value === "string" && withoutTrailingSeparator(value) === candidate; + if (!providerHomeVariables(provider).some((name) => started(env[name]))) return undefined; + if (!isAbsolute(candidate) || resolve(candidate) !== candidate) { + throw new Error("The selected account home must be an exact absolute folder."); + } + const pluginData = join(resolve(userDataPath), PLUGIN_DATA_FOLDER); + const segments = segmentsBelow(pluginData, candidate); + // /: never all of plugin data, never a whole plugin's data folder. + if (!segments || segments.length < 2) { + throw new Error("The selected account home must be a folder inside one plugin's data."); + } + try { + const canonicalRoot = join(realpathSync.native(resolve(userDataPath)), PLUGIN_DATA_FOLDER); + let path = pluginData; + let canonical = canonicalRoot; + for (const segment of ["", ...segments]) { + if (segment) { + path = join(path, segment); + canonical = join(canonical, segment); + } + const info = lstatSync(path); + if (!info.isDirectory() || info.isSymbolicLink() || realpathSync.native(path) !== canonical) throw new Error("not a real folder"); + } + } catch { + throw new Error("The selected account home is missing or crosses a symbolic link."); + } + return candidate; +} diff --git a/src/main/services/isolation/bubblewrap.ts b/src/main/services/isolation/bubblewrap.ts index bd0c47b25..c09506961 100644 --- a/src/main/services/isolation/bubblewrap.ts +++ b/src/main/services/isolation/bubblewrap.ts @@ -1,6 +1,9 @@ import { dirname, join } from "node:path"; import type { IsolationPaths } from "./isolationPaths.ts"; +/** What is at a path, as far as mounting goes; null when nothing is there. */ +export type PathKind = "file" | "directory" | "socket" | null; + /** The project's own git hooks folder. */ export function projectHooks(cwd: string): string { return join(cwd, ".git", "hooks"); @@ -13,8 +16,9 @@ function within(path: string, folder: string): boolean { /** * The bubblewrap arguments of one isolated agent on Linux: the whole file system read-only, the project, this * launch's temporary folder and the CLI's own folders writable, an empty tmpfs over every folder it must not read (and - * /dev/null over such files), its own PID namespace (it cannot signal the person's processes) and the person's user - * runtime folder hidden (the systemd user manager, D-Bus session bus, SSH agent and Podman sockets live there). + * /dev/null over such files and sockets), CanvasTTY's gateway socket folders read-only, its own PID namespace (it + * cannot signal the person's processes) and the person's user runtime folder hidden (the systemd user manager, D-Bus + * session bus, SSH agent and Podman sockets live there). * * Linux cannot filter which Unix sockets a process connects to, so a socket outside those hidden folders (a system * Docker socket the person's account may use) stays reachable; docs/installing-and-security.md says so. `exists` @@ -23,7 +27,7 @@ function within(path: string, folder: string): boolean { export function bubblewrapArguments( paths: IsolationPaths, launch: { command: string; args: readonly string[]; cwd: string; runtimeDir?: string }, - exists: (path: string) => "file" | "directory" | null + exists: (path: string) => PathKind ): string[] { const args = [ "--die-with-parent", @@ -39,19 +43,34 @@ export function bubblewrapArguments( seen.add(path); args.push("--bind", path, path); } + // CanvasTTY's data folder inside a writable project: its parent folders there become mount points (a mount point + // cannot be renamed: EBUSY), the folder itself read-only over it, then its writable folders bound again. + for (const path of paths.fixedFolders) { + if (exists(path) === "directory") args.push("--bind", path, path); + } + for (const path of paths.readOnlyFolders) { + if (exists(path) !== "directory") continue; + args.push("--ro-bind", path, path); + for (const folder of seen) if (folder !== path && within(folder, path)) args.push("--bind", folder, folder); + } for (const path of paths.writableFiles) { if (exists(path) === "file") args.push("--bind", path, path); } for (const path of paths.unreadable) { const kind = exists(path); if (kind === "directory") args.push("--tmpfs", path); - else if (kind === "file") args.push("--ro-bind", "/dev/null", path); + // A socket is hidden like a file: /dev/null over it, so connecting fails. + else if (kind === "file" || kind === "socket") args.push("--ro-bind", "/dev/null", path); } // What this launch was handed stays read-only (as in the macOS profile), except the CLI's own moved home, which is // one of its writable folders. for (const path of paths.readableAgain) { if (exists(path)) args.push(seen.has(path) ? "--bind" : "--ro-bind", path, path); } + // The project's repository folder becomes a mount point before anything is mounted inside it: a mount point cannot + // be renamed (EBUSY), so `.git` cannot be moved away from the read-only mounts below and back. + const repository = join(launch.cwd, ".git"); + if ([...seen].some((folder) => within(launch.cwd, folder)) && exists(repository) === "directory") args.push("--bind", repository, repository); // Last, over every bind above (a rebound home included): the read-only files and the project's git hooks. for (const path of [...paths.gitHooks, ...(paths.protectedDirectories ?? []), ...paths.protectedWrites]) { const kind = exists(path); @@ -83,16 +102,29 @@ export function bubblewrapArguments( else if (!exists(info)) args.push("--tmpfs", info); } if (launch.runtimeDir && exists(launch.runtimeDir) === "directory") args.push("--tmpfs", launch.runtimeDir); - // CanvasTTY's gateway sockets may sit in the hidden runtime folder or in /tmp; bind their folders back. + // CanvasTTY's gateway sockets may sit in the hidden runtime folder or in /tmp; bind their folders back read-only. + // Connecting to a Unix socket needs no write access to its folder, and a writable one would let an agent remove a + // gateway socket and bind its own in its place, receiving other agents' requests. This launch's own temporary + // folder is already bound writable above. + // A bind takes the host folder as it is, so one over a hidden path would show it again: refused, never mounted. + // Over a writable folder (the project, a CLI home) it would make that read-only and undo the throwaway hooks + // folder above: refused too. Core gateways use dedicated folders, so neither happens with them. + const hidden = [...paths.unreadable, ...(launch.runtimeDir ? [launch.runtimeDir] : [])]; + const writable = [...seen, ...paths.writableFiles]; for (const path of paths.socketFolders) { - if (exists(path) === "directory" && !seen.has(path)) args.push("--bind", path, path); + if (exists(path) !== "directory" || seen.has(path)) continue; + const covered = hidden.find((secret) => within(secret, path)); + if (covered) throw new Error(`The gateway socket folder ${path} contains ${covered}, which the agent may not read.`); + const shadowed = writable.find((folder) => within(folder, path)); + if (shadowed) throw new Error(`The gateway socket folder ${path} contains ${shadowed}, which the agent writes; it is not mounted over it.`); + args.push("--ro-bind", path, path); } args.push("--chdir", launch.cwd, "--", launch.command, ...launch.args); return args; } /** A missing `path` whose nearest existing ancestor is one of the writable folders (or inside one). */ -export function creatableInside(path: string, writable: readonly string[], exists: (path: string) => "file" | "directory" | null): boolean { +export function creatableInside(path: string, writable: readonly string[], exists: (path: string) => PathKind): boolean { let current = dirname(path); for (let i = 0; i < 128 && !exists(current); i++) { const parent = dirname(current); diff --git a/src/main/services/isolation/isolationPaths.ts b/src/main/services/isolation/isolationPaths.ts index 5cdcb9a5e..b94390146 100644 --- a/src/main/services/isolation/isolationPaths.ts +++ b/src/main/services/isolation/isolationPaths.ts @@ -28,20 +28,51 @@ export interface IsolationPathInput { grantedPrivate?: readonly string[]; /** Extra socket folders this launch may connect to (the control grant's endpoint folder). */ socketFolders?: readonly string[]; + /** + * The sockets of CanvasTTY's own gateways for this launch, as the host set them up. Their folders are bound back; + * a CANVASTTY_*_ADDRESS in the launch environment (the person's own or a plugin's) never is. + */ + gatewaySockets?: readonly string[]; + /** + * Folders the host itself placed this launch in (an environment's worktree, the folder the person chose for the + * card the launch descends from). A project inside plugin data is reopened only when it is inside one of them. + */ + placedProjects?: readonly PlacedProject[]; /** Plan: the project is readable only (the CLI's own folders stay writable). */ readOnlyProject?: boolean; } +/** + * A folder the host placed a launch in. `pluginId`: an environment of that plugin placed it, so it counts only inside + * that plugin's own data folder; without it, the person chose the folder. + */ +export interface PlacedProject { + folder: string; + pluginId?: string; +} + export interface IsolationPaths { /** Writable folders (subpaths), every spelling. */ writable: string[]; + /** + * Read-only as a whole although a writable folder holds it: CanvasTTY's own data folder when the project contains it + * (a project at HOME). Its settings and the plugins it runs are not the agent's to change; the writable folders above + * that sit inside it (an account home, this launch's own folders) stay writable. + */ + readOnlyFolders: string[]; + /** + * The folders between a writable folder and a read-only folder inside it (`~/.config` for `~/.config/canvastty` in a + * project at HOME): what is inside stays writable, but they cannot be renamed or replaced, which would move the + * read-only folder out of its protected path and let a different one take its place. + */ + fixedFolders: string[]; /** Writable single files (with their `.lock` / `.tmp` / backup siblings). */ writableFiles: string[]; /** Folders that may be created (empty) on the way to a writable one: `~/.local/state` for `~/.local/state/opencode`. */ creatableFolders: string[]; - /** Inside writable folders, still not writable: the repository's git config, the CLIs' own permission settings. */ + /** Inside writable folders, still not writable: the repository's git config and `commondir`, the CLIs' own permission settings. */ protectedWrites: string[]; - /** Permission-bearing agent directories; empty directories are neutral, unlike a fabricated build.md. */ + /** Permission-bearing agent directories and the repository's linked-worktree records; empty directories are neutral. */ protectedDirectories?: string[]; /** The project's git hook folders: only `*.sample` files may be written there (what `git init` creates). */ gitHooks: string[]; @@ -113,6 +144,11 @@ const HOME_VARIABLES: Partial> = { /** A CLI's configuration file named by a variable (it may hold keys): unreadable to other CLIs, never writable to its own. */ const CONFIG_FILE_VARIABLES: Partial> = { opencode: ["OPENCODE_CONFIG"] }; +/** The environment variables that name this provider's CLI home (empty when it has none). */ +export function providerHomeVariables(provider: ProviderId): readonly string[] { + return HOME_VARIABLES[provider] ?? []; +} + function movedProviderHomes(provider: ProviderId, env: IsolationPathInput["env"], variables = HOME_VARIABLES): string[] { return (variables[provider] ?? []) .map((name) => env[name]) @@ -138,13 +174,16 @@ function sensitiveHomeFolders(home: string, xdgConfig: string): string[] { ].map((name) => join(home, name)).concat([join(xdgConfig, "gh"), join(xdgConfig, "gcloud")]); } +/** Plugins' private data below userData (PluginManager's `/plugin-data/`). */ +export const PLUGIN_DATA_FOLDER = "plugin-data"; + /** - * CanvasTTY's own private data under userData (the same list base protection refuses): tokens, connection records, - * secret stores, account homes and prepared launch files. Its gateways' runtime folders stay readable: they hold - * token-authenticated sockets and per-run hook settings the CLI must read. + * CanvasTTY's own private data under userData: tokens, connection records, secret stores, account homes, prepared + * launch files and every plugin's private data (an accounts plugin keeps all of its CLI homes there). Its gateways' + * runtime folders stay readable: they hold token-authenticated sockets and per-run hook settings the CLI must read. */ export function privateAppData(userDataPath: string): string[] { - return ["agent-control", "provider-secrets.bin", "plugin-secrets", "account-homes", "github-oauth.json", "launch-runs"] + return ["agent-control", "provider-secrets.bin", "plugin-secrets", "account-homes", "github-oauth.json", "launch-runs", PLUGIN_DATA_FOLDER] .map((name) => join(userDataPath, name)); } @@ -212,8 +251,18 @@ export function isolationPaths(input: IsolationPathInput): IsolationPaths { // variable pointing at HOME or above the project would hide them, so it is not followed. .filter((folder) => !ownFolders.includes(folder) && !hides(folder)); const privateData = privateAppData(input.userDataPath); - const grants = [...(input.grantedPrivate ?? []), join(input.userDataPath, "launch-runs", safeSegment(input.sessionId))]; const project = input.cwd; + // A project inside plugin data (a plugin environment's worktree) stays this launch's project: that exact folder is + // reopened, never its siblings, its plugin's other data or all plugin data. Only when the host placed the launch + // there: a folder an agent asked for (a subagent's cwd) could otherwise be a plugin's account homes. + const pluginDataRoot = join(input.userDataPath, PLUGIN_DATA_FOLDER); + const projectDepth = pluginDataDepth(pathSpellings, pluginDataRoot, project); + if (projectDepth >= 0 && projectDepth < 2) throw new Error(`The project ${project} is CanvasTTY's private plugin data.`); + if (projectDepth >= 2 && !(input.placedProjects ?? []).some((placed) => placesProject(pathSpellings, pluginDataRoot, placed, project))) { + throw new Error(`The project ${project} is inside CanvasTTY's private plugin data, and CanvasTTY did not place this launch there. Only a plugin environment's own folder or a folder the person chose is opened inside plugin data.`); + } + const grants = [...(input.grantedPrivate ?? []), join(input.userDataPath, "launch-runs", safeSegment(input.sessionId)), + ...(projectDepth >= 2 ? [project] : [])]; const trustedOwn = providerFolders(input.provider, {}, hostHome); const trustedOthers = AGENT_PROVIDERS.filter((provider) => provider !== input.provider).flatMap((provider) => { const defaults = providerFolders(provider, {}, hostHome); @@ -243,35 +292,47 @@ export function isolationPaths(input: IsolationPathInput): IsolationPaths { if (!withinGrant && !ownAccount) throw new Error(`CLI home ${folder} overlaps protected CanvasTTY data.`); } } + const gitDir = join(project, ".git"); + const repository = existsSync(join(gitDir, "config")); const configSources = input.provider === "opencode" ? openCodeConfigPaths(input.env, input.cwd) : { jsonFiles: [], agentDirectories: [] }; const socketFolders = [ - ...Object.entries(input.env) - .filter(([name, value]) => /^CANVASTTY_.*_ADDRESS$/u.test(name) && typeof value === "string" && isAbsolute(value)) - .map(([, value]) => dirname(value!)), + ...(input.gatewaySockets ?? []).filter((path) => isAbsolute(path)).map((path) => dirname(path)), ...(input.socketFolders ?? []), input.sessionTemp, join(input.userDataPath, "browser", "runtime"), join(input.userDataPath, "lifecycle", "runtime"), join(input.userDataPath, "orchestration", "runtime") ]; + const writable = all([ + ...(input.readOnlyProject ? [] : [project]), + input.sessionTemp, + ...ownFolders, + join(home, ".npm"), join(home, ".bun"), join(xdg.cache, "npm"), join(xdg.cache, "bun") + ]); + const userData = pathSpellings(input.userDataPath); + const readOnlyFolders = writable.some((folder) => userData.some((path) => isWithin(path, folder))) ? userData : []; + const fixedFolders = readOnlyFolders.flatMap((path) => writable.filter((folder) => folder !== path && isWithin(path, folder)) + .flatMap((folder) => ancestorsBelow(folder, path))); return { - writable: all([ - ...(input.readOnlyProject ? [] : [project]), - input.sessionTemp, - ...ownFolders, - join(home, ".npm"), join(home, ".bun"), join(xdg.cache, "npm"), join(xdg.cache, "bun") - ]), + writable, + readOnlyFolders, + fixedFolders: [...new Set(fixedFolders)], writableFiles: all(own.files), creatableFolders: all([...ownFolders, join(home, ".npm"), join(home, ".bun")].flatMap((folder) => ancestorsBelow(home, folder))), protectedWrites: all([ // A repository that exists keeps its config (hooksPath, fsmonitor, filters run code when the person uses git - // later, outside the layer); a new one may be created, which writes its config. - ...(existsSync(join(project, ".git", "config")) ? [join(project, ".git", "config"), join(project, ".git", "config.lock")] : []), + // later, outside the layer), and its per-worktree config; a new one may be created, which writes its config. + ...(repository ? [join(gitDir, "config"), join(gitDir, "config.lock"), join(gitDir, "config.worktree"), join(gitDir, "config.worktree.lock")] : []), + // Where git reads the shared config and hooks from. Protected where it exists: an empty placeholder (Linux) would + // stop git from working; seatbelt also denies creating it (projectRoots). + ...(existsSync(join(gitDir, "commondir")) ? [join(gitDir, "commondir")] : []), ...(input.provider === "codex" ? ownFolders.map((folder) => join(folder, "config.toml")) : []), ...(input.provider === "claude" ? ownFolders.flatMap((folder) => [join(folder, "settings.json"), join(folder, "settings.local.json")]) : []), ...configSources.jsonFiles ]), - protectedDirectories: all(configSources.agentDirectories.flatMap((folder) => [join(folder, "agent"), join(folder, "agents")])), + // Linked worktrees' records (their git folder, `commondir` and config): what git reads when the person works there. + protectedDirectories: all([...configSources.agentDirectories.flatMap((folder) => [join(folder, "agent"), join(folder, "agents")]), + ...(repository ? [join(gitDir, "worktrees")] : [])]), gitHooks: all([join(project, ".git", "hooks")]), projectRoots: input.readOnlyProject ? [] : all([project]), unreadable: all([...sensitive, ...others, ...privateData]), @@ -283,6 +344,30 @@ export function isolationPaths(input: IsolationPathInput): IsolationPaths { }; } +/** How many folders below plugin data `path` is (0: plugin data itself), in any spelling; -1 when it is not inside. */ +function pluginDataDepth(spellingsOf: (path: string) => string[], pluginDataRoot: string, path: string): number { + const roots = spellingsOf(pluginDataRoot); + return Math.max(-1, ...spellingsOf(path).flatMap((candidate) => roots + .filter((root) => isWithin(candidate, root)) + .map((root) => relative(root, candidate).split(/[\\/]+/u).filter(Boolean).length))); +} + +/** + * `project` is inside a folder the host placed the launch in, and that folder is itself a folder inside one plugin's + * data (never all of it or a whole plugin's data); an environment's folder counts only inside its own plugin's data. + */ +function placesProject(spellingsOf: (path: string) => string[], pluginDataRoot: string, placed: PlacedProject, project: string): boolean { + if (!isAbsolute(placed.folder)) return false; + if (pluginDataDepth(spellingsOf, pluginDataRoot, placed.folder) < 2) return false; + if (placed.pluginId !== undefined) { + // One folder name (PluginManager's `/`), and the placed folder strictly inside it. + if (!/^[^/\\.][^/\\]*$/u.test(placed.pluginId)) return false; + if (pluginDataDepth(spellingsOf, join(pluginDataRoot, placed.pluginId), placed.folder) < 1) return false; + } + const folders = spellingsOf(placed.folder); + return spellingsOf(project).some((candidate) => folders.some((folder) => isWithin(candidate, folder))); +} + /** `path` is `folder` or inside it, by the host's path rules (separators, a drive on Windows). */ function isWithin(path: string, folder: string): boolean { const rest = relative(folder, path); diff --git a/src/main/services/isolation/seatbelt.ts b/src/main/services/isolation/seatbelt.ts index 1342435f0..166ef8f56 100644 --- a/src/main/services/isolation/seatbelt.ts +++ b/src/main/services/isolation/seatbelt.ts @@ -5,7 +5,8 @@ import type { IsolationPaths } from "./isolationPaths.ts"; * except what is listed, and a later rule wins over an earlier one: * * - writes: nowhere, then the project (every spelling), this launch's own temporary folder, the CLI's own state, - * config and cache folders and the package caches, then again nowhere in the project's git hooks and git config + * config and cache folders and the package caches, then again nowhere in CanvasTTY's own data folder when the project + * contains it (its writable folders inside excepted), the project's git hooks and git config * (a hook runs outside the layer the next time the person commits) and the CLI's own permission settings; * - reads: not of other CLIs' credential folders, SSH/cloud keys and CanvasTTY's own tokens and secret stores, except * what this launch was handed (its control grant, its account home, its plugin launch files); @@ -31,6 +32,16 @@ export function seatbeltProfile(paths: IsolationPaths): string { if (paths.writableFiles.length > 0) { lines.push(`(allow file-write*${paths.writableFiles.map((path) => ` (regex ${regex(`^${escapeRegex(path)}(\\.[^/]*)?$`)})`).join("")})`); } + // CanvasTTY's data folder inside a writable project: read-only again, except the writable folders inside it. + if (paths.readOnlyFolders.length > 0) { + lines.push(`(deny file-write*${paths.readOnlyFolders.map((path) => ` (subpath ${quote(path)})`).join("")})`); + const inside = paths.writable.filter((path) => paths.readOnlyFolders.some((folder) => path.startsWith(`${folder}/`))); + if (inside.length > 0) lines.push(`(allow file-write*${inside.map((path) => ` (subpath ${quote(path)})`).join("")})`); + } + // Its parent folders inside the project: entries can be created in them, but they cannot be renamed or replaced. + if (paths.fixedFolders.length > 0) { + lines.push(`(deny file-write-unlink${paths.fixedFolders.map((path) => ` (literal ${quote(path)})`).join("")})`); + } if (paths.unreadable.length > 0) { lines.push(`(deny file-read* file-write*${paths.unreadable.map((path) => ` (subpath ${quote(path)})`).join("")})`); } @@ -53,8 +64,14 @@ export function seatbeltProfile(paths: IsolationPaths): string { lines.push(`(deny file-write*${hookRules.map((pattern) => ` (regex ${regex(pattern)})`).join("")})`); lines.push(`(allow file-write*${hookRules.map((pattern) => ` (regex ${regex(`${pattern}[^/]+\\.sample$`)})`).join("")})`); } + // Also what git reads its config and paths from, existing or not: `commondir`, `config.worktree` and the linked + // worktrees' records (`worktrees/`). Objects, refs, the index, logs and HEAD stay writable: add, commit, checkout work. if (paths.projectRoots.length > 0) { - lines.push(`(deny file-write*${paths.projectRoots.map((path) => ` (regex ${regex(`^${escapeRegex(path)}/(.*/)?\\.git/info/attributes$`)})`).join("")})`); + const files = "(info/attributes|commondir|config\\.worktree(\\.lock)?|worktrees(/.*)?)"; + lines.push(`(deny file-write*${paths.projectRoots.map((path) => ` (regex ${regex(`^${escapeRegex(path)}/(.*/)?\\.git/${files}$`)})`).join("")})`); + // These rules match by path: a `.git`, `hooks` or `info` folder renamed away and back would carry changes past + // them. They cannot be renamed or removed; entries inside them can still be created where allowed. + lines.push(`(deny file-write-unlink${paths.projectRoots.map((path) => ` (regex ${regex(`^${escapeRegex(path)}/(.*/)?\\.git(/hooks|/info)?$`)})`).join("")})`); } const protectedPaths = [...paths.protectedWrites, ...(paths.protectedDirectories ?? [])]; if (protectedPaths.length > 0) { diff --git a/src/main/services/materials/HandoffService.ts b/src/main/services/materials/HandoffService.ts index 85abce3fe..dfafee2de 100644 --- a/src/main/services/materials/HandoffService.ts +++ b/src/main/services/materials/HandoffService.ts @@ -63,6 +63,8 @@ export interface HandoffImageOps { } interface HandoffTiming { + /** How long a paste waits for the agent's CLI to take input (OpenCode drops keys until its input field is up). */ + readyMs: number; attachMs: number; pasteMs: number; pollMs: number; @@ -114,7 +116,7 @@ type Checked = { ok: true; plan: HandoffPlan } | { ok: false; reason: HandoffBlo type Reserved = { ok: true; plan: HandoffPlan; screen: SessionScreen | null } | { ok: false; reason: HandoffBlock | MaterialFailure }; -const DEFAULT_TIMING: HandoffTiming = { attachMs: 8_000, pasteMs: 6_000, pollMs: 120 }; +const DEFAULT_TIMING: HandoffTiming = { readyMs: 15_000, attachMs: 8_000, pasteMs: 6_000, pollMs: 120 }; export class HandoffService { private readonly options: HandoffServiceOptions; @@ -508,7 +510,7 @@ export class HandoffService { const paste = async (data: string): Promise => { const error = this.deliveryError(plan, screen); if (error) return error; - const written = await this.options.terminals.deliverInput(id, `${PASTE_START}${data}${PASTE_END}`, 0); + const written = await this.options.terminals.deliverInput(id, `${PASTE_START}${data}${PASTE_END}`, this.timing.readyMs); return written.delivered ? null : written.reason; }; if (!screen) { @@ -553,7 +555,7 @@ export class HandoffService { ? failed(error) : { state: "pasted", note: "enter-failed", sentAt: this.now(), imagesAttached: attached, error }; this.submitting.set(plan.id, { sessionId: id, turnStartedAt: null }); - const submitted = await this.options.terminals.deliverInput(id, "\r", 0); + const submitted = await this.options.terminals.deliverInput(id, "\r", this.timing.readyMs); return submitted.delivered ? { state: "submitted", sentAt: this.now(), imagesAttached: attached } : { state: "pasted", note: "enter-failed", sentAt: this.now(), imagesAttached: attached, error: submitted.reason }; diff --git a/src/main/services/safety/SecretRedaction.ts b/src/main/services/safety/SecretRedaction.ts index 7a0df5608..7edc1dc20 100644 --- a/src/main/services/safety/SecretRedaction.ts +++ b/src/main/services/safety/SecretRedaction.ts @@ -16,15 +16,21 @@ * `Authorization:` values, URL credentials, secret-looking query values and assignments, and long * high-entropy runs. * + * All three are found in the text as it came, and the union of their source ranges is masked: overlapping held + * values become one marker, so a later value that overlaps an earlier one shows none of its suffix, and a held + * value that happens to be a label (`password`, `Authorization`, `PRIVATE KEY`) cannot hide the credential the + * label introduces. With no held values the passes run one after another as before. + * * Each match becomes ``. Idempotent: a marker is never masked again. Values live in memory only * and are never logged, sent or shown. */ -const SECRET_MARKER = ''; /** Shorter values are not keys, and removing them would only garble ordinary text. */ const MIN_SECRET_CHARS = 8; /** Long enough for a PEM key or a service-account JSON; the search costs the same for any length. */ const MAX_SECRET_CHARS = 65_536; +/** Held forms up to this length are found with the native `indexOf`; longer ones with Knuth-Morris-Pratt. */ +const MAX_NATIVE_INDEXOF_PATTERN_CHARS = 128; const MAX_VALUES_PER_OWNER = 64; const MAX_OWNERS = 512; /** What wrapping may put between two characters of a key: whitespace and line breaks, box-drawing sides. */ @@ -108,9 +114,11 @@ export class SecretRedactionRegistry { redact(text: string): string { if (typeof text !== 'string' || text.length === 0) return typeof text === 'string' ? text : ''; - let result = maskKnownValues(text, this.knownForms()); - result = result.replace(JSON_SECRET_VALUE, (_match, prefix: string, closing: string) => `${prefix}${SECRET_MARKER}${closing}`); - return redactCredentials(result); + const forms = this.knownForms(); + // With no held values there is no structural text for a known match to erase, so reuse the direct generic + // pipeline instead of allocating and merging empty-registry source-range streams on ordinary output. + if (forms.bare.length === 0 && forms.exact.length === 0) return redactJsonAndCredentials(text); + return redactKnownAndCredentials(text, forms); } /** @@ -185,82 +193,211 @@ function withoutWrapCharacters(text: string): string { return from === 0 ? text : result + text.slice(from); } -/** - * Replaces every held value in `text`: the wrap-free forms also where wrapping put up to MAX_WRAP_GAP wrap - * characters between two of their characters, the exact forms as written. Matches are taken leftmost first, the - * longest at a position, and never overlap. - */ -function maskKnownValues(text: string, forms: KnownForms): string { - const endAt = knownMatchEnds(text, forms); - if (!endAt) return text; - let result = ""; - let kept = 0; - for (let at = 0; at < text.length;) { - const end = endAt[at]; - if (end === 0) { at++; continue; } - result += text.slice(kept, at) + SECRET_MARKER; - kept = end; - at = end; +/** All held-value matches in start order, keeping only the longest valid match at each start. */ +type KnownMatch = { start: number; end: number; kind?: string; priority?: number }; +type MatchStream = { iterator: Generator; current: KnownMatch }; + +function* knownMatches(text: string, forms: KnownForms): Generator { + if (forms.bare.length === 0 && forms.exact.length === 0) return; + const bareText = forms.bare.length ? withoutWrapCharacters(text) : ''; + const hasWrapCharacters = forms.bare.length > 0 && bareText !== text; + + const streams = forms.exact.map(form => candidateOccurrences(text, form)); + if (hasWrapCharacters) streams.push(wrappedMatches(text, bareText, forms.bare)); + else for (const form of forms.bare) streams.push(candidateOccurrences(text, form)); + yield* mergeMatches(streams); +} + +/** Merge already ordered hit streams by source start, choosing the longest at each start. */ +function* mergeMatches(streams: Generator[]): Generator { + const occurrences = orderedOccurrences(streams); + let next = occurrences.next(); + while (!next.done) { + const start = next.value.start; + let end = next.value.end; + next = occurrences.next(); + while (!next.done && next.value.start === start) { + end = Math.max(end, next.value.end); + next = occurrences.next(); + } + yield { start, end }; } - return kept === 0 ? text : result + text.slice(kept); } -/** - * For each position of `text`, the end of the longest held-value match that starts there (0: none), or null when - * nothing matches. The wrap-free forms are searched in the text with its wrap characters taken out (a map leads - * back to the original positions), the exact forms in the text itself; each with Knuth-Morris-Pratt, linear in - * the text plus the form, whatever either holds. - */ -function knownMatchEnds(text: string, forms: KnownForms): Int32Array | null { - if (forms.bare.length === 0 && forms.exact.length === 0) return null; - const exact = forms.exact.filter(form => text.includes(form)); - const bare = forms.bare.length ? withoutWrapCharacters(text) : ''; - const present = forms.bare.filter(form => bare.includes(form)); - if (present.length === 0 && exact.length === 0) return null; - const endAt = new Int32Array(text.length); - for (const form of exact) { - for (const start of occurrences(text, form)) endAt[start] = Math.max(endAt[start], start + form.length); +/** Merge hit streams into start order without discarding same-start candidates. */ +function* orderedOccurrences(streams: Generator[]): Generator { + if (streams.length === 1) { + yield* streams[0]!; + return; } - if (present.length === 0) return endAt; - // Where each character of `bare` stands in `text`, and how many oversized gaps lie before it (a match may - // not cross one). - const positions = new Int32Array(bare.length); - const oversized = new Int32Array(bare.length + 1); - for (let i = 0, count = 0, previous = -1; i < text.length; i++) { - if (isWrapCharacter(text.charCodeAt(i))) continue; - oversized[count + 1] = oversized[count] + (previous >= 0 && i - previous - 1 > MAX_WRAP_GAP ? 1 : 0); - positions[count++] = i; - previous = i; + const heap: MatchStream[] = []; + const push = (stream: MatchStream): void => { + let index = heap.length; + heap.push(stream); + while (index > 0) { + const parent = (index - 1) >> 1; + if (heap[parent]!.current.start <= stream.current.start) break; + heap[index] = heap[parent]!; + index = parent; + } + heap[index] = stream; + }; + const pop = (): MatchStream => { + const first = heap[0]!; + const last = heap.pop()!; + if (heap.length > 0) { + let index = 0; + while (true) { + const left = index * 2 + 1; + if (left >= heap.length) break; + const right = left + 1; + const child = right < heap.length && heap[right]!.current.start < heap[left]!.current.start ? right : left; + if (heap[child]!.current.start >= last.current.start) break; + heap[index] = heap[child]!; + index = child; + } + heap[index] = last; + } + return first; + }; + const enqueueFirst = (iterator: Generator): void => { + const next = iterator.next(); + if (!next.done) push({ iterator, current: next.value }); + }; + const advance = (stream: MatchStream): void => { + const next = stream.iterator.next(); + if (!next.done) { + stream.current = next.value; + push(stream); + } + }; + + for (const iterator of streams) enqueueFirst(iterator); + while (heap.length > 0) { + const first = pop(); + const current = first.current; + advance(first); + yield current; } - for (const form of present) { - for (const start of occurrences(bare, form)) { - // Only the gaps inside the match count, not the one before its first character. - if (oversized[start + form.length] - oversized[start + 1] !== 0) continue; - const from = positions[start]; - endAt[from] = Math.max(endAt[from], positions[start + form.length - 1] + 1); +} + +function* candidateOccurrences(text: string, pattern: string): Generator { + if (pattern.length > MAX_NATIVE_INDEXOF_PATTERN_CHARS) { + yield* kmpOccurrences(text, pattern); + return; + } + yield* exactOccurrences(text, pattern); +} + +function* exactOccurrences(text: string, pattern: string): Generator { + for (let from = 0; from <= text.length - pattern.length;) { + const start = text.indexOf(pattern, from); + if (start < 0) return; + yield { start, end: start + pattern.length }; + // Advancing by one preserves overlapping hits for the leftmost/longest merger. + from = start + 1; + } +} + +/** Keep worst-case scan cost linear for long patterns whose repetitive near-matches fool native indexOf. */ +function* kmpOccurrences(text: string, pattern: string): Generator { + const firstCode = pattern.charCodeAt(0); + const patternLength = pattern.length; + const searchPrefix = pattern.slice(0, MAX_NATIVE_INDEXOF_PATTERN_CHARS); + const prefixLength = searchPrefix.length; + const firstStart = text.indexOf(searchPrefix); + if (firstStart < 0) return; + + const failure = kmpFailure(pattern); + for (let i = firstStart + prefixLength - 1, k = prefixLength - 1; i < text.length; i++) { + // Skip ordinary spans in native code when no KMP prefix is active. The bounded prefix stays code-unit based. + if (k === 0) { + const start = text.indexOf(searchPrefix, i); + if (start < 0) return; + i = start + prefixLength - 1; + k = prefixLength - 1; + } + const textCode = text.charCodeAt(i); + let patternCode = k === 0 ? firstCode : pattern.charCodeAt(k); + while (k > 0 && textCode !== patternCode) { + k = failure[k - 1]!; + patternCode = k === 0 ? firstCode : pattern.charCodeAt(k); + } + if (textCode === patternCode) k++; + if (k === patternLength) { + yield { start: i - k + 1, end: i + 1 }; + k = failure[k - 1]!; } } - return endAt; } -/** Every start of `pattern` in `text`, overlapping ones included (Knuth-Morris-Pratt). */ -function* occurrences(text: string, pattern: string): Generator { +/** Prefix lengths for Knuth-Morris-Pratt. Long candidate patterns use this to retain linear scan cost. */ +function kmpFailure(pattern: string): Int32Array { const failure = new Int32Array(pattern.length); for (let i = 1, k = 0; i < pattern.length; i++) { - while (k > 0 && pattern.charCodeAt(i) !== pattern.charCodeAt(k)) k = failure[k - 1]; + while (k > 0 && pattern.charCodeAt(i) !== pattern.charCodeAt(k)) k = failure[k - 1]!; if (pattern.charCodeAt(i) === pattern.charCodeAt(k)) k++; failure[i] = k; } - for (let i = 0, k = 0; i < text.length; i++) { - while (k > 0 && text.charCodeAt(i) !== pattern.charCodeAt(k)) k = failure[k - 1]; - if (text.charCodeAt(i) === pattern.charCodeAt(k)) k++; - if (k === pattern.length) { - yield i - k + 1; - k = failure[k - 1]; + return failure; +} + +/** + * Find wrap-tolerant matches from one shared compact-text merge and one source-offset ring. All bare starts are + * processed monotonically; a ring as long as the longest form plus one retains every start through the farthest + * end considered at that start, including overlapping candidates at later starts. + */ +function* wrappedMatches(text: string, bareText: string, patterns: readonly string[]): Generator { + const occurrences = orderedOccurrences(patterns.map(pattern => candidateOccurrences(bareText, pattern))); + let next = occurrences.next(); + if (next.done) return; + + const longestPattern = patterns[0]!.length; // knownForms sorts patterns longest first + const ringLength = longestPattern + 1; + const positions = new Uint32Array(ringLength); + const oversizedAfter = new Uint32Array(ringLength); + let raw = 0; + let bare = 0; + let previous = -1; + let oversizedGaps = 0; + + const translate = (start: number, end: number): KnownMatch | null => { + while (bare < end) { + const code = text.charCodeAt(raw); + if (!isWrapCharacter(code)) { + if (previous >= 0 && raw - previous - 1 > MAX_WRAP_GAP) oversizedGaps++; + const slot = bare % ringLength; + positions[slot] = raw; + oversizedAfter[slot] = oversizedGaps; + bare++; + previous = raw; + } + raw++; + } + const startSlot = start % ringLength; + const endSlot = (end - 1) % ringLength; + // Exclude the gap before the first matched character; only interior gaps can invalidate this candidate. + if (oversizedAfter[endSlot]! - oversizedAfter[startSlot]! > 0) return null; + return { start: positions[startSlot]!, end: positions[endSlot]! + 1 }; + }; + + while (!next.done) { + const compactStart = next.value.start; + let rawStart = -1; + let rawEnd = -1; + while (!next.done && next.value.start === compactStart) { + // Check every candidate before selecting the longest. A longer form can cross an oversized gap while a + // shorter form at the same start remains valid. + const translated = translate(next.value.start, next.value.end); + if (translated) { + rawStart = translated.start; + rawEnd = Math.max(rawEnd, translated.end); + } + next = occurrences.next(); } + if (rawEnd >= 0) yield { start: rawStart, end: rawEnd }; } } - function lastChars(text: string, maxChars: number): string { return text.length <= maxChars ? text : text.slice(text.length - maxChars); } @@ -346,15 +483,15 @@ function isLineStartUncrossed(text: string, b: number, last: number, forms: Know // Held-value matches that start before `b` lie within knownSpan of it. const checkFrom = Math.max(0, Math.min(last, quote)); const from = Math.max(0, checkFrom - knownSpan); - const endAt = knownMatchEnds(text.slice(from, Math.min(text.length, b + knownSpan)), forms); - if (!endAt) return true; - for (let at = 0; from + at < b; at++) { - if (endAt[at] !== 0 && from + endAt[at] > checkFrom) return false; + const window = text.slice(from, Math.min(text.length, b + knownSpan)); + for (const match of knownMatches(window, forms)) { + if (from + match.start >= b) break; + if (from + match.end > checkFrom) return false; } return true; } -type Rule = { kind: string; pattern: RegExp; replace?: (match: string, ...groups: string[]) => string }; +type Rule = { kind: string; pattern?: RegExp; transform?: (text: string) => string; replace?: (match: string, ...groups: string[]) => string }; const marker = (kind: string): string => ``; const NOT_MASKED = '(?!|[:=])\\s*)(?:"${NOT_MASKED}[^"\\r\\n]{4,}"|'${NOT_MASKED}[^'\\r\\n]{4,}'|\`${NOT_MASKED}[^\`\\r\\n]{4,}\`|${NOT_MASKED}[^\\s"'\`<>,;]{4,})`, 'gu'), replace: (_match, prefix) => `${prefix}${marker('assignment')}` }, // A random run the terminal wrapped over lines, judged as one run. - { kind: 'high-entropy', pattern: new RegExp(`(? { const joined = match.replace(/[\s\u2500-\u257f|]/gu, ''); return joined.length >= 32 && looksRandom(joined) ? marker('high-entropy') : match; } }, + { kind: 'high-entropy', transform: text => maskEntropyRanges(text, wrappedEntropyRanges(text)) }, // A long run that mixes upper case, lower case and digits with high entropy. Pure hex (a commit SHA) has no // upper case and survives; paths never form one run because `/` and `.` end it. - { kind: 'high-entropy', pattern: /(? looksRandom(match) ? marker('high-entropy') : match } + { kind: 'high-entropy', transform: text => maskEntropyRanges(text, unwrappedEntropyRanges(text)) } ]; -function looksRandom(value: string): boolean { - const upper = value.match(/[A-Z]/gu)?.length ?? 0, lower = value.match(/[a-z]/gu)?.length ?? 0, digits = value.match(/[0-9]/gu)?.length ?? 0; +/** + * Redact registered values and credential shapes against their original source ranges together. Applying the + * registry first can destroy a structural label (for example `PRIVATE KEY`, `password`, or `Authorization`) + * before its shape is recognized. Applying generic rules first has the inverse problem: a generic match can + * split a longer registered value. Keeping both sets of source ranges lets the renderer mask their union. + */ +function redactKnownAndCredentials(text: string, forms: KnownForms): string { + const streams: Generator[] = []; + const known = forms.bare.length > 0 || forms.exact.length > 0 ? [...knownMatches(text, forms)] : []; + if (known.length > 0) { + streams.push((function* (): Generator { + for (const match of known) yield { ...match, kind: 'secret', priority: 0 }; + })()); + } + streams.push(jsonSecretRanges(text)); + for (let index = 0; index < RULES.length; index++) { + const rule = RULES[index]!; + if (rule.pattern) streams.push(ruleSecretRanges(text, rule, 10 + index)); + } + // A random run is judged without the held values in it, as when they were masked first: otherwise a name joined + // to a held value (`PROVIDER_API_KEY=`) would form one random-looking run and lose its name. + const entropyText = withHeldValuesBroken(text, known); + streams.push(wrappedEntropyRanges(entropyText)); + streams.push(unwrappedEntropyRanges(entropyText)); + + let result = ''; + let copied = 0; + let active: KnownMatch | null = null; + let chosen: KnownMatch | null = null; + for (const candidate of orderedOccurrences(streams)) { + if (!candidate.kind || candidate.end <= candidate.start) continue; + if (active && candidate.start < active.end) { + active.end = Math.max(active.end, candidate.end); + if (preferMask(candidate, chosen!)) chosen = candidate; + continue; + } + if (active && chosen) { + result += text.slice(copied, active.start) + marker(chosen.kind!); + copied = active.end; + } + active = { start: candidate.start, end: candidate.end }; + chosen = candidate; + } + if (!active || !chosen) return text; + return result + text.slice(copied, active.start) + marker(chosen.kind!) + text.slice(active.end); +} + +/** `text` with every held-value match replaced by as many characters that end a random run (offsets unchanged). */ +function withHeldValuesBroken(text: string, known: readonly KnownMatch[]): string { + if (known.length === 0) return text; + let result = ''; + let copied = 0; + for (const { start, end } of known) { + if (end <= copied) continue; + const from = Math.max(start, copied); + result += text.slice(copied, from) + '\u0001'.repeat(end - from); + copied = end; + } + return result + text.slice(copied); +} + +/** Prefer the most complete source match, while preserving JSON's earlier guard against later generic rules. */ +function preferMask(candidate: KnownMatch, current: KnownMatch): boolean { + const candidateIsJson = candidate.priority === 1; + const currentIsJson = current.priority === 1; + const candidateIsGeneric = (candidate.priority ?? 0) >= 10; + const currentIsGeneric = (current.priority ?? 0) >= 10; + if (candidateIsJson && currentIsGeneric) return !privateKeyContainsJson(current, candidate); + if (currentIsJson && candidateIsGeneric) return privateKeyContainsJson(candidate, current); + const candidateLength = candidate.end - candidate.start; + const currentLength = current.end - current.start; + if (candidateLength !== currentLength) return candidateLength > currentLength; + return (candidate.priority ?? Number.MAX_SAFE_INTEGER) < (current.priority ?? Number.MAX_SAFE_INTEGER); +} + +/** PEM masking runs after the JSON-value pass and can still consume a nested JSON marker inside its body. */ +function privateKeyContainsJson(privateKey: KnownMatch, json: KnownMatch): boolean { + return privateKey.kind === 'private-key' && privateKey.start <= json.start && privateKey.end >= json.end + && (privateKey.start < json.start || privateKey.end > json.end); +} + +function* jsonSecretRanges(text: string): Generator { + const pattern = new RegExp(JSON_SECRET_VALUE.source, JSON_SECRET_VALUE.flags); + for (let match = pattern.exec(text); match; match = pattern.exec(text)) { + const prefix = match[1] ?? ''; + const closing = match[2] ?? ''; + const start = match.index + prefix.length; + const end = match.index + match[0].length - closing.length; + if (end > start) yield { start, end, kind: 'secret', priority: 1 }; + } +} + +function* ruleSecretRanges(text: string, rule: Rule, priority: number): Generator { + const originalPattern = rule.pattern!; + const flags = originalPattern.flags.includes('g') ? originalPattern.flags : `${originalPattern.flags}g`; + const pattern = new RegExp(originalPattern.source, flags); + for (let match = pattern.exec(text); match; match = pattern.exec(text)) { + let start = match.index; + let end = match.index + match[0].length; + switch (rule.kind) { + case 'authorization': + // Its replacement keeps the Authorization label and masks an optional Bearer/Basic/etc. scheme too. + start += (match[1] ?? '').length; + break; + case 'bearer': { + const prefix = /^Bearer\s+/u.exec(match[0]); + start += prefix?.[0].length ?? 0; + break; + } + case 'url-credentials': { + const scheme = match[1] ?? ''; + const userinfo = match[2] ?? ''; + // Keep the existing rule's allowance for a short, plain user name without a password. + if (!userinfo.includes(':') && userinfo.length < 16) continue; + start += scheme.length; + end = start + userinfo.length; + break; + } + case 'url-secret': + start += (match[1] ?? '').length; + break; + case 'assignment': { + // JSON_SECRET_VALUE runs first in the public pipeline. When it recognizes the same quoted key/value, + // its marker blocks this later assignment match and preserves the JSON quotes. + const jsonMatch = new RegExp(JSON_SECRET_VALUE.source, JSON_SECRET_VALUE.flags).exec(match[0]); + if (jsonMatch?.index === 0 && jsonMatch[0].length === match[0].length) continue; + start += (match[1] ?? '').length; + break; + } + } + if (end > start) yield { start, end, kind: rule.kind, priority }; + } +} + +function* wrappedEntropyRanges(text: string): Generator { + for (let at = 0; at < text.length;) { + if (!isEntropyRunCharacter(text.charCodeAt(at)) || (at > 0 && isEntropyRunCharacter(text.charCodeAt(at - 1)))) { + at++; + continue; + } + const firstEnd = entropyRunEnd(text, at); + if (firstEnd - at < 12) { + at = firstEnd; + continue; + } + let end = firstEnd; + let joinedLength = firstEnd - at; + let continuations = 0; + while (true) { + let next = end; + if (text.charCodeAt(next) === 0x0d && text.charCodeAt(next + 1) === 0x0a) next += 2; + else if (text.charCodeAt(next) === 0x0a) next++; + else break; + let decoration = 0; + while (decoration < 8 && next < text.length) { + const code = text.charCodeAt(next); + if (!(code === 0x20 || code === 0x09 || code === 0x7c || (code >= 0x2500 && code <= 0x257f))) break; + next++; + decoration++; + } + const nextEnd = entropyRunEnd(text, next); + if (nextEnd - next < 4 || !hasDigit(text, next, nextEnd)) break; + joinedLength += nextEnd - next; + continuations++; + end = nextEnd; + } + if (continuations > 0) { + if (joinedLength >= 32 && looksRandomRange(text, at, end, true)) { + yield { start: at, end, kind: 'high-entropy', priority: 10 + RULES.length }; + } + // A low-entropy wrapped candidate consumes its continuation too. + at = end; + } else at = firstEnd; + } +} + +function* unwrappedEntropyRanges(text: string): Generator { + for (let at = 0; at < text.length;) { + if (!isEntropyRunCharacter(text.charCodeAt(at)) || (at > 0 && isEntropyRunCharacter(text.charCodeAt(at - 1)))) { + at++; + continue; + } + const end = entropyRunEnd(text, at); + if (end - at >= 32 && looksRandomRange(text, at, end, false)) { + yield { start: at, end, kind: 'high-entropy', priority: 11 + RULES.length }; + } + at = end; + } +} + +function looksRandomRange(value: string, from: number, to: number, skipWrap: boolean): boolean { + let upper = 0, lower = 0, digits = 0; + for (let i = from; i < to && (upper < 2 || lower < 2 || digits < 2); i++) { + const code = value.charCodeAt(i); + if (skipWrap && (code === 0x7c || isWrapCharacter(code))) continue; + if (code >= 0x41 && code <= 0x5a) upper++; + else if (code >= 0x61 && code <= 0x7a) lower++; + else if (code >= 0x30 && code <= 0x39) digits++; + } if (upper < 2 || lower < 2 || digits < 2) return false; - const counts = new Map(); - for (const character of value) counts.set(character, (counts.get(character) ?? 0) + 1); + const counts = new Map(); + let length = 0; + for (let i = from; i < to; i++) { + const code = value.charCodeAt(i); + if (skipWrap && (code === 0x7c || isWrapCharacter(code))) continue; + counts.set(code, (counts.get(code) ?? 0) + 1); + length++; + } let entropy = 0; - for (const count of counts.values()) { const p = count / value.length; entropy -= p * Math.log2(p); } + for (const count of counts.values()) { const p = count / length; entropy -= p * Math.log2(p); } return entropy >= 4.2; } +const isEntropyRunCharacter = (code: number): boolean => code >= 0x30 && code <= 0x39 + || code >= 0x41 && code <= 0x5a || code >= 0x61 && code <= 0x7a + || code === 0x2b || code === 0x3d || code === 0x5f || code === 0x2d; + +function entropyRunEnd(text: string, from: number): number { + let end = from; + while (end < text.length && isEntropyRunCharacter(text.charCodeAt(end))) end++; + return end; +} + +function hasDigit(text: string, from: number, to: number): boolean { + for (let i = from; i < to; i++) { + const code = text.charCodeAt(i); + if (code >= 0x30 && code <= 0x39) return true; + } + return false; +} + +/** Render the same bounded entropy scans used when registered values overlap credential shapes. */ +function maskEntropyRanges(text: string, ranges: Iterable): string { + let result = ''; + let copied = 0; + for (const { start, end } of ranges) { + result += text.slice(copied, start) + marker('high-entropy'); + copied = end; + } + return copied === 0 ? text : result + text.slice(copied); +} + /** The generic shapes alone (no registered values). */ export function redactCredentials(text: string): string { if (typeof text !== 'string' || text.length === 0) return typeof text === 'string' ? text : ''; let result = text; for (const rule of RULES) { + if (rule.transform) { + result = rule.transform(result); + continue; + } + if (!rule.pattern) continue; result = result.replace(rule.pattern, (match: string, ...rest: unknown[]) => { // Arguments after the match: the capture groups, then the numeric offset and the whole input. const end = rest.findIndex(value => typeof value === 'number'); @@ -420,3 +792,9 @@ export function redactCredentials(text: string): string { } return result; } + +/** The registry's pre-range pipeline when it has no held values to mask. */ +function redactJsonAndCredentials(text: string): string { + const withJsonValues = text.replace(JSON_SECRET_VALUE, (_match, prefix: string, closing: string) => `${prefix}${marker('secret')}${closing}`); + return redactCredentials(withJsonValues); +} diff --git a/src/main/services/safety/baseProtection.ts b/src/main/services/safety/baseProtection.ts index 2675e022d..ee1116da3 100644 --- a/src/main/services/safety/baseProtection.ts +++ b/src/main/services/safety/baseProtection.ts @@ -1,5 +1,5 @@ import { tmpdir } from 'node:os'; -import { join } from 'node:path'; +import { isAbsolute, join, normalize } from 'node:path'; import { isPathInside } from '../../../agent-runtime/path-inside.mjs'; import { analyzeAction, commandFromArgv, realish, type HardFacts, type PrivateData, type ToolAction } from './commandFacts.ts'; @@ -10,7 +10,7 @@ import { analyzeAction, commandFromArgv, realish, type HardFacts, type PrivateDa * model: local rules only, no git, no network. */ -const BASE_DENY_RULES = ['app-private', 'elevation', 'pipe-to-shell', 'download-exec', 'disk', 'fork-bomb', 'delete-outside', 'write-outside', 'unknown-target'] as const; +const BASE_DENY_RULES = ['app-private', 'elevation', 'pipe-to-shell', 'download-exec', 'disk', 'fork-bomb', 'delete-outside', 'write-outside', 'unknown-target', 'uncheckable'] as const; export type BaseDenyRule = typeof BASE_DENY_RULES[number]; /** What the model reads: why the call was refused and what to do instead. */ @@ -23,27 +23,57 @@ const DENY_MESSAGES: Readonly> = { 'fork-bomb': 'CanvasTTY blocked this command: it would exhaust the computer\'s processes. Do not run it.', 'delete-outside': 'CanvasTTY blocked this command: it deletes files outside the project folder (or the folder itself). Delete only inside the project; if something elsewhere must go, ask the person.', 'write-outside': 'CanvasTTY blocked this command: it writes outside the project folder. Keep changes inside the project; if a file elsewhere must change, ask the person.', - 'unknown-target': 'CanvasTTY blocked this command: it deletes or moves a path held in a shell variable or command output that CanvasTTY cannot resolve, so it cannot tell whether the path is inside the project folder. Write the path out, or set the variable in the same command to a path inside the project, and run it again.' + 'unknown-target': 'CanvasTTY blocked this command: it deletes or moves a path held in a shell variable or command output that CanvasTTY cannot resolve, so it cannot tell whether the path is inside the project folder. Write the path out, or set the variable in the same command to a path inside the project, and run it again.', + uncheckable: 'CanvasTTY blocked this tool call: it names more paths, longer paths or more nested commands than base protection checks in one call, so it cannot tell what the call reaches. Split it into smaller calls (for example at most 32 files per read), write paths out plainly, and run it again.' }; const TEMP_WRITE_MESSAGE = 'CanvasTTY blocked this command: it writes to the temporary folder (/tmp or $TMPDIR), which is outside the project folder. Make a scratch folder inside the project instead (for example ./tmp, added to .gitignore if needed) and use that; if a file elsewhere must change, ask the person.'; +const ANALYZER_FAILURE_MESSAGE = 'CanvasTTY blocked this tool call: base protection failed while checking it, so it cannot tell what the call reaches. Write the command and paths out plainly and run it again; if it keeps failing, ask the person.'; + export interface BaseVerdict { rule: BaseDenyRule; message: string } /** Tool names of the CLIs' hooks: shells and file writes. Anything else is not checked. */ const SHELL_TOOLS = new Set(['Bash', 'bash', 'run_shell_command', 'shell', 'local_shell', 'exec_command']); const EDIT_TOOLS = new Set(['Write', 'Edit', 'MultiEdit', 'NotebookEdit', 'write_file', 'edit', 'replace', 'apply_patch']); +/** + * Read-only file tools: Claude Code's Read, NotebookRead, Grep, Glob and LS; Qwen Code's read_file, read_many_files, + * grep_search (search_file_content), glob and list_directory; OpenCode's read, grep, glob and list arrive under the + * Claude names. They are checked only against CanvasTTY's private data, never against the project folder. + */ +const READ_TOOLS = new Set(['Read', 'NotebookRead', 'Grep', 'Glob', 'LS', 'read_file', 'read_many_files', 'grep_search', 'search_file_content', 'glob', 'list_directory']); +/** Read-only tools whose `path` is a folder they walk (search), rather than one file or one listing. */ +const SEARCH_TOOLS = new Set(['Grep', 'Glob', 'grep_search', 'search_file_content', 'glob']); const MAX_PATHS = 32; +/** Files one `apply_patch` may touch and still be checked. */ +const MAX_PATCH_PATHS = 256; const MAX_PATH_CHARS = 4_096; -/** The files an `apply_patch` envelope touches. */ -export function patchPaths(patch: string): string[] { +/** + * A path field of a tool's input: the path, null when it is absent, or `false` when it is there but cannot be checked + * (not a string, or longer than MAX_PATH_CHARS even after `./` and `a/..` padding is folded away). + */ +function pathField(value: unknown): string | null | false { + if (value === undefined || value === null || value === '') return null; + if (typeof value !== 'string') return false; + if (value.length <= MAX_PATH_CHARS) return value; + const folded = normalize(value); + return folded.length <= MAX_PATH_CHARS ? folded : false; +} + +/** The files an `apply_patch` envelope touches; `uncheckable` when there are more, or longer ones, than are checked. */ +function patchTargets(patch: string): { paths: string[]; uncheckable: boolean } { const paths: string[] = []; for (const match of patch.matchAll(/^\*\*\* (?:Add|Update|Delete) File: (.+)$|^\*\*\* Move to: (.+)$/gmu)) { - const path = (match[1] ?? match[2] ?? '').trim(); - if (path && path.length <= MAX_PATH_CHARS) paths.push(path); - if (paths.length >= MAX_PATHS) break; + const path = pathField((match[1] ?? match[2] ?? '').trim()); + if (path === false || paths.length >= MAX_PATCH_PATHS) return { paths, uncheckable: true }; + if (path) paths.push(path); } - return paths; + return { paths, uncheckable: false }; +} + +/** The files an `apply_patch` envelope touches. */ +export function patchPaths(patch: string): string[] { + return patchTargets(patch).paths; } const PREVIEW_PATH = /"(?:file_path|filePath|path|notebook_path|absolute_path)"\s*:\s*"((?:[^"\\]|\\.){1,4096})"/gu; @@ -55,14 +85,24 @@ const PREVIEW_PATH = /"(?:file_path|filePath|path|notebook_path|absolute_path)"\ */ export function actionFromHook(toolName: string, toolInput: unknown, preview: string | null = null): ToolAction { const raw = toolInput && typeof toolInput === 'object' && !Array.isArray(toolInput) ? toolInput as Record : null; - const text = (value: unknown): string | null => typeof value === 'string' && value.length > 0 && value.length <= MAX_PATH_CHARS ? value : null; - const commandCwd = text(raw?.cwd) ?? text(raw?.workdir) ?? text(raw?.directory); + // A path field that is there but cannot be checked makes the call uncheckable: it is refused, never skipped. + let uncheckable = false; + const field = (...values: unknown[]): string | null => { + for (const value of values) { + const path = pathField(value); + if (path === false) uncheckable = true; + else if (path) return path; + } + return null; + }; + const commandCwd = field(raw?.cwd, raw?.workdir, raw?.directory); if (SHELL_TOOLS.has(toolName)) { const candidate = raw?.command ?? raw?.cmd; const command = typeof candidate === 'string' ? candidate : Array.isArray(candidate) && candidate.length && candidate.every(item => typeof item === 'string') ? commandFromArgv(candidate as string[]) : null; - return { kind: 'shell', command, commandCwd, paths: [] }; + return { kind: 'shell', command, commandCwd, paths: [], ...(uncheckable ? { uncheckable } : {}) }; } + if (READ_TOOLS.has(toolName)) return readAction(toolName, raw, preview); if (!EDIT_TOOLS.has(toolName)) return { kind: null, command: null, commandCwd: null, paths: [] }; if (!raw && preview && toolName !== 'apply_patch') { const paths: string[] = []; @@ -73,31 +113,71 @@ export function actionFromHook(toolName: string, toolInput: unknown, preview: st return { kind: 'edit', command: null, commandCwd: null, paths }; } if (toolName === 'apply_patch') { - const patch = text(raw?.command) ?? (typeof raw?.patch === 'string' ? raw.patch : typeof raw?.input === 'string' ? raw.input : typeof raw?.patchText === 'string' ? raw.patchText : ''); - return { kind: 'edit', command: null, commandCwd, paths: patchPaths(patch ?? '') }; + const patch = (typeof raw?.command === 'string' && raw.command ? raw.command : null) ?? (typeof raw?.patch === 'string' ? raw.patch : typeof raw?.input === 'string' ? raw.input : typeof raw?.patchText === 'string' ? raw.patchText : ''); + const targets = patchTargets(patch ?? ''); + return { kind: 'edit', command: null, commandCwd, paths: targets.paths, ...(uncheckable || targets.uncheckable ? { uncheckable: true } : {}) }; + } + const path = field(raw?.file_path, raw?.filePath, raw?.path, raw?.notebook_path, raw?.absolute_path); + return { kind: 'edit', command: null, commandCwd: null, paths: path ? [path] : [], ...(uncheckable ? { uncheckable } : {}) }; +} + +/** A read-only tool's files, glob patterns and searched folders. Cut input: the paths at the start of its preview. */ +function readAction(toolName: string, raw: Record | null, preview: string | null): ToolAction { + const action: ToolAction = { kind: 'read', command: null, commandCwd: null, paths: [], searched: [] }; + if (!raw) { + for (const match of (preview ?? '').matchAll(PREVIEW_PATH)) { + try { const value: unknown = JSON.parse(`"${match[1]}"`); if (typeof value === 'string' && value) action.searched!.push(value); } catch { /* a cut escape */ } + if (action.searched!.length >= 4) break; + } + return action; } - const path = text(raw?.file_path) ?? text(raw?.filePath) ?? text(raw?.path) ?? text(raw?.notebook_path) ?? text(raw?.absolute_path); - return { kind: 'edit', command: null, commandCwd: null, paths: path ? [path] : [] }; + // A path field that is there but cannot be checked (not a string, too long) makes the call uncheckable: refused. + const field = (...values: unknown[]): string | null => { + for (const value of values) { + const path = pathField(value); + if (path === false) action.uncheckable = true; + else if (path) return path; + } + return null; + }; + const file = field(raw.file_path, raw.filePath, raw.notebook_path, raw.absolute_path); + if (file) action.paths.push(file); + const folder = field(raw.path, raw.dir_path, raw.directory); + if (folder) (SEARCH_TOOLS.has(toolName) ? action.searched! : action.paths).push(folder); + // A glob pattern names files itself (`/…/canvastty/provider-*`); relative ones resolve against the searched folder. + const pattern = toolName === 'Glob' || toolName === 'glob' ? field(raw.pattern) : null; + if (pattern) action.paths.push(folder && !isAbsolute(pattern) && !pattern.startsWith('~') && !pattern.startsWith('$') ? join(folder, pattern) : pattern); + // Qwen Code's read_many_files: `paths` and `include` are both globs it expands and reads. More than are checked is + // refused, not cut short. + const many = [raw.paths, raw.include].flatMap(list => list === undefined || list === null ? [] : Array.isArray(list) ? list : [list]); + if (many.length > MAX_PATHS) { action.uncheckable = true; return action; } + for (const item of many) { const path = field(item); if (path) action.searched!.push(path); } + return action; } /** * CanvasTTY's own private data under its userData folder (the app passes `app.getPath('userData')`; tests pass a * temporary folder): the control token and descriptor, the gateways' connection records and sockets, the secret - * stores, the per-account homes and the prepared launch runs. Its settings and layouts are not listed. + * stores, the per-account homes, the prepared launch runs, the per-card Claude Code draft folders (claude-draft, where + * the app and one card exchange that card's prompt draft) and the plugins' data (plugin-data: an accounts plugin keeps + * every account's sign-in there). Its settings and layouts are not listed. A card's own account home or project inside + * them stays its own (agentRoots, the project). plugin-data is not a marker: a project or account home inside it names + * that folder in its own paths. */ export function canvasTtyPrivateData(userDataPath: string): PrivateData { const names = ['agent-control', join('browser', 'runtime'), join('lifecycle', 'runtime'), join('orchestration', 'runtime'), - 'provider-secrets.bin', 'plugin-secrets', 'account-homes', 'github-oauth.json', 'launch-runs']; + 'provider-secrets.bin', 'plugin-secrets', 'account-homes', 'github-oauth.json', 'launch-runs', 'claude-draft', 'plugin-data']; return { appRoots: [userDataPath], paths: names.map(name => join(userDataPath, name)), - markers: ['agent-control', 'provider-secrets', 'plugin-secrets', 'account-homes', 'github-oauth'] + markers: ['agent-control', 'provider-secrets', 'plugin-secrets', 'account-homes', 'github-oauth', 'claude-draft'] }; } /** The first deny rule a set of facts breaks, in a fixed order. */ export function denyRule(facts: HardFacts): BaseDenyRule | null { if (facts.appPrivate) return 'app-private'; + if (facts.uncheckable) return 'uncheckable'; if (facts.elevation) return 'elevation'; if (facts.pipeToShell) return 'pipe-to-shell'; if (facts.downloadExec) return 'download-exec'; @@ -112,7 +192,8 @@ export function denyRule(facts: HardFacts): BaseDenyRule | null { /** * Base protection for one hook call: a deny with its message, or null (no opinion). `root` is the session's * working folder; `agentRoots` the agent's own config folders, whose plan and memory folders are not "outside". - * Any failure is null: the rules only ever add a deny. + * A failure while checking a shell, file-edit or read call denies it as `uncheckable` (fail closed); any other tool + * has no opinion (null). */ export function checkBaseProtection(input: { toolName: string; toolInput: unknown; preview?: string | null; root: string; commandCwd?: string | null; @@ -123,6 +204,7 @@ export function checkBaseProtection(input: { try { const action = actionFromHook(input.toolName, input.toolInput, input.preview ?? null); if (action.kind === null) return null; + if (action.uncheckable) return { rule: 'uncheckable', message: DENY_MESSAGES.uncheckable }; // The agent's current folder (the hook reports it) resolves relative paths; the working folder stays the root. if (!action.commandCwd && input.commandCwd) action.commandCwd = input.commandCwd; const facts = analyzeAction(action, input.root, { @@ -134,7 +216,8 @@ export function checkBaseProtection(input: { if (!rule) return null; return { rule, message: rule === 'write-outside' && writesOnlyToTemp(facts) ? TEMP_WRITE_MESSAGE : DENY_MESSAGES[rule] }; } catch { - return null; + const checked = SHELL_TOOLS.has(input.toolName) || EDIT_TOOLS.has(input.toolName) || READ_TOOLS.has(input.toolName); + return checked ? { rule: 'uncheckable', message: ANALYZER_FAILURE_MESSAGE } : null; } } diff --git a/src/main/services/safety/commandFacts.ts b/src/main/services/safety/commandFacts.ts index fe515213d..6a24628fb 100644 --- a/src/main/services/safety/commandFacts.ts +++ b/src/main/services/safety/commandFacts.ts @@ -12,15 +12,19 @@ import { lexShell, shellQuote, type Segment, type Word } from './shellParse.ts'; * data (its access tokens, secret stores and control sockets). */ -/** One tool call, source-neutral: a shell command or the files a file tool writes. */ +/** One tool call, source-neutral: a shell command, the files a file tool writes, or the files a read-only tool reads. */ export interface ToolAction { - kind: 'shell' | 'edit' | null; + kind: 'shell' | 'edit' | 'read' | null; /** The shell command; null when none could be read. */ command: string | null; /** The command's own working directory, if the agent said one. */ commandCwd: string | null; - /** Paths a file tool writes. */ + /** Paths a file tool writes, or (`read`) the files and glob patterns a read-only tool reads. */ paths: string[]; + /** `read` only: folders a search or listing tool walks (Grep's or Glob's `path`). */ + searched?: string[]; + /** The call names more paths, or longer ones, than are checked: it is refused, not checked in part. */ + uncheckable?: boolean; } export type Where = 'inside' | 'outside' | 'unresolved'; @@ -51,6 +55,11 @@ export interface HardFacts { * (inside the project or not) is unknown. */ unknownTarget: boolean; + /** + * More than the rules check in one call (paths against the private data, brace expansions, nested shells): the rest + * was not looked at, so the call is refused rather than let through unchecked. + */ + uncheckable: boolean; } /** @@ -228,7 +237,7 @@ interface Acc { ctx: PathContext; writes: Target[]; deletes: Target[]; - flags: { elevation: boolean; pipeToShell: boolean; downloadExec: boolean; disk: boolean; forkBomb: boolean; appPrivate: boolean; unknownTarget: boolean }; + flags: { elevation: boolean; pipeToShell: boolean; downloadExec: boolean; disk: boolean; forkBomb: boolean; appPrivate: boolean; unknownTarget: boolean; uncheckable: boolean }; depth: number; budget: number; /** Nesting of argv inside wrappers (nohup, env, xargs, su -c, busybox …): bounded like substitutions. */ @@ -243,7 +252,8 @@ function wordOf(text: string): Word { return { text, quoted: false, vars: [], su /** Analyses one shell command string (recursively for `bash -c`, `eval`, substitutions). */ function analyzeText(command: string, cwd: string | null, acc: Acc): string | null { - if (acc.depth > 4 || --acc.budget < 0) return cwd; + // Deeper or more nested commands than are followed: the rest is not checked, so the call is refused. + if (acc.depth > 4 || --acc.budget < 0) { acc.flags.uncheckable = true; return cwd; } acc.depth++; try { const lexed = lexShell(command); @@ -543,7 +553,11 @@ function runInterpreter(program: string, argWords: Word[], cwd: string | null, a } if (stdin.pipeIn) acc.flags.pipeToShell = true; // A program read from a heredoc (`python3 - < privateHit(text, cwd, acc, false)) || codeNamesPrivate(stdin.heredoc, acc.ctx))) acc.flags.appPrivate = true; + if (stdin.heredoc !== null) { + const found = privateCandidates(stdin.heredoc, acc.ctx); + if (found === null) acc.flags.uncheckable = true; + else if (found.some(text => privateHit(text, cwd, acc, false)) || codeNamesPrivate(stdin.heredoc, acc.ctx)) acc.flags.appPrivate = true; + } return cwd; } @@ -560,14 +574,20 @@ const CONTROL_CLI = /^canvastty-control(?:\.mjs)?$/u; /** Programs that walk folders by themselves: naming a folder that holds private data reads it. */ const RECURSIVE = new Set(['rg', 'ag', 'ack', 'find', 'fd', 'tar', 'bsdtar', 'zip', '7z', 'rsync', 'ditto', 'scp', 'rclone']); const GLOB_CHAR = /[*?[]/u; -const MAX_PRIVATE_CHECKS = 256; +const MAX_PRIVATE_CHECKS = 1024; +/** Paths found inside one word or program text (`--opt=/a:/b`, interpreter code). */ +const MAX_CANDIDATES = 256; const MAX_SCANNED_TEXT = 64 * 1024; const parts = (path: string): string[] => path.split(/[\\/]+/u).filter(Boolean); +/** macOS (APFS, HFS+) and Windows file systems ignore case by default: a pattern there matches names in any case. */ +const FOLD_CASE = process.platform === 'darwin' || process.platform === 'win32'; +const MAX_BRACE_EXPANSIONS = 64; + /** A glob component (`token-*`, `ctty-control-????`) against one real name. */ function componentMatches(pattern: string, name: string): boolean { - if (!GLOB_CHAR.test(pattern)) return pattern === name; + if (!GLOB_CHAR.test(pattern)) return FOLD_CASE ? pattern.toLowerCase() === name.toLowerCase() : pattern === name; let source = ''; for (let i = 0; i < pattern.length; i++) { const char = pattern[i]!; @@ -580,7 +600,80 @@ function componentMatches(pattern: string, name: string): boolean { i = close; } else source += char.replace(/[.+^${}()|\\\]]/gu, '\\$&'); } - try { return new RegExp(`^${source}$`, 'u').test(name); } catch { return true; } + try { return new RegExp(`^${source}$`, FOLD_CASE ? 'iu' : 'u').test(name); } catch { return true; } +} + +/** + * A glob's components against a real path's: `full` when the glob names the path or something inside it, `ancestor` + * when the whole glob already matched a folder above it. `**` matches zero or more components; a glob that only + * passes through the path on its way to other names (`~/**` then `*.ts`, past a private folder) does not name it. + */ +function globAgainst(pattern: readonly string[], target: readonly string[]): { full: boolean; ancestor: boolean } { + const close = (states: Set): Set => { + for (const state of [...states]) for (let i = state; i < pattern.length && pattern[i] === '**'; i++) states.add(i + 1); + return states; + }; + let states = close(new Set([0])); + let named = new Set(); + let ancestor = false; + for (const name of target) { + if (states.has(pattern.length)) ancestor = true; + const matched = new Set(); + const within = new Set(); + for (const state of states) { + if (state >= pattern.length) continue; + if (pattern[state] === '**') within.add(state); + else if (componentMatches(pattern[state]!, name)) matched.add(state + 1); + } + named = close(matched); + states = close(new Set([...named, ...within])); + if (!states.size) return { full: false, ancestor }; + } + // Named: the last name matched a component of the glob, or all that is left of the glob is `**`. + const full = [...states].some(state => named.has(state) || pattern.slice(state).every(part => part === '**')); + return { full, ancestor }; +} + +/** + * The words a shell's brace expansion makes of `text` (`a.{bin,json}` is `a.bin` and `a.json`); `text` itself when it + * has none. Sequences (`{1..3}`) are not expanded. Null when it makes more than MAX_BRACE_EXPANSIONS words: they are + * not checked, and the call is refused. + */ +function braceExpansions(text: string): string[] | null { + const out: string[] = []; + let over = false; + const walk = (value: string): void => { + if (over) return; + if (out.length >= MAX_BRACE_EXPANSIONS) { over = true; return; } + for (let open = value.indexOf('{'); open >= 0; open = value.indexOf('{', open + 1)) { + let depth = 0; + let end = -1; + const commas: number[] = []; + for (let i = open; i < value.length; i++) { + if (value[i] === '{') depth++; + else if (value[i] === '}' && --depth === 0) { end = i; break; } + else if (value[i] === ',' && depth === 1) commas.push(i); + } + if (end < 0) break; + if (!commas.length) continue; + const bounds = [open, ...commas, end]; + for (let k = 0; k + 1 < bounds.length; k++) walk(value.slice(0, open) + value.slice(bounds[k]! + 1, bounds[k + 1]) + value.slice(end + 1)); + return; + } + out.push(value); + }; + walk(text); + return over ? null : out; +} + +/** The gateway socket folder under a temporary folder that holds `abs` (`$TMPDIR/ctty-control-XXXX`), or null. */ +function socketFolder(abs: string, ctx: PathContext): string | null { + for (const temp of ctx.tempRoots) { + if (!isPathInside(temp, abs, { allowRoot: false })) continue; + const first = parts(relative(temp, abs))[0] ?? ''; + if (SOCKET_DIR.test(first)) return join(temp, first); + } + return null; } /** @@ -589,33 +682,31 @@ function componentMatches(pattern: string, name: string): boolean { */ function privatePath(abs: string, glob: boolean, recursive: boolean, ctx: PathContext): boolean { if (!glob) { - // The project, and the agent's own config folder (an account home it was launched with), are its own. - if (isPathInside(ctx.rootReal, abs) || ctx.agentRoots.some(dir => isPathInside(dir, abs))) return false; - if (ctx.privatePaths.some(path => isPathInside(path, abs))) return true; if (recursive && ctx.privatePaths.some(path => isPathInside(abs, path)) && ctx.appRoots.some(root => isPathInside(root, abs))) return true; - return ctx.tempRoots.some(temp => isPathInside(temp, abs, { allowRoot: false }) && SOCKET_DIR.test(parts(relative(temp, abs))[0] ?? '')); + const socket = socketFolder(abs, ctx); + const holders = [...ctx.privatePaths.filter(path => isPathInside(path, abs)), ...(socket ? [socket] : [])]; + if (!holders.length) return false; + // The project, and the agent's own config folder (an account home it was launched with), are its own, but only + // below the private data: a broad folder (HOME, or one that holds the app's data or a temporary folder) does not + // make the private data inside it the agent's. A folder inside a private path (that account home) stays its own. + const owns = (root: string): boolean => isPathInside(root, abs) && !holders.some(holder => isPathInside(root, holder)); + return !(owns(ctx.rootReal) || ctx.agentRoots.some(owns)); } // A glob: the folders before its first wildcard resolved, the rest matched name by name. const all = parts(abs); const first = all.findIndex(part => GLOB_CHAR.test(part)); const prefix = realish((abs.startsWith('/') ? '/' : '') + all.slice(0, first).join('/')); const pattern = [...parts(prefix), ...all.slice(first)]; - const matchesFrom = (target: string[]): number => { - let i = 0; - while (i < pattern.length && i < target.length && componentMatches(pattern[i]!, target[i]!)) i++; - return i; - }; for (const path of ctx.privatePaths) { - const target = parts(path); - const matched = matchesFrom(target); - if (matched >= target.length) return true; - if (matched === pattern.length && recursive && ctx.appRoots.some(root => parts(root).length <= pattern.length)) return true; + const { full, ancestor } = globAgainst(pattern, parts(path)); + if (full) return true; + if (ancestor && recursive && ctx.appRoots.some(root => parts(root).length <= pattern.length)) return true; } return ctx.tempRoots.some(temp => { const target = parts(temp); - if (matchesFrom(target) < target.length || pattern.length <= target.length) return false; + if (pattern.length <= target.length || !globAgainst(pattern.slice(0, target.length), target).full) return false; const next = pattern[target.length]!; - return SOCKET_DIR.test(next) || next.startsWith('ctty') && GLOB_CHAR.test(next); + return next === '**' || SOCKET_DIR.test(next) || next.startsWith('ctty') && GLOB_CHAR.test(next); }); } @@ -627,11 +718,19 @@ function codePath(text: string, ctx: PathContext): string | null { return isAbsolute(value) && value.length > 1 ? value : null; } -/** Paths inside a word or a program text: after `=` or `:` (`--unix-socket=P`, `UNIX-CONNECT:P`), quoted strings, bare tokens. */ -function privateCandidates(text: string, ctx: PathContext): string[] { - if (text.length > MAX_SCANNED_TEXT) text = text.slice(0, MAX_SCANNED_TEXT); +/** + * Paths inside a word or a program text: after `=` or `:` (`--unix-socket=P`, `UNIX-CONNECT:P`), quoted strings, bare + * tokens. Null when the text is longer, or holds more paths, than are checked. + */ +function privateCandidates(text: string, ctx: PathContext): string[] | null { + if (text.length > MAX_SCANNED_TEXT) return null; const found = new Set(); - const add = (value: string | undefined): void => { const path = value ? codePath(value, ctx) : null; if (path && found.size < 64) found.add(path); }; + let over = false; + const add = (value: string | undefined): void => { + const path = value ? codePath(value, ctx) : null; + if (!path || found.has(path)) return; + if (found.size >= MAX_CANDIDATES) over = true; else found.add(path); + }; for (const match of text.matchAll(/[=:]((?:~|\$\{?(?:HOME|TMPDIR)\}?|[A-Za-z]:[\\/]|\\\\|\/)[^\s,;'"`()<>|&]*)/gu)) { // An HTTP(S) authority is not a UNC path. Keep scanning other paths in the same text. if (match[0].startsWith('://') && /(?:^|[^A-Za-z0-9+.-])https?$/iu.test(text.slice(0, match.index))) continue; @@ -639,7 +738,7 @@ function privateCandidates(text: string, ctx: PathContext): string[] { } for (const match of text.matchAll(/(['"`])([^'"`\n]{1,4096}?)\1/gu)) add(match[2]); for (const token of text.split(/[\s,;()[\]{}<>|&'"`=]+/u)) if (/^(?:~|\$\{?(?:HOME|TMPDIR)\}?|[A-Za-z]:[\\/]|\\\\|\/)/u.test(token)) add(token); - return [...found]; + return over ? null : [...found]; } /** Interpreter code that builds a private path from pieces: an app folder's name together with a private name. */ @@ -650,7 +749,8 @@ function codeNamesPrivate(code: string, ctx: PathContext): boolean { } function privateHit(text: string, cwd: string | null, acc: Acc, recursive: boolean, glob = GLOB_CHAR.test(text)): boolean { - if (--acc.privateBudget < 0) return false; + // More paths than are checked: the call is refused rather than checked in part. + if (--acc.privateBudget < 0) { acc.flags.uncheckable = true; return false; } if (!isAbsolute(text) && cwd === null) return false; return privatePath(glob ? resolve(cwd ?? acc.ctx.rootReal, text) : realish(resolve(cwd ?? acc.ctx.rootReal, text)), glob, recursive, acc.ctx); } @@ -669,7 +769,8 @@ function runsControlCli(words: readonly Word[]): boolean { * the command uses them. CanvasTTY's own control CLI is the one program that may name its descriptor. */ function checkPrivate(segment: Segment, words: Word[], cwd: string | null, acc: Acc): void { - if (acc.flags.appPrivate || acc.privateBudget <= 0) return; + if (acc.flags.appPrivate) return; + if (acc.privateBudget <= 0) { acc.flags.uncheckable = true; return; } if (runsControlCli(words)) return; const recursive = words.some(word => RECURSIVE.has(programName(word.text)) || /^(?:-[a-zA-Z]*[rR][a-zA-Z]*|--recursive|--archive)$/u.test(word.text)) || programName(words[0]?.text ?? '') === 'cp' && words.some(word => /^-[a-zA-Z]*a/u.test(word.text)); @@ -679,7 +780,11 @@ function checkPrivate(segment: Segment, words: Word[], cwd: string | null, acc: if (word.substitution) continue; const text = expand(word, cwd, acc.ctx); if (text === null || text === '') continue; - const candidates = new Set([text, ...(text.length > 1 && /[=:'"\s]/u.test(text) ? privateCandidates(text, acc.ctx) : [])]); + // `provider-secrets.{bin,}`: the shell expands braces before the program sees the words. + const expanded = word.brace ? braceExpansions(text) : []; + const inner = text.length > 1 && /[=:'"\s]/u.test(text) ? privateCandidates(text, acc.ctx) : []; + if (expanded === null || inner === null) { acc.flags.uncheckable = true; return; } + const candidates = new Set([text, ...expanded, ...inner]); for (const candidate of candidates) { if (!privateHit(candidate, cwd, acc, recursive, candidate === text ? word.glob : GLOB_CHAR.test(candidate))) continue; acc.flags.appPrivate = true; @@ -1118,13 +1223,22 @@ export function commandFromArgv(argv: readonly string[]): string { return argv.m export function analyzeAction(action: ToolAction, root: string, options: { home?: string; agentRoots?: readonly string[]; privateData?: PrivateData } = {}): HardFacts { const ctx = pathContext(root, options.home, options.agentRoots, options.privateData); const acc: Acc = { ctx, writes: [], deletes: [], depth: 0, budget: 64, argvDepth: 0, privateBudget: MAX_PRIVATE_CHECKS, - flags: { elevation: false, pipeToShell: false, downloadExec: false, disk: false, forkBomb: false, appPrivate: false, unknownTarget: false } }; + flags: { elevation: false, pipeToShell: false, downloadExec: false, disk: false, forkBomb: false, appPrivate: false, unknownTarget: false, uncheckable: Boolean(action.uncheckable) } }; const commandCwd = action.commandCwd ? resolveTarget(action.commandCwd, ctx.rootReal, ctx) : null; const cwd = commandCwd ? commandCwd.abs : ctx.rootReal; if (action.kind === 'shell' && action.command) analyzeText(action.command, cwd, acc); else if (action.kind === 'edit') { acc.writes.push(...action.paths.map(path => resolveTarget(path, cwd, ctx))); if (action.paths.some(path => { const text = codePath(path, ctx) ?? path; return privateHit(text, cwd, acc, false, false); })) acc.flags.appPrivate = true; + } else if (action.kind === 'read') { + // A read-only tool is only checked against CanvasTTY's private data: reading elsewhere is not "outside". + // Glob tools expand braces (`provider-secrets.{bin,json}`); a literal name is checked as written too. + const hit = (path: string, recursive: boolean): boolean => { + const expanded = braceExpansions(path); + if (expanded === null) { acc.flags.uncheckable = true; return false; } + return [...new Set([path, ...expanded])].some(item => privateHit(codePath(item, ctx) ?? item, cwd, acc, recursive)); + }; + if (action.paths.some(path => hit(path, false)) || (action.searched ?? []).some(path => hit(path, true))) acc.flags.appPrivate = true; } const outside = acc.writes.filter(t => t.where === 'outside' && !t.device && !(t.abs && isAgentServicePath(t.abs, ctx))); return { diff --git a/src/main/services/safety/shellParse.ts b/src/main/services/safety/shellParse.ts index 71cb47430..e46c50fb0 100644 --- a/src/main/services/safety/shellParse.ts +++ b/src/main/services/safety/shellParse.ts @@ -23,6 +23,8 @@ export interface Word { glob: boolean; /** Starts with an unquoted `~`. */ tilde: boolean; + /** An unquoted `{` (a brace expansion such as `a.{bin,json}` may follow). */ + brace?: boolean; } export interface Redirect { op: string; target: Word | null; fdDup: boolean } @@ -223,6 +225,7 @@ export function lexShell(input: string): Lexed { const w = current(); if (!w.quoted && w.text === '' && c === '~') w.tilde = true; if (c === '*' || c === '?' || c === '[') w.glob = true; + if (c === '{') w.brace = true; // cmd.exe variables: %NAME%. if (c === '%') { const m = /^%([A-Za-z_][A-Za-z0-9_()]*)%/u.exec(text.slice(i)); diff --git a/src/main/services/shutdownSteps.ts b/src/main/services/shutdownSteps.ts new file mode 100644 index 000000000..670249712 --- /dev/null +++ b/src/main/services/shutdownSteps.ts @@ -0,0 +1,13 @@ +export interface ShutdownStep { + name: string; + run(): void | Promise; +} + +/** Cleanup remains ordered, but one failed save must not skip independent services. */ +export async function runShutdownSteps(steps: readonly ShutdownStep[], + warn: (message: string, error: unknown) => void = console.warn): Promise { + for (const step of steps) { + try { await step.run(); } + catch (error) { warn(`CanvasTTY shutdown: ${step.name} failed.`, error); } + } +} diff --git a/src/main/services/terminalScreenText.ts b/src/main/services/terminalScreenText.ts new file mode 100644 index 000000000..6adf8db5b --- /dev/null +++ b/src/main/services/terminalScreenText.ts @@ -0,0 +1,35 @@ +import { lazyRequire } from "../lazyRequire.ts"; + +const xterm = lazyRequire("@xterm/headless"); + +/** Rows of scrollback the rendering keeps above the screen: the tail an orchestrator reads is far shorter. */ +const SCROLLBACK_ROWS = 500; +const DEFAULT_GEOMETRY = { cols: 120, rows: 40 }; + +/** + * What a terminal shows for `data` as plain text: the scrollback rows and the screen (the alternate screen of a + * full-screen TUI such as OpenCode, which keeps no scrollback), as a headless xterm renders them, so cursor moves, + * redraws, erases and colors leave only the text a person sees. A row the terminal wrapped continues its line; + * trailing blanks and empty rows at the end are dropped. + */ +export async function terminalScreenText(data: string, geometry: { cols: number; rows: number } | null): Promise { + const { cols, rows } = geometry && geometry.cols > 0 && geometry.rows > 0 ? geometry : DEFAULT_GEOMETRY; + const terminal = new (xterm().Terminal)({ cols, rows, scrollback: SCROLLBACK_ROWS, allowProposedApi: true }); + try { + await new Promise((resolve) => terminal.write(data, resolve)); + const buffer = terminal.buffer.active; + const lines: string[] = []; + for (let row = 0; row < buffer.length; row++) { + const line = buffer.getLine(row); + if (!line) continue; + // Untrimmed, so a space the wrap fell on stays between the words; each line is trimmed once it is whole. + const text = line.translateToString(false); + if (line.isWrapped && lines.length > 0) lines[lines.length - 1] += text; + else lines.push(text); + } + while (lines.length > 0 && lines[lines.length - 1]!.trim() === "") lines.pop(); + return lines.map((line) => line.trimEnd()).join("\n"); + } finally { + terminal.dispose(); + } +} diff --git a/tests/account-home-isolation.test.mjs b/tests/account-home-isolation.test.mjs new file mode 100644 index 000000000..ee8690675 --- /dev/null +++ b/tests/account-home-isolation.test.mjs @@ -0,0 +1,371 @@ +import assert from "node:assert/strict"; +import { spawnSync } from "node:child_process"; +import { existsSync } from "node:fs"; +import { mkdir, mkdtemp, readFile, realpath, rm, symlink, writeFile } from "node:fs/promises"; +import { tmpdir } from "node:os"; +import { join } from "node:path"; +import test from "node:test"; +import { LaunchPipeline } from "../src/main/services/LaunchPipeline.ts"; +import { AgentIsolation, SANDBOX_EXEC } from "../src/main/services/isolation/AgentIsolation.ts"; +import { isolationPaths } from "../src/main/services/isolation/isolationPaths.ts"; +import { bubblewrapArguments } from "../src/main/services/isolation/bubblewrap.ts"; +import { TerminalManager } from "../src/main/services/TerminalManager.ts"; +import { DecisionHooks } from "../src/main/services/DecisionHooks.ts"; +import { canvasTtyPrivateData } from "../src/main/services/safety/baseProtection.ts"; + +// Any plugin id: the host treats every plugin's data the same way. +const PLUGIN = "com.example.accounts"; +const OTHER = "com.example.other"; +const mac = process.platform === "darwin" && existsSync(SANDBOX_EXEC); +const onMac = { skip: mac ? false : "macOS seatbelt (sandbox-exec) only" }; + +async function fixture(t) { + // Short: Unix socket paths are limited to ~104 bytes, so /tmp rather than macOS's long per-user folder. + const root = await realpath(await mkdtemp(join(process.platform === "win32" ? tmpdir() : "/tmp", "ctty-acct-"))); + t.after(() => rm(root, { recursive: true, force: true })); + const userData = join(root, "u"); + const pluginData = join(userData, "plugin-data"); + const accountData = join(pluginData, PLUGIN); + const otherData = join(pluginData, OTHER); + const homes = join(accountData, "homes"); + const project = join(root, "p"); + const temp = join(root, "t"); + const home = join(root, "h"); + await Promise.all([homes, otherData, project, temp, home].map((path) => mkdir(path, { recursive: true }))); + return { root, userData, pluginData, accountData, otherData, homes, project, temp, home }; +} + +async function prepare(f, options = {}) { + const provider = options.provider ?? "codex"; + const pluginId = options.pluginId ?? PLUGIN; + const dataDir = Object.hasOwn(options, "dataDir") ? options.dataDir : join(f.pluginData, pluginId); + const env = options.env ?? { CODEX_HOME: join(f.homes, "codex-alternate") }; + const requests = []; + const pipeline = new LaunchPipeline({ + contributors: () => [{ pluginId, pluginName: "Accounts", serviceId: "accounts", launch: { fields: [] }, secrets: false, + ...(dataDir === undefined ? {} : { dataDir }) }], + call: async (calledId, _serviceId, _method, params) => { + requests.push({ calledId, params }); + return { env, ...(options.answerExtra ?? {}) }; + }, + secret: async () => null, + runsRoot: join(f.userData, "launch-runs") + }); + const result = await pipeline.prepare({ sessionId: "fixture", provider, profile: "normal", role: "agent", cwd: f.project, + restoring: false, resume: false, options: { [pluginId]: { account: "alternate" } }, environment: null }); + return { result, requests }; +} + +function isolation(f, extra = {}) { + return new AgentIsolation({ + userDataPath: f.userData, + enabled: () => true, + tempRoot: f.temp, + hostEnvironment: { HOME: f.home, PATH: "/usr/bin:/bin" }, + ...extra + }); +} + +const fakeSeatbelt = { platform: "darwin", exists: (path) => path === SANDBOX_EXEC }; + +test("the selected account home is the exact CLI home a chosen contributor set inside its own data, and stays out of plugin RPC", async (t) => { + const f = await fixture(t); + const accountHome = join(f.homes, "codex-alternate"); + const { result, requests } = await prepare(f, { env: { CODEX_HOME: accountHome } }); + assert.equal(result.ok, true); + assert.equal(result.accountHome, accountHome); + assert.deepEqual(result.env, { CODEX_HOME: accountHome }); + assert.equal(requests.length, 1); + assert.equal(requests[0].calledId, PLUGIN); + for (const hostOnly of ["dataDir", "accountHome", "grantedPrivate"]) { + assert.equal(Object.hasOwn(requests[0].params, hostOnly), false, `${hostOnly} is host-only`); + } + await result.cleanup(); + + for (const [provider, key] of [["claude", "CLAUDE_CONFIG_DIR"], ["grok", "GROK_HOME"]]) { + const own = join(f.homes, `${provider}-alternate`); + const other = await prepare(f, { provider, env: { [key]: own } }); + assert.equal(other.result.ok, true); + assert.equal(other.result.accountHome, own, provider); + await other.result.cleanup(); + } +}); + +test("no account home without host authority, outside the contributor's own data, at its data root or through another variable", async (t) => { + const f = await fixture(t); + const cases = [ + { name: "no host data folder", dataDir: undefined, env: { CODEX_HOME: join(f.homes, "codex-a") } }, + { name: "another plugin's data", env: { CODEX_HOME: join(f.otherData, "codex-a") } }, + { name: "the plugin's whole data folder", env: { CODEX_HOME: f.accountData } }, + { name: "all plugin data", env: { CODEX_HOME: f.pluginData } }, + { name: "a path that climbs out", env: { CODEX_HOME: `${f.homes}/../../${OTHER}/codex-a` } }, + { name: "another provider's variable", env: { CLAUDE_CONFIG_DIR: join(f.homes, "claude-a") } }, + { name: "a variable that is not a CLI home", env: { SOME_DIR: join(f.homes, "codex-a") } }, + { name: "a Claude home for Codex", provider: "claude", env: { CODEX_HOME: join(f.homes, "codex-a") } }, + { name: "outside plugin data", env: { CODEX_HOME: join(f.root, "elsewhere") } } + ]; + for (const row of cases) { + const { result } = await prepare(f, row); + assert.equal(result.ok, true, row.name); + assert.equal(result.accountHome, undefined, row.name); + await result.cleanup(); + } +}); + +test("a plugin cannot name an isolation grant in its answer", async (t) => { + const f = await fixture(t); + const accountHome = join(f.homes, "codex-alternate"); + for (const key of ["accountHome", "grantedPrivate", "dataDir"]) { + const { result } = await prepare(f, { env: { CODEX_HOME: accountHome }, answerExtra: { [key]: [f.pluginData] } }); + assert.equal(result.ok, false, key); + assert.match(result.reason, /unknown key/u); + } +}); + +test("an isolated agent's paths hide all plugin data and reopen only the selected home", async (t) => { + const f = await fixture(t); + const selected = join(f.homes, "codex-alternate"); + const sibling = join(f.homes, "codex-other"); + await Promise.all([selected, sibling].map((path) => mkdir(path))); + const base = { provider: "codex", cwd: f.project, sessionTemp: join(f.temp, "s"), userDataPath: f.userData, sessionId: "s", + hostEnvironment: { HOME: f.home } }; + + const plain = isolationPaths({ ...base, env: { HOME: f.home } }); + assert.ok(plain.unreadable.includes(f.pluginData), "plugin data is private app data"); + assert.ok(!plain.readableAgain.some((path) => path.startsWith(f.pluginData)), "nothing of it is reopened without a selection"); + + const paths = isolationPaths({ ...base, env: { HOME: f.home, CODEX_HOME: selected }, grantedPrivate: [selected] }); + assert.ok(paths.unreadable.includes(f.pluginData)); + assert.deepEqual(paths.readableAgain.filter((path) => path.startsWith(f.pluginData)), [selected]); + assert.ok(paths.writable.includes(selected), "the CLI's own home stays writable"); + assert.ok(!paths.readableAgain.includes(sibling) && !paths.readableAgain.includes(f.homes) && !paths.readableAgain.includes(f.accountData)); + + // The CLI started with a home in plugin data that the launch was not handed: refused, never left readable. + assert.throws(() => isolationPaths({ ...base, env: { HOME: f.home, CODEX_HOME: sibling }, grantedPrivate: [selected] }), + /overlaps protected CanvasTTY data/u); + + // Linux: an empty tmpfs over all plugin data, then only the selected home bound back (writable: it is the CLI's). + // Its config.toml exists (LinuxHostPaths puts a placeholder there before a launch) and is read-only again. + const kinds = new Map([[f.project, "directory"], [f.pluginData, "directory"], [selected, "directory"], [sibling, "directory"], + [join(selected, "config.toml"), "file"]]); + const text = bubblewrapArguments(paths, { command: "/usr/bin/codex", args: [], cwd: f.project }, (path) => kinds.get(path) ?? null).join(" "); + const afterHidden = text.slice(text.indexOf(`--tmpfs ${f.pluginData}`)); + assert.ok(text.includes(`--tmpfs ${f.pluginData}`)); + assert.ok(afterHidden.includes(`--bind ${selected} ${selected}`)); + assert.ok(!text.includes(sibling)); +}); + +test("AgentIsolation.wrap reopens only a real folder inside one plugin's data, never through a symbolic link", async (t) => { + const f = await fixture(t); + const sandbox = isolation(f, fakeSeatbelt); + const wrap = (home, env = { CODEX_HOME: home }) => sandbox.wrap({ sessionId: "a", provider: "codex", cwd: f.project, command: "/bin/sh", + args: [], env: { HOME: f.home, PATH: "/usr/bin:/bin", ...env }, profile: "acceptEdits", accountHome: home }); + + const target = join(f.otherData, "secret"); + const alias = join(f.homes, "codex-alias"); + await mkdir(target, { recursive: true }); + await symlink(target, alias, process.platform === "win32" ? "junction" : "dir"); + // The refusal says once that the agent was not started. + const refusal = (why) => ({ name: "LaunchRefusal", message: `agent isolation could not be set up: ${why} The agent was not started without it.` }); + assert.throws(() => wrap(alias), refusal("The selected account home is missing or crosses a symbolic link.")); + assert.throws(() => wrap(join(f.homes, "codex-missing")), refusal("The selected account home is missing or crosses a symbolic link.")); + assert.throws(() => wrap(f.accountData), refusal("The selected account home must be a folder inside one plugin's data.")); + assert.throws(() => wrap("relative/home"), refusal("The selected account home must be an exact absolute folder.")); + + const selected = join(f.homes, "codex-alternate"); + await mkdir(selected); + const ok = wrap(selected); + try { + const profile = await readFile(ok.args[1], "utf8"); + assert.ok(profile.includes(`(deny file-read* file-write*`) && profile.includes(`(subpath "${f.pluginData}")`)); + assert.match(profile, new RegExp(`\\(allow file-read\\*[^\\n]*\\(subpath "${selected.replace(/[.*+?^${}()|[\]\\]/gu, "\\$&")}"\\)`, "u")); + } finally { + ok.cleanup(); + } + // An environment that replaced the CLI's home: nothing of plugin data is reopened, and the launch is not refused for it. + const elsewhere = join(f.root, "elsewhere"); + await mkdir(elsewhere); + const replaced = wrap(selected, { CODEX_HOME: elsewhere }); + try { + assert.ok(!(await readFile(replaced.args[1], "utf8")).includes(`(subpath "${selected}")`)); + } finally { + replaced.cleanup(); + } +}); + +test("a project inside plugin data that CanvasTTY placed there is reopened exactly; all of plugin data or a whole plugin's data is never a project", async (t) => { + const f = await fixture(t); + const worktree = join(f.otherData, "worktrees", "one"); + const sibling = join(f.otherData, "worktrees", "two"); + await Promise.all([worktree, sibling].map((path) => mkdir(path, { recursive: true }))); + const base = { provider: "codex", sessionTemp: join(f.temp, "s"), userDataPath: f.userData, sessionId: "s", env: { HOME: f.home } }; + // An environment of the plugin that owns the folder placed the launch there (TerminalManager.placedProjects). + const placed = { placedProjects: [{ folder: worktree, pluginId: OTHER }] }; + const paths = isolationPaths({ ...base, cwd: worktree, ...placed }); + assert.ok(paths.unreadable.includes(f.pluginData)); + assert.deepEqual(paths.readableAgain.filter((path) => path.startsWith(f.pluginData)), [worktree]); + assert.ok(paths.writable.includes(worktree)); + const plan = isolationPaths({ ...base, cwd: worktree, readOnlyProject: true, ...placed }); + assert.ok(plan.readableAgain.includes(worktree) && !plan.writable.includes(worktree), "plan keeps it read-only"); + for (const cwd of [f.pluginData, f.otherData]) { + assert.throws(() => isolationPaths({ ...base, cwd, placedProjects: [{ folder: cwd }] }), /private plugin data/u, cwd); + } +}); + +test("seatbelt, for real: an isolated agent reads and writes only its selected home in plugin data", onMac, async (t) => { + const f = await fixture(t); + const selected = join(f.homes, "codex-alternate"); + const sibling = join(f.homes, "codex-other"); + const otherPlugin = join(f.otherData, "worktrees", "sibling"); + const settings = join(f.accountData, "accounts.json"); + await Promise.all([selected, sibling, otherPlugin].map((path) => mkdir(path, { recursive: true }))); + await writeFile(join(selected, "marker"), "selected fixture home\n"); + await writeFile(join(sibling, "marker"), "sibling fixture home\n"); + await writeFile(join(otherPlugin, "marker"), "other plugin fixture\n"); + await writeFile(settings, "private account settings\n"); + const wrapped = isolation(f).wrap({ + sessionId: "selected-account", + provider: "codex", + cwd: f.project, + command: "/bin/sh", + args: ["-c", [ + 'cat "$CODEX_HOME/marker" >/dev/null && echo OWN-READ-ok', + 'echo actor > "$CODEX_HOME/actor.txt" && echo OWN-WRITE-ok', + 'echo actor > "$CODEX_HOME/config.toml" 2>/dev/null || echo CLI-SETTINGS-WRITE-denied', + `cat ${JSON.stringify(join(sibling, "marker"))} >/dev/null 2>&1 || echo ACCOUNT-SIBLING-READ-denied`, + `echo actor >> ${JSON.stringify(join(sibling, "marker"))} 2>/dev/null || echo ACCOUNT-SIBLING-WRITE-denied`, + `cat ${JSON.stringify(join(otherPlugin, "marker"))} >/dev/null 2>&1 || echo OTHER-PLUGIN-READ-denied`, + `cat ${JSON.stringify(settings)} >/dev/null 2>&1 || echo SETTINGS-READ-denied`, + `echo actor >> ${JSON.stringify(settings)} 2>/dev/null || echo SETTINGS-WRITE-denied`, + `ls ${JSON.stringify(f.pluginData)} >/dev/null 2>&1 || echo PLUGIN-DATA-LIST-denied` + ].join("; ")], + env: { HOME: f.home, PATH: "/usr/bin:/bin", CODEX_HOME: selected }, + profile: "acceptEdits", + accountHome: selected + }); + try { + const result = spawnSync(wrapped.command, wrapped.args, { cwd: f.project, env: wrapped.env, encoding: "utf8", timeout: 20_000 }); + assert.equal(result.status, 0, result.stderr); + assert.deepEqual(result.stdout.split("\n").filter(Boolean), [ + "OWN-READ-ok", "OWN-WRITE-ok", "CLI-SETTINGS-WRITE-denied", "ACCOUNT-SIBLING-READ-denied", "ACCOUNT-SIBLING-WRITE-denied", + "OTHER-PLUGIN-READ-denied", "SETTINGS-READ-denied", "SETTINGS-WRITE-denied", "PLUGIN-DATA-LIST-denied" + ]); + assert.equal(await readFile(join(selected, "actor.txt"), "utf8"), "actor\n"); + assert.equal(existsSync(join(selected, "config.toml")), false); + assert.equal(await readFile(join(sibling, "marker"), "utf8"), "sibling fixture home\n"); + assert.equal(await readFile(settings, "utf8"), "private account settings\n"); + } finally { + wrapped.cleanup(); + } +}); + +test("seatbelt, for real: without a selected home an isolated agent reads no plugin data", onMac, async (t) => { + const f = await fixture(t); + const home = join(f.homes, "codex-a"); + await mkdir(home); + await writeFile(join(home, "auth.json"), "FAKE-TOKEN\n"); + const wrapped = isolation(f).wrap({ sessionId: "plain", provider: "codex", cwd: f.project, command: "/bin/sh", + args: ["-c", `cat ${JSON.stringify(join(home, "auth.json"))} 2>/dev/null || echo TOKEN-READ-denied`], + env: { HOME: f.home, PATH: "/usr/bin:/bin" }, profile: "acceptEdits" }); + try { + const result = spawnSync(wrapped.command, wrapped.args, { cwd: f.project, env: wrapped.env, encoding: "utf8", timeout: 20_000 }); + assert.equal(result.stdout.trim(), "TOKEN-READ-denied"); + } finally { + wrapped.cleanup(); + } +}); + +test("a launch hands the isolation layer the account home its contributor selected", async (t) => { + const f = await fixture(t); + const selected = join(f.homes, "codex-alternate"); + await mkdir(selected); + const pipeline = new LaunchPipeline({ + contributors: () => [{ pluginId: PLUGIN, pluginName: "Accounts", serviceId: "accounts", dataDir: f.accountData, secrets: false, + launch: { fields: [{ key: "account", label: "Account", kind: "text", default: "" }] } }], + call: async () => ({ env: { CODEX_HOME: selected } }), + secret: async () => null, + runsRoot: join(f.userData, "launch-runs") + }); + const calls = []; + const spawner = (command, args, options) => { + calls.push({ command, args, options }); + return { pid: 41_000 + calls.length, process: command, write() {}, resize() {}, kill() {}, pause() {}, resume() {}, + onData() { return { dispose() {} }; }, onExit() { return { dispose() {} }; } }; + }; + const registry = { + get: (provider) => ({ state: "available", provider, executable: `/resolved/${provider}`, launcher: "native", environment: { PATH: "/usr/bin" }, checked: [] }), + snapshot: () => ({}) + }; + const terminals = new TerminalManager(() => undefined, registry, undefined, undefined, true, spawner); + t.after(() => terminals.shutdown()); + terminals.configureLaunchPipeline(pipeline); + const wraps = []; + const layer = isolation(f, fakeSeatbelt); + terminals.configureIsolation({ + containment: () => true, + decide: (input) => layer.decide(input), + wrap: (launch) => { + wraps.push(launch); + return { command: SANDBOX_EXEC, args: ["-f", "/p.sb", launch.command, ...launch.args], env: launch.env, cleanup() {} }; + } + }); + terminals.create({ provider: "codex", profile: "acceptEdits", cwd: f.project, position: { x: 0, y: 0 }, + launchOptions: { [PLUGIN]: { account: "alternate" } } }); + const deadline = Date.now() + 5_000; + while (wraps.length === 0 && Date.now() < deadline) await new Promise((resolve) => setTimeout(resolve, 10)); + assert.equal(wraps.length, 1); + assert.equal(wraps[0].accountHome, selected); + assert.equal(wraps[0].env.CODEX_HOME, selected); + assert.equal(calls.at(-1).command, SANDBOX_EXEC); +}); + +test("base protection: a card's own selected account home is its own; other accounts' homes and the rest of plugin data stay private", async (t) => { + const f = await fixture(t); + const selected = join(f.homes, "claude-alternate"); + const neighbour = join(f.homes, "claude-other"); + await Promise.all([selected, neighbour].map((path) => mkdir(path))); + const pipeline = new LaunchPipeline({ + contributors: () => [{ pluginId: PLUGIN, pluginName: "Accounts", serviceId: "accounts", dataDir: f.accountData, secrets: false, + launch: { fields: [{ key: "account", label: "Account", kind: "text", default: "" }] } }], + call: async (_pluginId, _serviceId, _method, params) => ({ env: params.provider === "codex" ? { CODEX_HOME: `${selected}/` } : { CLAUDE_CONFIG_DIR: selected } }), + secret: async () => null, + runsRoot: join(f.userData, "launch-runs") + }); + const spawned = []; + const spawner = (command) => { + spawned.push(command); + return { pid: 42_000 + spawned.length, process: command, write() {}, resize() {}, kill() {}, pause() {}, resume() {}, + onData() { return { dispose() {} }; }, onExit() { return { dispose() {} }; } }; + }; + const registry = { + get: (provider) => ({ state: "available", provider, executable: `/resolved/${provider}`, launcher: "native", environment: { PATH: "/usr/bin" }, checked: [] }), + snapshot: () => ({}) + }; + const terminals = new TerminalManager(() => undefined, registry, undefined, undefined, true, spawner); + t.after(() => terminals.shutdown()); + terminals.configureLaunchPipeline(pipeline); + const hooks = new DecisionHooks({ baseProtection: () => true, services: () => [], call: async () => null, home: f.home, + privateData: canvasTtyPrivateData(f.userData), session: (id) => terminals.decisionContext(id) }); + const decide = async (id, toolName, toolInput) => + (await hooks.decide(id, { toolName, toolInput, toolInputPreview: null, cwd: null, truncated: false }, new AbortController().signal)).behavior; + // Person-launched (no isolation layer): base protection is the guard here. + const withAccount = (provider) => terminals.create({ provider, profile: "auto", cwd: f.project, position: { x: 0, y: 0 }, + launchOptions: { [PLUGIN]: { account: "alternate" } } }).id; + const plain = terminals.create({ provider: "claude", profile: "auto", cwd: f.project, position: { x: 0, y: 0 } }).id; + const cards = [withAccount("claude"), withAccount("codex")]; + const deadline = Date.now() + 5_000; + while (spawned.length < 3 && Date.now() < deadline) await new Promise((resolve) => setTimeout(resolve, 10)); + assert.equal(spawned.length, 3); + for (const id of cards) { + assert.ok(terminals.decisionContext(id).configDirs.includes(selected), id); + assert.equal(await decide(id, "Read", { file_path: join(selected, "auth.json") }), "none", "own home"); + assert.equal(await decide(id, "Bash", { command: `cat '${join(selected, "auth.json")}'` }), "none", "own home"); + assert.equal(await decide(id, "Read", { file_path: join(neighbour, "auth.json") }), "deny", "another account's home"); + assert.equal(await decide(id, "Bash", { command: `cat '${join(neighbour, "auth.json")}'` }), "deny", "another account's home"); + assert.equal(await decide(id, "Read", { file_path: join(f.otherData, "state.json") }), "deny", "another plugin's data"); + assert.equal(await decide(id, "Read", { file_path: join(f.project, "a.ts") }), "none", "the project"); + } + assert.equal(terminals.decisionContext(plain).configDirs.includes(selected), false); + assert.equal(await decide(plain, "Read", { file_path: join(selected, "auth.json") }), "deny", "not this card's home"); + assert.equal(await decide(plain, "Bash", { command: "cat a.ts && grep -r todo ." }), "none", "a normal project is unaffected"); +}); diff --git a/tests/agent-control.test.mjs b/tests/agent-control.test.mjs index 58ad664bc..8b7ce971d 100644 --- a/tests/agent-control.test.mjs +++ b/tests/agent-control.test.mjs @@ -709,7 +709,7 @@ test("send appends submit unless told otherwise and rejects exited sessions", as terminals.disposeAll(); }); -test("observe returns a capped terminal tail and result reflects exit state", () => { +test("observe returns a capped terminal tail and result reflects exit state", async () => { const { terminals, control } = serviceFixture(); const parent = terminals.create({ provider: "codex", @@ -717,11 +717,11 @@ test("observe returns a capped terminal tail and result reflects exit state", () profile: "normal", position: { x: 0, y: 0 } }); - const observation = control.observe(parent.id); + const observation = await control.observe(parent.id); assert.equal(observation.sessionId, parent.id); assert.equal(observation.output.length <= 8_192, true); - const running = control.result(parent.id); + const running = await control.result(parent.id); assert.equal(running.state, "running"); assert.equal(running.exitCode, null); terminals.disposeAll(); @@ -746,8 +746,8 @@ test("cancel disposes the subagent and plain terminals are not agents", async () assert.equal(terminals.list().some((session) => session.id === child.id), false); assert.throws(() => control.send(terminal.id, "text"), /not agents/u); - assert.throws(() => control.observe(terminal.id), /not agents/u); - assert.throws(() => control.result(terminal.id), /not agents/u); + await assert.rejects(control.observe(terminal.id), /not agents/u); + await assert.rejects(control.result(terminal.id), /not agents/u); terminals.disposeAll(); }); @@ -805,8 +805,8 @@ test("an orchestrator tool call looks its sessions up by id: no other card's scr assert.equal(control.status(child.id).id, child.id); assert.equal(control.isInSubtree(parent.id, child.id), true); assert.deepEqual(control.children(parent.id).map((session) => session.id), [child.id]); - control.observe(child.id); - control.result(child.id); + await control.observe(child.id); + await control.result(child.id); assert.throws(() => control.status("missing"), /does not exist/u); assert.deepEqual(copied, [child.id, child.id], "only the observed card's own scrollback is read, and no list() snapshot of every card"); diff --git a/tests/agent-isolation.test.mjs b/tests/agent-isolation.test.mjs index 89b2266aa..2324b719f 100644 --- a/tests/agent-isolation.test.mjs +++ b/tests/agent-isolation.test.mjs @@ -1,17 +1,18 @@ import assert from "node:assert/strict"; import { spawn, spawnSync } from "node:child_process"; -import { existsSync } from "node:fs"; +import { existsSync, readFileSync } from "node:fs"; import { mkdir, mkdtemp, readFile, readdir, realpath, rm, symlink, writeFile } from "node:fs/promises"; import { createServer } from "node:net"; import { tmpdir } from "node:os"; import { delimiter, join } from "node:path"; import test from "node:test"; -import { AgentIsolation, ISOLATION_FOLDER_PREFIX, probeBubblewrap } from "../src/main/services/isolation/AgentIsolation.ts"; +import { AgentIsolation, gatewaySocketAddresses, ISOLATION_FOLDER_PREFIX, probeBubblewrap } from "../src/main/services/isolation/AgentIsolation.ts"; import { isolationPaths } from "../src/main/services/isolation/isolationPaths.ts"; import { seatbeltProfile } from "../src/main/services/isolation/seatbelt.ts"; import { bubblewrapArguments } from "../src/main/services/isolation/bubblewrap.ts"; import { AgentControlService } from "../src/main/services/AgentControlService.ts"; import { codexInsideIsolation } from "../src/main/services/terminalLaunch.ts"; +import { LaunchPipeline } from "../src/main/services/LaunchPipeline.ts"; import { TerminalManager } from "../src/main/services/TerminalManager.ts"; import { availableRegistry, fakeSpawner } from "./helpers/terminal.mjs"; @@ -180,6 +181,32 @@ test("bubblewrap: read-only root, writable project, tmpfs over what may not be r assert.deepEqual(args.slice(-5), ["--chdir", w.project, "--", "/usr/bin/claude", "--x"]); }); +test("bubblewrap: CanvasTTY's gateway socket folders are read-only, a hidden socket is masked", async (t) => { + const w = await world(t); + const lifecycle = join(w.userData, "lifecycle", "runtime"); + const gateway = join(w.base, "ctty-1000-gw"); + const sessionTemp = join(w.temp, "s"); + const paths = isolationPaths({ provider: "codex", cwd: w.project, sessionTemp, + env: w.env, gatewaySockets: [join(lifecycle, "r.sock"), join(gateway, "gateway.sock")], userDataPath: w.userData, sessionId: "s" }); + const ssh = join(w.home, ".ssh", "agent.sock"); + const kinds = new Map([[w.project, "directory"], [sessionTemp, "directory"], [lifecycle, "directory"], [gateway, "directory"], + [join(w.userData, "browser", "runtime"), "directory"], [join(w.userData, "orchestration", "runtime"), "directory"], [ssh, "socket"]]); + const args = bubblewrapArguments({ ...paths, unreadable: [...paths.unreadable, ssh] }, { command: "/usr/bin/codex", args: [], cwd: w.project }, + (path) => kinds.get(path) ?? null); + const pairs = args.map((arg, i) => `${arg} ${args[i + 1]} ${args[i + 2]}`); + // Connecting to a Unix socket needs no write access to its folder; a writable folder would let one agent replace a + // gateway socket and receive another agent's requests. + for (const folder of [lifecycle, gateway, join(w.userData, "browser", "runtime"), join(w.userData, "orchestration", "runtime")]) { + assert.ok(pairs.includes(`--ro-bind ${folder} ${folder}`), `${folder} read-only`); + assert.ok(!pairs.includes(`--bind ${folder} ${folder}`), `${folder} not writable`); + } + // This launch's own temporary folder (also a socket folder) stays writable, bound once. + assert.equal(pairs.filter((pair) => pair === `--bind ${sessionTemp} ${sessionTemp}`).length, 1); + assert.ok(!pairs.includes(`--ro-bind ${sessionTemp} ${sessionTemp}`)); + // A socket it may not reach is hidden like a file, never left in place. + assert.ok(pairs.includes(`--ro-bind /dev/null ${ssh}`)); +}); + test("the paths: another CLI's home moved by its own variable is unreadable, never this CLI's own", async (t) => { const w = await world(t); const custom = (name) => join(w.base, "creds", name); @@ -304,6 +331,86 @@ test("the launch: wrapped when the layer applies, refused (never unwrapped) when assert.deepEqual(cleaned, [auto.id], "its folder goes with the card"); }); +test("the launch: only CanvasTTY's own gateway sockets reach the layer, never a CANVASTTY_*_ADDRESS from the person or a plugin", async (t) => { + const w = await world(t); + const calls = []; + const runtimeSocket = join(w.userData, "lifecycle", "runtime", "r.sock"); + const runtime = { + prepareLaunch: () => ({ args: [], environment: { CANVASTTY_RUNTIME_ADDRESS: runtimeSocket, CANVASTTY_RUNTIME_CAPABILITY: "c" }, cleanup() {} }), + currentStatus: () => null + }; + const terminals = new TerminalManager(() => undefined, availableRegistry(), undefined, runtime, true, fakeSpawner(calls)); + t.after(() => terminals.shutdown()); + const launches = []; + terminals.configureIsolation({ + containment: () => true, + decide: (input) => new AgentIsolation({ userDataPath: w.userData, enabled: () => true, platform: "darwin", exists: () => true }).decide(input), + wrap: (launch) => { launches.push(launch); return { command: "/usr/bin/sandbox-exec", args: [launch.command, ...launch.args], env: launch.env, cleanup() {} }; } + }); + const before = process.env.CANVASTTY_FOO_ADDRESS; + process.env.CANVASTTY_FOO_ADDRESS = join(w.home, "foo.sock"); + t.after(() => { if (before === undefined) delete process.env.CANVASTTY_FOO_ADDRESS; else process.env.CANVASTTY_FOO_ADDRESS = before; }); + terminals.create({ provider: "claude", profile: "auto", cwd: w.project, position: at }); + assert.equal(launches.length, 1); + assert.equal(launches[0].env.CANVASTTY_FOO_ADDRESS, join(w.home, "foo.sock"), "the variable itself still reaches the agent"); + assert.deepEqual(launches[0].gatewaySockets, [runtimeSocket]); + assert.deepEqual(gatewaySocketAddresses({ CANVASTTY_A_ADDRESS: "/a/s.sock", CANVASTTY_B_ADDRESS: "127.0.0.1:1", OTHER_ADDRESS: "/o" }, + undefined, { CANVASTTY_C_ADDRESS: "/a/s.sock" }), ["/a/s.sock"], "absolute CanvasTTY addresses only, once each"); +}); + +test("the launch: only the control grant CanvasTTY gave an orchestrator is reopened, never one a plugin names", async (t) => { + const w = await world(t); + const calls = []; + const terminals = new TerminalManager(() => undefined, availableRegistry(), undefined, undefined, true, fakeSpawner(calls)); + t.after(() => terminals.shutdown()); + const launches = []; + terminals.configureIsolation({ + containment: () => true, + decide: (input) => new AgentIsolation({ userDataPath: w.userData, enabled: () => true, platform: "darwin", exists: () => true }).decide(input), + wrap: (launch) => { launches.push(launch); return { command: "/usr/bin/sandbox-exec", args: [launch.command, ...launch.args], env: launch.env, cleanup() {} }; } + }); + const own = join(w.userData, "agent-control", "sessions", "own", "connection.json"); + terminals.setControlConnection({ connectionPath: join(w.userData, "agent-control", "connection.json"), cliPath: "/cli.mjs", grant: () => own }); + // A launch contributor names another session's grant on an ordinary (non-orchestrator) launch. + const victim = join(w.userData, "agent-control", "sessions", "victim", "connection.json"); + const runsRoot = join(w.temp, "runs"); + await mkdir(runsRoot, { recursive: true }); + const pipeline = new LaunchPipeline({ + contributors: () => [{ pluginId: "p.one", pluginName: "Plugin p.one", serviceId: "launcher", secrets: false, + launch: { fields: [{ key: "on", label: "On", kind: "boolean", default: true }] } }], + call: async () => ({ env: { CANVASTTY_CONTROL_CONNECTION: victim } }), + secret: async () => null, + runsRoot + }); + const waitFor = async (predicate) => { + for (let tries = 0; !predicate() && tries < 500; tries++) await new Promise((resolve) => setTimeout(resolve, 10)); + assert.ok(predicate(), "in time"); + }; + // The launch pipeline refuses the name already... + terminals.configureLaunchPipeline(pipeline); + const refused = terminals.create({ provider: "claude", profile: "auto", cwd: w.project, position: at, launchOptions: { "p.one": {} } }); + const card = () => terminals.list().find((session) => session.id === refused.id); + await waitFor(() => card().status === "failed"); + assert.match(card().failureDetails, /env CANVASTTY_CONTROL_CONNECTION is reserved for CanvasTTY/u); + assert.equal(launches.length, 0); + // ...and the isolation layer does not depend on that: a value that reaches the launch environment reopens nothing. + terminals.configureLaunchPipeline({ + normalizeOptions: (...args) => pipeline.normalizeOptions(...args), + unavailable: (options) => pipeline.unavailable(options), + forgetSession: async () => undefined, + prepare: async () => ({ ok: true, env: { CANVASTTY_CONTROL_CONNECTION: victim }, args: [], secrets: [], + envSources: { CANVASTTY_CONTROL_CONNECTION: "Plugin p.one" }, thirdPartyModel: false, cleanup: async () => undefined }) + }); + terminals.create({ provider: "claude", profile: "auto", cwd: w.project, position: at, launchOptions: { "p.one": {} } }); + await waitFor(() => launches.length === 1); + assert.equal(launches[0].env.CANVASTTY_CONTROL_CONNECTION, victim, "the variable itself reaches the agent"); + assert.equal(launches[0].grantedPrivate, undefined, "but no folder is reopened for it"); + + terminals.create({ provider: "claude", profile: "auto", cwd: w.project, position: at, role: "orchestrator" }); + assert.equal(launches.length, 2); + assert.deepEqual(launches[1].grantedPrivate, [join(w.userData, "agent-control", "sessions", "own")], "an orchestrator's own grant is"); +}); + test("without a layer a subagent runs in normal (it asks) and the card says why", async (t) => { const w = await world(t); const calls = []; @@ -423,6 +530,52 @@ test("an isolated launch gets an empty git template, so git init writes no hooks } finally { wrapped.cleanup(); } }); +test("an isolated Claude keeps its Bash and editor files in the launch's own temp folder, on both layers; other CLIs are untouched", async (t) => { + const w = await world(t); + const uid = typeof process.getuid === "function" ? process.getuid() : 501; + const layers = [ + ["seatbelt", isolation(w, { platform: "darwin", exists: () => true })], + ["bubblewrap", isolation(w, { platform: "linux", bubblewrapPath: "/usr/bin/bwrap", bubblewrapProbe: () => null, exists: () => true })] + ]; + for (const [layer, isolated] of layers) { + // A CLAUDE_CODE_TMPDIR from the person's environment would point Claude back at a shared folder: replaced. + const claude = isolated.wrap({ sessionId: "c", provider: "claude", cwd: w.project, command: "/usr/bin/claude", args: [], env: { ...w.env, CLAUDE_CODE_TMPDIR: "/tmp" } }); + const codex = isolated.wrap({ sessionId: "x", provider: "codex", cwd: w.project, command: "/usr/bin/codex", args: [], env: w.env }); + try { + assert.equal(claude.env.CLAUDE_CODE_TMPDIR, claude.env.TMP, layer); + assert.equal(`${claude.env.CLAUDE_CODE_TMPDIR}/`, claude.env.TMPDIR, layer); + assert.match(claude.env.TMP, new RegExp(`^${w.temp}/${ISOLATION_FOLDER_PREFIX}`, "u"), layer); + assert.equal(codex.env.CLAUDE_CODE_TMPDIR, undefined, `${layer}: only Claude reads it`); + // The folder all of the person's Claude sessions share is not opened for writing. + const text = layer === "seatbelt" ? readFileSync(claude.args[1], "utf8") : claude.args.join(" "); + assert.ok(!text.includes(`claude-${uid}`), `${layer}: no grant of the shared /tmp/claude-${uid}`); + } finally { + claude.cleanup(); + codex.cleanup(); + } + } +}); + +test("seatbelt, for real: Claude's scratch folder is in its own temp folder; the shared /tmp/claude- stays read-only", onMac, async (t) => { + const w = await world(t); + const shared = `/private/tmp/claude-${process.getuid()}/ctty-iso-probe-${process.pid}`; + await mkdir(shared, { recursive: true }); + await writeFile(join(shared, "other-session.output"), "another session's output"); + t.after(() => rm(shared, { recursive: true, force: true })); + t.after(() => rm(`/private/tmp/claude-${process.getuid()}/ctty-iso-new-${process.pid}`, { recursive: true, force: true })); + const { lines, env } = run(w, [ + // What Claude Code does for its Bash tool: ${CLAUDE_CODE_TMPDIR:-/tmp}/claude-///tasks. + 'd="${CLAUDE_CODE_TMPDIR:-/tmp}/claude-$(id -u)/p/s/tasks"; mkdir -p "$d" && echo out > "$d/b.output" && cat "$d/b.output" >/dev/null && echo SCRATCH-ok || echo SCRATCH-denied', + `echo x >> "${shared}/other-session.output" 2>/dev/null && echo SHARED-WRITE-ok || echo SHARED-WRITE-denied`, + `: > "${shared}/planted" 2>/dev/null && echo SHARED-CREATE-ok || echo SHARED-CREATE-denied`, + `mkdir "/private/tmp/claude-$(id -u)/ctty-iso-new-${process.pid}" 2>/dev/null && echo SHARED-MKDIR-ok || echo SHARED-MKDIR-denied` + ].join("; "), { provider: "claude" }); + assert.deepEqual(lines, ["SCRATCH-ok", "SHARED-WRITE-denied", "SHARED-CREATE-denied", "SHARED-MKDIR-denied"]); + assert.equal(env.CLAUDE_CODE_TMPDIR, env.TMP); + assert.equal(await readFile(join(shared, "other-session.output"), "utf8"), "another session's output"); + assert.equal(existsSync(join(shared, "planted")), false); +}); + test("seatbelt, for real: Claude saves a refreshed sign-in in the login keychain; nothing else there; plan is read-only", onMac, async (t) => { const w = await world(t); const folder = join(w.home, "Library", "Keychains"); diff --git a/tests/agent-runtime-gateway.test.mjs b/tests/agent-runtime-gateway.test.mjs index 61ec999da..67e8bc0df 100644 --- a/tests/agent-runtime-gateway.test.mjs +++ b/tests/agent-runtime-gateway.test.mjs @@ -158,6 +158,32 @@ test("RuntimeGateway: a late start of an earlier turn does not take over from th assert.equal(gateway.currentStatus("terminal-turns"), "idle"); }); +test("RuntimeGateway: a turn whose start hook was lost still counts after an idle, but not while another turn runs", POSIX_RUNTIME_GATEWAY_TEST, async (t) => { + const root = await fixture(t); + const signals = []; + const gateway = new RuntimeGateway({ runtimeDirectory: root, onSignal: (id, signal) => signals.push({ id, signal }) }); + await gateway.start(); + t.after(() => gateway.close()); + const capability = gateway.registerSession("terminal-lost-start", "codex"); + await send(capability.address, message(capability, "working", "UserPromptSubmit", "turn-1")); + // An event of another turn while turn 1 runs is out of order. + await send(capability.address, message(capability, "working", "PostToolUse", "turn-x")); + await send(capability.address, message(capability, "idle", "Stop", "turn-1")); + // Turn 2's UserPromptSubmit never arrived: its own events still report it. + await send(capability.address, message(capability, "working", "PostToolUse", "turn-2")); + assert.equal(gateway.currentStatus("terminal-lost-start"), "working"); + // A late event of the earlier turn stays stale. + await send(capability.address, message(capability, "idle", "Stop", "turn-1")); + await send(capability.address, message(capability, "idle", "Stop", "turn-2")); + // After that idle, an id already seen does not open a turn again. + await send(capability.address, message(capability, "working", "PostToolUse", "turn-1")); + await new Promise((resolve) => setImmediate(resolve)); + assert.deepEqual(signals.map(({ signal }) => `${signal.state}:${signal.event}:${signal.turnId}`), [ + "working:UserPromptSubmit:turn-1", "idle:Stop:turn-1", "working:PostToolUse:turn-2", "idle:Stop:turn-2" + ]); + assert.equal(gateway.currentStatus("terminal-lost-start"), "idle"); +}); + test("RuntimeGateway propagates threadId for codex sessions with canonical UUID", POSIX_RUNTIME_GATEWAY_TEST, async (t) => { const root = await fixture(t); const signals = []; diff --git a/tests/agent-runtime-provider-launch.test.mjs b/tests/agent-runtime-provider-launch.test.mjs index c56a310b2..c3966d872 100644 --- a/tests/agent-runtime-provider-launch.test.mjs +++ b/tests/agent-runtime-provider-launch.test.mjs @@ -46,6 +46,7 @@ test("Claude and Codex receive automatic lifecycle hooks without prompt or respo assert.equal(settings.showStatusInTerminalTab, true); assert.match(settings.hooks.UserPromptSubmit[0].hooks[0].command, /working.*UserPromptSubmit/u); assert.match(settings.hooks.PermissionRequest[0].hooks[0].command, /needs_approval.*PermissionRequest/u); + assert.match(settings.hooks.PostToolUseFailure[0].hooks[0].command, /working.*PostToolUseFailure/u); assert.match(settings.hooks.Stop[0].hooks[0].command, /idle.*Stop/u); assert.match(settings.hooks.Stop[0].hooks[0].command, /^ELECTRON_RUN_AS_NODE='1' /u); assert.equal(JSON.stringify(settings).includes('"prompt":'), false); @@ -497,3 +498,25 @@ function runtimeOptionsFor(root, qwenSystemSettingsPath, pluginHooks) { environment: {} }; } + +test("with status hooks off a result-capturing subagent gets no lease and no lifecycle hooks", async (t) => { + const root = await fixture(t); + const registrations = []; + const gateway = { + registerSession(id, provider, capture, _grant, decisions) { + registrations.push({ id, capture, decisions }); + return { address: join(root, "runtime.sock"), terminalSessionId: id, provider, capabilityToken: "token" }; + }, + revokeTerminalSession() {}, + currentStatus: () => null + }; + const bridge = new AgentRuntimeBridge(gateway, { ...runtimeOptionsFor(root), coreHooksEnabled: false }); + for (const provider of ["codex", "opencode"]) { + const launch = bridge.prepareLaunch({ terminalSessionId: provider, provider, cwd: root, captureResult: true, decisions: false }); + assert.equal(launch.environment[AGENT_RUNTIME_ENV.capabilityToken], undefined, `${provider}: turning status hooks off means no hooks`); + assert.deepEqual(launch.args.filter((arg) => /^hooks\./u.test(arg)), []); + assert.equal(launch.environment.OPENCODE_CONFIG_CONTENT, undefined); + launch.cleanup(); + } + assert.deepEqual(registrations, []); +}); diff --git a/tests/base-protection-app-private.test.mjs b/tests/base-protection-app-private.test.mjs index e19321d0d..b7a46ed03 100644 --- a/tests/base-protection-app-private.test.mjs +++ b/tests/base-protection-app-private.test.mjs @@ -11,6 +11,7 @@ import { join } from "node:path"; import { analyzeAction } from "../src/main/services/safety/commandFacts.ts"; import { canvasTtyPrivateData, checkBaseProtection, denyRule } from "../src/main/services/safety/baseProtection.ts"; import { DecisionHooks } from "../src/main/services/DecisionHooks.ts"; +import { guardedCall } from "../src/agent-runtime/opencode-decisions.mjs"; const base = realpathSync(mkdtempSync(join(tmpdir(), "canvastty-app-private-"))); const home = join(base, "home"); @@ -102,10 +103,113 @@ test("CanvasTTY's own tokens, secret stores and sockets are refused for every re for (const command of DENY) assert.equal(rule(command), "app-private", command); }); +test("the per-card Claude Code draft folders are private data too", () => { + const draft = join(userData, "claude-draft", "card-2", "draft.1"); + assert.equal(rule(`cat ${q(draft)}`), "app-private"); + assert.equal(rule(`touch ${q(join(userData, "claude-draft", "card-2", "req.1"))}`), "app-private"); + assert.equal(rule("cat \"$HOME/Library/Application Support/canvastty/claude-draft/card-2/draft.1\""), "app-private"); + assert.equal(rule("cat notes/claude-drafting.md"), null, "a project file with a similar name stays allowed"); +}); + test("look-alikes in the project, the app's other files and other sockets stay allowed", () => { for (const command of ALLOW) assert.equal(rule(command), null, command); }); +test("a broad project (HOME, or a folder holding the app's data) cannot reopen the private data below it", () => { + const secrets = join(userData, "provider-secrets.bin"); + const otherAccount = join(userData, "account-homes", "acct-2", "auth.json"); + for (const root of [home, join(home, "Library"), userData]) { + const broad = (command, extra = {}) => denyRule(analyzeAction(shell(command), root, { home, privateData, ...extra })); + const tool = (toolName, toolInput, extra = {}) => checkBaseProtection({ toolName, toolInput, root, home, privateData, ...extra })?.rule ?? null; + for (const command of [ + `cat ${q(secrets)}`, + "cat ~/Library/Application\\ Support/canvastty/provider-secrets.bin", + `cat ${q(token)}`, + `cat ${q(join(userData, "github-oauth.json"))}`, + `cat ${q(otherAccount)}`, + `ls ${q(join(userData, "plugin-secrets"))}`, + `cat ${q(join(userData, "lifecycle", "runtime", "connection.json"))}`, + `cd ${q(userData)} && cat provider-secrets.bin`, + `python3 -c "print(open('${secrets}','rb').read())"`, + `echo x > ${q(secrets)}`, + `cp ./x ${q(token)}`, + `grep -r token ${q(userData)}` + ]) assert.equal(broad(command), "app-private", `${root}: ${command}`); + assert.equal(tool("Write", { file_path: secrets, content: "x" }), "app-private", `${root}: Write`); + assert.equal(tool("Edit", { file_path: join(userData, "github-oauth.json"), old_string: "a", new_string: "b" }), "app-private", `${root}: Edit`); + assert.equal(tool("apply_patch", { input: `*** Begin Patch\n*** Update File: ${token}\n*** End Patch` }), "app-private", `${root}: apply_patch`); + assert.equal(tool("Bash", { command: `cat ${q(secrets)}` }), "app-private", `${root}: Bash`); + // The app's other files, and the project's own ordinary files, stay usable. + assert.equal(broad(`cat ${q(join(userData, "settings.json"))}`), null, root); + assert.equal(broad(`ls ${q(userData)}`), null, root); + // The agent's own account home (strictly inside account-homes) stays its own; its neighbours do not. + const accountHome = join(userData, "account-homes", "acct-1", ".claude"); + assert.equal(broad(`cat ${q(join(accountHome, "projects", "p", "memory", "MEMORY.md"))}`, { agentRoots: [accountHome] }), null, root); + assert.equal(tool("Write", { file_path: join(accountHome, "plans", "p.md"), content: "x" }, { agentRoots: [accountHome] }), null, root); + assert.equal(broad(`cat ${q(otherAccount)}`, { agentRoots: [accountHome] }), "app-private", root); + assert.equal(broad(`cat ${q(secrets)}`, { agentRoots: [accountHome] }), "app-private", root); + } + const atHome = (command) => denyRule(analyzeAction(shell(command), home, { home, privateData })); + assert.equal(atHome(`cat ${q(join(home, "notes.txt"))}`), null); + assert.equal(atHome("grep -rn todo ."), null, "a recursive search of the project itself is not a read of the app's folder"); +}); + +test("a broad agent config folder does not reopen the private data below it either", () => { + assert.equal(rule(`cat ${q(token)}`, { agentRoots: [home] }), "app-private"); + assert.equal(rule(`cat ${q(join(home, "notes.txt"))}`, { agentRoots: [home] }), null); +}); + +test("a project in a temporary folder does not make the gateways' socket folders its own", () => { + const temp = realpathSync(tmpdir()); + const atTemp = (command) => denyRule(analyzeAction(shell(command), temp, { home, privateData })); + assert.equal(atTemp(`nc -U ${q(join(temp, "ctty-control-Ab12Cd", "c.sock"))}`), "app-private"); + assert.equal(atTemp(`cat ${q(join(temp, "ctty-runtime-501-ab12", "connection.json"))}`), "app-private"); + assert.equal(atTemp(`cat ${q(join(temp, "notes.txt"))}`), null); +}); + +test("a project inside the app's private data keeps its own files, not its neighbours", () => { + const accountProject = join(userData, "account-homes", "acct-1"); + const inside = (command) => denyRule(analyzeAction(shell(command), accountProject, { home, privateData })); + assert.equal(inside("cat .claude/settings.json"), null); + assert.equal(inside(`cat ${q(join(userData, "account-homes", "acct-2", "auth.json"))}`), "app-private"); + assert.equal(inside(`cat ${q(join(userData, "provider-secrets.bin"))}`), "app-private"); +}); + +test("plugin data is private: another account's home is refused; the card's own selected home and a project placed there stay its own", () => { + const pluginData = join(userData, "plugin-data"); + const own = join(pluginData, "com.example.accounts", "homes", "acct-1"); + const other = join(pluginData, "com.example.accounts", "homes", "acct-2"); + const check = (toolName, toolInput, extra = {}) => checkBaseProtection({ toolName, toolInput, root: project, home, privateData, ...extra })?.rule ?? null; + const ownRoots = { agentRoots: [join(home, ".claude"), own] }; + for (const extra of [{}, ownRoots]) { + assert.equal(check("Bash", { command: `cat ${q(join(other, "auth.json"))}` }, extra), "app-private"); + assert.equal(check("Read", { file_path: join(other, "auth.json") }, extra), "app-private"); + assert.equal(check("Bash", { command: `cat ${q(join(pluginData, "com.example.accounts", "state.json"))}` }, extra), "app-private"); + assert.equal(check("LS", { path: pluginData }, extra), "app-private"); + assert.equal(check("Grep", { pattern: "token", path: join(pluginData, "com.example.accounts") }, extra), "app-private"); + assert.equal(check("Bash", { command: "cat ~/Library/Application\\ Support/canvastty/plugin-data/*/homes/*/auth.json" }, extra), "app-private"); + assert.equal(check("Bash", { command: `cat ${q(join(pluginData, "com.example.other", "data.json"))}` }, extra), "app-private"); + } + // Without the launch's record the selected home is plugin data like any other. + assert.equal(check("Read", { file_path: join(own, "auth.json") }), "app-private"); + assert.equal(check("Read", { file_path: join(own, "auth.json") }, ownRoots), null); + assert.equal(check("Bash", { command: `cat ${q(join(own, "auth.json"))} && echo x > ${q(join(own, "plans", "p.md"))}` }, ownRoots), null); + assert.equal(check("Grep", { pattern: "x", path: own }, ownRoots), null); + // A whole plugin's data folder is never one card's own. + assert.equal(check("Read", { file_path: join(other, "auth.json") }, { agentRoots: [pluginData] }), "app-private"); + // A project CanvasTTY placed inside plugin data (an environment's worktree) keeps its own files, code naming them too. + const worktree = join(pluginData, "com.example.environments", "worktrees", "w1"); + const inWorktree = (command) => denyRule(analyzeAction(shell(command), worktree, { home, privateData })); + assert.equal(inWorktree("cat src/a.ts && git status"), null); + assert.equal(inWorktree(`python3 -c "open(${JSON.stringify(join(worktree, "src", "a.ts"))}).read()"`), null); + assert.equal(inWorktree(`cat ${q(join(pluginData, "com.example.environments", "worktrees", "w2", "a.ts"))}`), "app-private"); + // A normal project is unaffected. + for (const command of ["cat src/a.ts", "grep -r todo .", `ls ${q(userData)}`, `cat ${q(join(project, "plugin-data", "x.json"))}`]) { + assert.equal(check("Bash", { command }), null, command); + } + assert.equal(check("Read", { file_path: join(project, "src", "a.ts") }), null); +}); + test("file tools that name private data are refused; the agent's own account home is its own", () => { const check = (toolName, toolInput, extra = {}) => checkBaseProtection({ toolName, toolInput, root: project, home, privateData, ...extra }); assert.equal(check("Write", { file_path: join(control, "token-x"), content: "x" })?.rule, "app-private"); @@ -116,6 +220,60 @@ test("file tools that name private data are refused; the agent's own account hom assert.equal(check("Bash", { command: `cat ${q(token)}` }, { agentRoots: [accountHome] })?.rule, "app-private"); }); +test("read-only file tools are refused only for CanvasTTY's private data", () => { + const secrets = join(userData, "provider-secrets.bin"); + const accountHome = join(userData, "account-homes", "acct-1", ".claude"); + for (const root of [project, home]) { + const read = (toolName, toolInput, extra = {}) => checkBaseProtection({ toolName, toolInput, root, home, privateData, ...extra })?.rule ?? null; + for (const [toolName, toolInput, extra] of [ + ["Read", { file_path: secrets }], + ["Read", { file_path: "~/Library/Application Support/canvastty/agent-control/token-0123abcd" }], + ["NotebookRead", { notebook_path: join(userData, "github-oauth.json") }], + ["Grep", { pattern: "token", path: userData }], + ["Grep", { pattern: "x", path: join(userData, "plugin-secrets") }], + ["Glob", { pattern: "*", path: control }], + ["Glob", { pattern: join(userData, "provider-*") }], + ["LS", { path: join(userData, "account-homes") }], + ["read_file", { absolute_path: secrets }], + ["read_many_files", { paths: [join(userData, "plugin-secrets")] }], + ["grep_search", { pattern: "x", path: userData }], + ["list_directory", { path: control }], + ["Read", { file_path: join(userData, "account-homes", "acct-2", "auth.json") }, { agentRoots: [accountHome] }] + ]) assert.equal(read(toolName, toolInput, extra), "app-private", `${root}: ${toolName} ${JSON.stringify(toolInput)}`); + // Everything else stays as it was: reads outside the project are not "outside", and nothing new is refused. + for (const [toolName, toolInput] of [ + ["Read", { file_path: join(project, "src", "a.ts") }], + ["Read", { file_path: "src/a.ts" }], + ["Read", { file_path: "/etc/hosts" }], + ["Read", { file_path: join(userData, "settings.json") }], + ["Grep", { pattern: "todo", path: "src" }], + ["Grep", { pattern: "todo" }], + ["Grep", { pattern: "todo", path: home }], + ["Glob", { pattern: "**/*.ts" }], + ["LS", { path: userData }], + ["read_file", { absolute_path: join(project, "src", "a.ts") }] + ]) assert.equal(read(toolName, toolInput), null, `${root}: ${toolName} ${JSON.stringify(toolInput)}`); + // The agent's own account home stays its own. + assert.equal(read("Read", { file_path: join(accountHome, "settings.json") }, { agentRoots: [accountHome] }), null, root); + assert.equal(read("Grep", { pattern: "x", path: accountHome }, { agentRoots: [accountHome] }), null, root); + } +}); + +test("decision hooks refuse a private read without asking decision plugins about reads", async () => { + let asked = 0; + const hooks = new DecisionHooks({ + baseProtection: () => true, services: () => [{ pluginId: "p", pluginName: "P", serviceId: "s", mayAllow: false }], + call: async () => { asked++; return { behavior: "deny", message: "plugin" }; }, home, privateData, + session: () => ({ provider: "claude", role: "agent", cwd: home, configDirs: [] }) + }); + const request = (toolInput) => ({ toolName: "Read", toolInput, toolInputPreview: null, cwd: null, truncated: false }); + const denied = await hooks.decide("s1", request({ file_path: join(userData, "provider-secrets.bin") }), new AbortController().signal); + assert.equal(denied.behavior, "deny"); + assert.match(denied.message, /Orchestrator role/u); + assert.equal((await hooks.decide("s1", request({ file_path: join(home, "notes.txt") }), new AbortController().signal)).behavior, "none"); + assert.equal(asked, 0, "plugins are not asked about read-only tools"); +}); + test("without the app's folder the socket folders are still known; the userData paths are not guessed", () => { assert.equal(denyRule(analyzeAction(shell(`nc -U ${sock}`), project, { home })), "app-private"); assert.equal(denyRule(analyzeAction(shell(`cat ${q(token)}`), project, { home })), null); @@ -141,3 +299,128 @@ test("decision hooks pass the app's private data to base protection", async () = const ordinary = await hooks.decide("s1", { toolName: "bash", toolInput: { command: "cat src/a.ts" }, toolInputPreview: null, cwd: null, truncated: false }, new AbortController().signal); assert.equal(ordinary.behavior, "none"); }); + +test("Qwen read_many_files: more than 32 paths is refused, and its include globs are checked", () => { + const secrets = join(userData, "provider-secrets.bin"); + const read = (toolInput) => checkBaseProtection({ toolName: "read_many_files", toolInput, root: project, home, privateData }) ?? null; + const benign = Array.from({ length: 32 }, (_, i) => `src/f${i}.ts`); + assert.equal(read({ paths: benign }), null, "32 project paths are checked and allowed"); + const many = read({ paths: [...benign, secrets] }); + assert.equal(many?.rule, "uncheckable", "the 33rd path is not skipped"); + assert.match(many.message, /at most 32/u); + assert.equal(read({ paths: [...benign, "src/g.ts"] })?.rule, "uncheckable"); + assert.equal(read({ paths: benign.slice(0, 20), include: benign.slice(0, 13) })?.rule, "uncheckable", "include counts toward the limit"); + assert.equal(read({ paths: ["src/a.ts"], include: [secrets] })?.rule, "app-private"); + assert.equal(read({ paths: ["src/a.ts"], include: [join(userData, "account-homes", "**", "*.json")] })?.rule, "app-private"); + assert.equal(read({ paths: ["src/a.ts"], include: ["**/*.ts"] }), null); +}); + +test("OpenCode's read-only tools reach base protection: a grep without a path searches the current folder, a glob's pattern is checked", () => { + const check = (tool, args, root) => { + const call = guardedCall(tool, args, true); + assert.ok(call, `${tool} is forwarded`); + return checkBaseProtection({ toolName: call.toolName, toolInput: call.toolInput, root, home, privateData })?.rule ?? null; + }; + assert.equal(check("grep", { pattern: "token" }, userData), "app-private", "grep with no path, run in the app's folder"); + assert.equal(check("grep", { pattern: "token", include: "*.json" }, control), "app-private"); + assert.equal(check("grep", { pattern: "token" }, project), null); + assert.equal(check("glob", { pattern: join(userData, "provider-*") }, project), "app-private"); + assert.equal(check("glob", { pattern: "agent-control/token-*", path: userData }, project), "app-private"); + assert.equal(check("glob", { pattern: "**/*.ts" }, project), null); +}); + +test("globs: ** matches any number of folders and braces are expanded before matching the private data", () => { + const secrets = q(join(userData, "provider-secrets.bin")).slice(1, -1); + for (const command of [ + "cat ~/Library/**/provider-secrets.bin", + "cat ~/**/provider-secrets.bin", + "cat ~/Library/Application\\ Support/canvastty/**", + "cat ~/Library/**/account-homes/*/auth.json", + `cat '${userData}'/provider-secrets.{bin,}`, + `cat '${userData}'/{settings.json,provider-secrets.bin}`, + `cat '${userData}'/agent-control/{connection.json,token-*}`, + `cp '${userData}'/{provider-secrets,x}.bin ./copy` + ]) assert.equal(rule(command), "app-private", command); + // Read-only tools: a Glob pattern with braces or ** is matched the same way. + const glob = (toolInput, root = project) => checkBaseProtection({ toolName: "Glob", toolInput, root, home, privateData })?.rule ?? null; + assert.equal(glob({ pattern: "provider-secrets.{bin,json}", path: userData }), "app-private"); + assert.equal(glob({ pattern: join(home, "Library", "**", "provider-secrets.bin") }), "app-private"); + // Ordinary globs stay allowed: in a normal project, and ones that only pass by the app's folder on the way. + for (const command of [ + "ls **/*.ts", + "rg todo **/*.ts", + "cat src/{a,b}.ts", + "cat ~/**/*.ts", + "ls ~/Library/**/*.plist", + `cat '${secrets}x'.{md,txt}`, + "echo {a,b}" + ]) assert.equal(rule(command), null, command); + assert.equal(glob({ pattern: "**/*.ts" }), null); + assert.equal(glob({ pattern: "**/*.ts" }, home), null, "a ** glob that only passes through the app's folder"); + assert.equal(glob({ pattern: "src/**/*.{ts,tsx}" }), null); +}); + +test("macOS and Windows: a glob matches the private data in any letter case", { skip: process.platform !== "darwin" && process.platform !== "win32" }, () => { + assert.equal(rule("cat ~/library/application\\ support/CANVASTTY/Provider-Secrets.*"), "app-private"); + assert.equal(rule("cat ~/LIBRARY/**/PROVIDER-SECRETS.BIN"), "app-private"); +}); + +test("what base protection cannot check is refused, not skipped: long or odd path fields, too many paths, expansions or nested shells", () => { + const secrets = join(userData, "provider-secrets.bin"); + const check = (toolName, toolInput, root = project) => checkBaseProtection({ toolName, toolInput, root, home, privateData })?.rule ?? null; + // Padding that folds away is checked as the path it names. + assert.equal(check("Read", { file_path: userData + "/.".repeat(3000) + "/provider-secrets.bin" }), "app-private"); + assert.equal(check("Read", { file_path: userData + "/x/..".repeat(1500) + "/provider-secrets.bin" }), "app-private"); + assert.equal(check("Write", { file_path: userData + "/./".repeat(2000) + "provider-secrets.bin", content: "x" }), "app-private"); + // A path field that cannot be checked: too long after folding, or not a string. + const long = "/" + "a".repeat(5000); + for (const [toolName, toolInput] of [ + ["Read", { file_path: long }], + ["Read", { file_path: 42 }], + ["Grep", { pattern: "x", path: long }], + ["Glob", { pattern: { glob: "*" } }], + ["read_many_files", { paths: ["src/a.ts", 7] }], + ["read_many_files", { paths: ["src/a.ts"], include: long }], + ["Write", { file_path: long, content: "x" }], + ["Bash", { command: "ls", cwd: long }], + ["apply_patch", { command: `*** Begin Patch\n*** Add File: ${"b".repeat(5000)}\n+x\n*** End Patch` }], + ["apply_patch", { command: `*** Begin Patch\n${Array.from({ length: 257 }, (_, i) => `*** Add File: f${i}.txt\n+x`).join("\n")}\n*** End Patch` }] + ]) assert.equal(check(toolName, toolInput), "uncheckable", `${toolName} ${JSON.stringify(toolInput).slice(0, 80)}`); + // A patch longer than one path is still read in full. + assert.equal(check("apply_patch", { command: `*** Begin Patch\n*** Add File: notes.md\n+${"x".repeat(5000)}\n*** Add File: ${secrets}\n+x\n*** End Patch` }), "app-private"); + // More paths, brace expansions or nested shells than are checked. + assert.equal(rule(`echo ${"a ".repeat(1100)}; cat ${q(secrets)}`) !== null, true, "the private path after the budget is not let through"); + assert.equal(rule(`echo ${"a ".repeat(1100)}`), "uncheckable"); + assert.equal(rule(`cat ${q(userData)}/{a,b}{c,d}{e,f}{g,h}{i,j}{k,l}{m,n}`), "uncheckable"); + assert.equal(check("Glob", { pattern: "{a,b}{c,d}{e,f}{g,h}{i,j}{k,l}{m,n}", path: project }), "uncheckable"); + const sq = (text) => `'${text.replace(/'/gu, "'\\''")}'`; + let nested = "ls"; + for (let i = 0; i < 6; i++) nested = `bash -c ${sq(nested)}`; + assert.equal(rule(nested), "uncheckable", "shells nested deeper than are followed"); + assert.equal(rule("bash -c 'bash -c ls'"), null); + // Ordinary large calls stay allowed. + assert.equal(rule(`git add ${Array.from({ length: 300 }, (_, i) => `src/f${i}.ts`).join(" ")}`), null); + assert.equal(rule("cat src/{a,b,c}/{x,y}.ts"), null); + assert.equal(check("apply_patch", { command: `*** Begin Patch\n${Array.from({ length: 40 }, (_, i) => `*** Add File: f${i}.txt\n+x`).join("\n")}\n*** End Patch` }), null); +}); + +test("an analyzer failure refuses a shell, edit or read call as uncheckable; other tools keep no opinion", () => { + const throwing = (field) => Object.defineProperty({}, field, { enumerable: true, get() { throw new Error("fixture analyzer failure"); } }); + const brokenPrivateData = { appRoots: [userData], get paths() { throw new Error("fixture analyzer failure"); }, markers: [] }; + const check = (toolName, toolInput, data = privateData) => checkBaseProtection({ toolName, toolInput, root: project, home, privateData: data }); + for (const [toolName, toolInput] of [ + ["Bash", throwing("command")], + ["Read", throwing("file_path")], + ["Write", throwing("file_path")], + ["apply_patch", throwing("command")] + ]) { + const verdict = check(toolName, toolInput); + assert.equal(verdict?.rule, "uncheckable", toolName); + assert.match(verdict.message, /failed while checking it/u); + assert.doesNotMatch(verdict.message, /fixture analyzer failure/u, "the error itself stays out of the message"); + } + for (const [toolName, toolInput] of [["Bash", { command: "ls" }], ["Read", { file_path: "src/a.ts" }], ["Edit", { file_path: "src/a.ts" }]]) { + assert.equal(check(toolName, toolInput, brokenPrivateData)?.rule, "uncheckable", `${toolName}: a failure inside the analysis`); + } + assert.equal(check("WebFetch", throwing("url")), null, "a tool base protection does not check stays without an opinion"); +}); diff --git a/tests/base-protection.test.mjs b/tests/base-protection.test.mjs index 860392347..91ac9adb2 100644 --- a/tests/base-protection.test.mjs +++ b/tests/base-protection.test.mjs @@ -80,7 +80,7 @@ test("ordinary work inside the working folder has no deny", () => { assert.equal(rule(edit("src/new.ts")), null); }); -test("hook tool calls: Claude, Codex, Qwen and OpenCode names; only shells and file writes are checked", () => { +test("hook tool calls: Claude, Codex, Qwen and OpenCode names; shells and file writes are checked, reads only for private data", () => { const cases = [ ["Write", { file_path: join(home, "Downloads", "hello.txt"), content: "hello" }, "write-outside"], ["Edit", { file_path: "/etc/hosts", old_string: "a", new_string: "b" }, "write-outside"], @@ -99,7 +99,9 @@ test("hook tool calls: Claude, Codex, Qwen and OpenCode names; only shells and f assert.equal(check(tool, input), null, `${tool} ${JSON.stringify(input)}`); } assert.deepEqual(patchPaths("*** Begin Patch\n*** Add File: a.txt\n*** Delete File: b.txt\n*** End Patch"), ["a.txt", "b.txt"]); - assert.deepEqual(actionFromHook("Grep", { pattern: "x" }).kind, null); + assert.deepEqual(actionFromHook("Grep", { pattern: "x" }), { kind: "read", command: null, commandCwd: null, paths: [], searched: [] }); + assert.deepEqual(actionFromHook("Read", { file_path: "/etc/passwd" }).paths, ["/etc/passwd"]); + assert.deepEqual(actionFromHook("WebFetch", { url: "https://example.com" }).kind, null); }); test("each deny tells the model what to do instead; a write only to the temporary folder suggests a folder in the project", () => { diff --git a/tests/claude-http-hooks.test.mjs b/tests/claude-http-hooks.test.mjs index fb649ea87..94791bd48 100644 --- a/tests/claude-http-hooks.test.mjs +++ b/tests/claude-http-hooks.test.mjs @@ -262,7 +262,7 @@ test("Claude's lifecycle hooks become HTTP hooks except SessionStart; the decisi assert.equal(settings.hooks.PreToolUse[0].hooks[0].type, "command"); assert.match(settings.hooks.PreToolUse[0].hooks[0].command, /permission-gate\.mjs' 'pretool'/u); for (const [event, state] of [["UserPromptSubmit", "working"], ["PermissionRequest", "needs_approval"], ["PostToolUse", "working"], - ["Stop", "idle"], ["StopFailure", "idle"], ["SessionEnd", "idle"], ["Notification", "needs_approval"]]) { + ["PostToolUseFailure", "working"], ["Stop", "idle"], ["StopFailure", "idle"], ["SessionEnd", "idle"], ["Notification", "needs_approval"]]) { const hook = settings.hooks[event][0].hooks[0]; assert.deepEqual(hook, { type: "http", diff --git a/tests/decision-hooks.test.mjs b/tests/decision-hooks.test.mjs index b7a4a2308..27af58f58 100644 --- a/tests/decision-hooks.test.mjs +++ b/tests/decision-hooks.test.mjs @@ -11,7 +11,7 @@ import { tmpdir } from "node:os"; import { join } from "node:path"; import test from "node:test"; import { fileURLToPath } from "node:url"; -import { AGENT_RUNTIME_ENV, OPENCODE_DECISIONS_ENV, RUNTIME_PROTOCOL_VERSION } from "../src/agent-runtime/runtime-protocol.mjs"; +import { AGENT_RUNTIME_ENV, OPENCODE_DECISIONS_ENV, OPENCODE_DECISION_READS_ENV, RUNTIME_PROTOCOL_VERSION } from "../src/agent-runtime/runtime-protocol.mjs"; import { buildRequest, hookOutput, parseDecision } from "../src/agent-runtime/permission-gate.mjs"; import { createOpenCodeDecisions, guardedCall } from "../src/agent-runtime/opencode-decisions.mjs"; import { DecisionHooks, TOO_LARGE_MESSAGE, mergeDecisions } from "../src/main/services/DecisionHooks.ts"; @@ -129,8 +129,10 @@ test("cut input under base protection: a shell call, or a file write whose targe // A file write whose path is at the start of the preview is judged by that path, as before. const inside = JSON.stringify({ file_path: join(project, "big.txt"), content: "x".repeat(300) }).slice(0, 200); assert.equal((await decisions.decide("s1", cut("Write", inside), live())).behavior, "none"); + // A cut read-only call may hide the file it reads (a long read_many_files list): denied too. + assert.deepEqual(await decisions.decide("s1", cut("read_many_files", "{\"paths\":[\"src/a.ts\""), live()), { behavior: "deny", message: TOO_LARGE_MESSAGE }); // Tools the rules do not read, and base protection off, are unchanged. - assert.equal((await decisions.decide("s1", cut("Read", "{\"file_path\":"), live())).behavior, "none"); + assert.equal((await decisions.decide("s1", cut("WebFetch", "{\"url\":"), live())).behavior, "none"); assert.equal((await hooks({ protect: false }).decide("s1", cut("Bash", "{\"command\":\"ls"), live())).behavior, "none"); }); @@ -302,6 +304,9 @@ test("launch: the decision hook is added only when wanted, per provider, with or const claude = JSON.parse(adapters.prepare("claude", "t1", false, true).args[1]); assert.deepEqual(Object.keys(claude.hooks), ["PreToolUse"]); assert.equal(claude.hooks.PreToolUse[0].matcher, "Bash|Write|Edit|MultiEdit|NotebookEdit"); + // Read-only file tools only when base protection was on at launch (they are checked for its private-data rule). + const claudeReads = JSON.parse(adapters.prepare("claude", "t1", false, true, undefined, undefined, true).args[1]); + assert.equal(claudeReads.hooks.PreToolUse[0].matcher, "Bash|Write|Edit|MultiEdit|NotebookEdit|Read|NotebookRead|Grep|Glob|LS"); assert.match(claude.hooks.PreToolUse[0].hooks[0].command, /permission-gate\.mjs['"] ['"]pretool['"]$/u); assert.ok(JSON.parse(adapters.prepare("claude", "t1", true, true).args[1]).hooks.Stop, "status hooks stay alongside"); const codex = adapters.prepare("codex", "t2", false, true).args.join(" "); @@ -309,8 +314,15 @@ test("launch: the decision hook is added only when wanted, per provider, with or const qwen = adapters.prepare("qwen", "t3", false, true); const qwenSettings = JSON.parse(await readFile(qwen.environment.QWEN_CODE_SYSTEM_SETTINGS_PATH, "utf8")); assert.equal(qwenSettings.hooks.PreToolUse[0].hooks[0].timeout, 15_000); + assert.equal(qwenSettings.hooks.PreToolUse[0].matcher, "^(run_shell_command|write_file|edit|replace)$"); qwen.releaseConfiguration(); + const qwenReads = adapters.prepare("qwen", "t3", false, true, undefined, undefined, true); + assert.equal(JSON.parse(await readFile(qwenReads.environment.QWEN_CODE_SYSTEM_SETTINGS_PATH, "utf8")).hooks.PreToolUse[0].matcher, + "^(run_shell_command|write_file|edit|replace|read_file|read_many_files|grep_search|search_file_content|glob|list_directory)$"); + qwenReads.releaseConfiguration(); assert.equal(adapters.prepare("opencode", "t4", false, true).environment[OPENCODE_DECISIONS_ENV], "1"); + assert.equal(adapters.prepare("opencode", "t4", false, true).environment[OPENCODE_DECISION_READS_ENV], undefined); + assert.equal(adapters.prepare("opencode", "t4", false, true, undefined, undefined, true).environment[OPENCODE_DECISION_READS_ENV], "1"); assert.equal(adapters.prepare("opencode", "t4", false, false).environment[OPENCODE_DECISIONS_ENV], undefined); assert.equal(adapters.decisionsSupported("kimi"), false); assert.deepEqual(adapters.prepare("kimi", "t5", false, true).args, []); @@ -325,9 +337,10 @@ test("launch: the decision hook is added only when wanted, per provider, with or }; const bridge = new AgentRuntimeBridge(gateway, { helper, runtimeDirectory, openCodePluginPath: "/opt/CanvasTTY/opencode-plugin.mjs", kimiHomeDirectory: join(runtimeDirectory, "kimi"), hermesHomeDirectory: join(runtimeDirectory, "hermes"), grokHomeDirectory: join(runtimeDirectory, "grok"), - permissionGate, coreHooksEnabled: true, wantsDecisions: (provider) => provider === "claude" }); + permissionGate, coreHooksEnabled: true, wantsDecisions: (provider) => provider === "claude", decisionReads: () => true }); const launched = bridge.prepareLaunch({ terminalSessionId: "a", provider: "claude", cwd: project }); assert.equal(launched.decisions, true); + assert.match(JSON.stringify(launched.args), /NotebookEdit\|Read\|NotebookRead/u, "base protection on: the hook sees reads"); assert.equal(registered[0][4], true); assert.equal(bridge.prepareLaunch({ terminalSessionId: "b", provider: "codex", cwd: project }).decisions, false); bridge.setCoreHooksEnabled(false); @@ -336,7 +349,7 @@ test("launch: the decision hook is added only when wanted, per provider, with or test("OpenCode: the guard throws a deny, remembers an allow and answers OpenCode's prompt once", async () => { const env = { - [OPENCODE_DECISIONS_ENV]: "1", [AGENT_RUNTIME_ENV.address]: "/tmp/x.sock", [AGENT_RUNTIME_ENV.terminalSessionId]: "t", + [OPENCODE_DECISIONS_ENV]: "1", [OPENCODE_DECISION_READS_ENV]: "1", [AGENT_RUNTIME_ENV.address]: "/tmp/x.sock", [AGENT_RUNTIME_ENV.terminalSessionId]: "t", [AGENT_RUNTIME_ENV.provider]: "opencode", [AGENT_RUNTIME_ENV.capabilityToken]: "c".repeat(40) }; const replies = []; @@ -354,10 +367,25 @@ test("OpenCode: the guard throws a deny, remembers an allow and answers OpenCode next = { behavior: "ask", message: "" }; await decisions.guard({ tool: "bash", callID: "c3" }, { args: { command: "ls" } }); assert.equal(await decisions.permissionAsked({ id: "r3", tool: { callID: "c3" } }), false, "ask leaves OpenCode's own prompt"); - await decisions.guard({ tool: "read", callID: "c4" }, { args: { filePath: "/etc/hosts" } }); - assert.equal(sent.length, 3, "tools that neither run commands nor write files are not sent"); + await decisions.guard({ tool: "webfetch", callID: "c4" }, { args: { url: "https://example.com" } }); + assert.equal(sent.length, 3, "tools that neither run commands, write files nor read them are not sent"); + await decisions.guard({ tool: "read", callID: "c5" }, { args: { filePath: "/etc/hosts" } }); + assert.equal(sent.length, 4, "a read-only file tool is sent for base protection's private-data rule"); + assert.equal(sent[3].toolName, "Read"); + // Base protection off at launch: reads are not sent at all. + const noReads = []; + await createOpenCodeDecisions({ client, env: { ...env, [OPENCODE_DECISION_READS_ENV]: undefined }, send: async (_address, message) => { noReads.push(message); return null; } }) + .guard({ tool: "read", callID: "c6" }, { args: { filePath: "/etc/hosts" } }); + assert.equal(noReads.length, 0); assert.equal(createOpenCodeDecisions({ env: { ...env, [OPENCODE_DECISIONS_ENV]: "0" } }).enabled, false); assert.deepEqual(guardedCall("apply_patch", { patchText: "*** Begin Patch" }), { toolName: "apply_patch", toolInput: { patch: "*** Begin Patch" } }); + assert.deepEqual(guardedCall("read", { filePath: "/p/a.ts" }, true), { toolName: "Read", toolInput: { file_path: "/p/a.ts" } }); + assert.deepEqual(guardedCall("grep", { pattern: "x", path: "/p" }, true), { toolName: "Grep", toolInput: { path: "/p" } }); + assert.deepEqual(guardedCall("glob", { pattern: "**/*.ts" }, true), { toolName: "Glob", toolInput: { pattern: "**/*.ts" } }); + assert.deepEqual(guardedCall("glob", { pattern: "*.md", path: "/p" }, true), { toolName: "Glob", toolInput: { path: "/p", pattern: "*.md" } }); + assert.deepEqual(guardedCall("list", { path: "/p" }, true), { toolName: "LS", toolInput: { path: "/p" } }); + assert.deepEqual(guardedCall("grep", { pattern: "x", include: "*.ts" }, true), { toolName: "Grep", toolInput: { path: "." } }, "a grep without a path searches the current folder"); + assert.equal(guardedCall("read", { filePath: "/p/a.ts" }), null, "reads are mapped only when the launch checks them"); // multiedit: every edit's new text reaches the decision, as Claude's MultiEdit tool_input does. assert.deepEqual(guardedCall("multiedit", { filePath: "/p/a.ts", edits: [{ oldString: "a", newString: "TOKEN=1" }, { oldString: "b", newString: "rm -rf /" }, { newString: 5 }] }), { toolName: "edit", toolInput: { file_path: "/p/a.ts", content: "TOKEN=1\nrm -rf /" } }); diff --git a/tests/file-access-service.test.mjs b/tests/file-access-service.test.mjs index 08f30967f..8f95583c4 100644 --- a/tests/file-access-service.test.mjs +++ b/tests/file-access-service.test.mjs @@ -3,7 +3,7 @@ import fs, { realpathSync } from "node:fs"; import { mkdir, mkdtemp, rm, symlink, writeFile } from "node:fs/promises"; import { tmpdir } from "node:os"; import { basename, join } from "node:path"; -import { syncBuiltinESMExports } from "node:module"; +import { createRequire, syncBuiltinESMExports } from "node:module"; import test from "node:test"; import { FileAccessService } from "../src/main/services/FileAccessService.ts"; @@ -11,6 +11,8 @@ async function makeTempDir(prefix) { return mkdtemp(join(tmpdir(), prefix)); } +const fsPromises = createRequire(import.meta.url)("node:fs/promises"); + function createService({ cwd, limits } = {}) { return new FileAccessService({ resolveSessionCwd: (sessionId) => (sessionId === "session-1" ? cwd : undefined), @@ -138,6 +140,35 @@ test("rejects symlinks whose real target escapes the root", async () => { } }); +test("rejects a read whose descriptor is not the file the contained path names", async (t) => { + // Simulates an intermediate directory swapped between realpath and open: the open lands on a file + // outside the root while the path re-resolved afterwards is still inside it. + const root = await makeTempDir("canvastty-files-swap-root-"); + const outside = await makeTempDir("canvastty-files-swap-outside-"); + const realOpen = fsPromises.open; + try { + const service = createService({ cwd: root }); + const { rootId } = service.registerSessionRoot("session-1"); + await writeFile(join(root, "notes.txt"), "inside"); + await writeFile(join(outside, "secret.txt"), "top secret"); + const redirected = join(outside, "secret.txt"); + t.mock.method(fsPromises, "open", (path, ...rest) => + String(path).endsWith("notes.txt") ? realOpen(redirected, ...rest) : realOpen(path, ...rest) + ); + syncBuiltinESMExports(); + + assert.deepEqual(await service.read(rootId, "notes.txt"), { + kind: "unsupported", + reason: "not-permitted" + }); + } finally { + t.mock.restoreAll(); + syncBuiltinESMExports(); + await rm(root, { recursive: true, force: true }); + await rm(outside, { recursive: true, force: true }); + } +}); + test("rejects non-regular files such as directories", async () => { const root = await makeTempDir("canvastty-files-nonregular-"); try { diff --git a/tests/helpers/terminal.mjs b/tests/helpers/terminal.mjs index c36d9d628..2747eb98f 100644 --- a/tests/helpers/terminal.mjs +++ b/tests/helpers/terminal.mjs @@ -25,6 +25,18 @@ export function availableRegistry() { * returns a fake PTY: `emitData`/`emitExit` drive its listeners, `lastResize` * holds the last size, and `onWrite(data, options)` sees what was written. */ +/** + * What OpenCode prints once its prompt has focus: the alternate screen, its hidden cursor while it draws, then the + * cursor shown in the input field (TerminalManager types queued input only after that for OpenCode). + */ +export const OPENCODE_INPUT_READY = "\x1b[?1049h\x1b[?2004h\x1b[?25l Ask anything\x1b[?25h"; + +/** Shows OpenCode's input field on the fake PTY OpenCode was spawned on, once that PTY exists. */ +export async function openCodeInputReady(calls, index = calls.length) { + for (let i = 0; i < 200 && calls.length <= index; i++) await new Promise((resolve) => setTimeout(resolve, 5)); + calls[index].process.emitData(OPENCODE_INPUT_READY); +} + export function fakeSpawner(calls, { pidBase = 20_000, onWrite } = {}) { return (command, args, options) => { let dataListener = () => undefined; @@ -53,3 +65,13 @@ export function fakeSpawner(calls, { pidBase = 20_000, onWrite } = {}) { return process; }; } + +/** fakeSpawner whose PTYs show OpenCode's input field shortly after they start (input waits for it for OpenCode). */ +export function inputReadySpawner(calls, options) { + const spawner = fakeSpawner(calls, options); + return (command, args, spawnOptions) => { + const process = spawner(command, args, spawnOptions); + setTimeout(() => process.emitData(OPENCODE_INPUT_READY), 10); + return process; + }; +} diff --git a/tests/isolation-git-metadata.test.mjs b/tests/isolation-git-metadata.test.mjs new file mode 100644 index 000000000..1972ad382 --- /dev/null +++ b/tests/isolation-git-metadata.test.mjs @@ -0,0 +1,160 @@ +import assert from "node:assert/strict"; +import { spawnSync } from "node:child_process"; +import { existsSync, statSync } from "node:fs"; +import { mkdir, mkdtemp, readFile, realpath, rm, writeFile } from "node:fs/promises"; +import { tmpdir } from "node:os"; +import { join } from "node:path"; +import test from "node:test"; +import { AgentIsolation } from "../src/main/services/isolation/AgentIsolation.ts"; +import { isolationPaths } from "../src/main/services/isolation/isolationPaths.ts"; +import { seatbeltProfile } from "../src/main/services/isolation/seatbelt.ts"; +import { bubblewrapArguments } from "../src/main/services/isolation/bubblewrap.ts"; +import { LinuxHostPaths } from "../src/main/services/isolation/linuxHostPaths.ts"; + +const mac = process.platform === "darwin" && existsSync("/usr/bin/sandbox-exec"); +const onMac = { skip: mac ? false : "macOS seatbelt (sandbox-exec) only" }; +const hasGit = spawnSync("git", ["--version"]).status === 0; +const withGit = { skip: process.platform === "win32" ? "POSIX paths" : hasGit ? false : "git is not installed" }; +const GIT_IDENTITY = "-c user.email=a@b -c user.name=n"; + +async function world(t) { + const base = await realpath(await mkdtemp(join(process.platform === "win32" ? tmpdir() : "/tmp", "ctty-git-test-"))); + t.after(() => rm(base, { recursive: true, force: true })); + const home = join(base, "h"); + const userData = join(base, "u"); + const project = join(base, "p"); + const temp = join(base, "t"); + const template = join(base, "tpl"); + for (const dir of [home, userData, project, temp, template]) await mkdir(dir, { recursive: true }); + const env = { HOME: home, PATH: "/usr/bin:/bin:/usr/sbin:/sbin", LANG: "en_US.UTF-8", GIT_CONFIG_NOSYSTEM: "1" }; + return { base, home, userData, project, temp, template, env }; +} + +function git(w, ...args) { + const result = spawnSync("git", args, { cwd: w.project, env: { ...w.env, GIT_TEMPLATE_DIR: w.template }, encoding: "utf8" }); + assert.equal(result.status, 0, result.stderr); + return result.stdout; +} + +function pathsFor(w, options = {}) { + return isolationPaths({ provider: "codex", cwd: w.project, sessionTemp: join(w.temp, "s"), env: w.env, userDataPath: w.userData, sessionId: "s", ...options }); +} + +test("the paths: an existing repository's per-worktree config, commondir and worktree records are protected", withGit, async (t) => { + const w = await world(t); + const gitDir = join(w.project, ".git"); + // No repository yet: nothing to protect by path (seatbelt's pattern still covers what `git init` makes). + const fresh = pathsFor(w); + assert.ok(!fresh.protectedWrites.some((path) => path.startsWith(`${gitDir}/`))); + assert.match(seatbeltProfile(fresh), /\\\.git\/\(info\/attributes\|commondir\|config\\\.worktree\(\\\.lock\)\?\|worktrees\(\/\.\*\)\?\)\$/u); + git(w, "init", "-q"); + const paths = pathsFor(w); + for (const file of ["config", "config.worktree", "config.worktree.lock"]) assert.ok(paths.protectedWrites.includes(join(gitDir, file)), file); + assert.ok(!paths.protectedWrites.includes(join(gitDir, "commondir")), "a missing commondir gets no placeholder: git would refuse an empty one"); + assert.ok(paths.protectedDirectories.includes(join(gitDir, "worktrees"))); + await writeFile(join(gitDir, "commondir"), ".\n"); + assert.ok(pathsFor(w).protectedWrites.includes(join(gitDir, "commondir")), "an existing one is protected"); + // Objects, refs, the index, logs and HEAD stay writable. + for (const name of ["objects", "refs", "index", "logs", "HEAD", "packed-refs", "ORIG_HEAD"]) { + const path = join(gitDir, name); + assert.ok(![...paths.protectedWrites, ...paths.protectedDirectories].some((p) => path === p || path.startsWith(`${p}/`)), name); + } +}); + +test("bubblewrap: the records are mounted read-only; a missing config.worktree and worktrees folder get neutral placeholders", withGit, async (t) => { + const w = await world(t); + git(w, "init", "-q"); + const gitDir = join(w.project, ".git"); + const kind = (path) => { + try { const stat = statSync(path); return stat.isDirectory() ? "directory" : stat.isFile() ? "file" : null; } catch { return null; } + }; + const paths = pathsFor(w); + const hostPaths = new LinuxHostPaths(); + const release = hostPaths.prepare(paths); + try { + assert.equal(await readFile(join(gitDir, "config.worktree"), "utf8"), "", "an empty config.worktree means no per-worktree config"); + assert.equal(kind(join(gitDir, "worktrees")), "directory"); + assert.equal(existsSync(join(gitDir, "commondir")), false, "no commondir placeholder"); + // The person's git outside the layer keeps working with them in place. + git(w, "status", "--short"); + assert.equal(git(w, "rev-parse", "--git-common-dir").trim(), ".git"); + const args = bubblewrapArguments(paths, { command: "/usr/bin/codex", args: [], cwd: w.project }, kind).join(" "); + const project = args.indexOf(`--bind ${w.project} ${w.project}`); + const repository = args.indexOf(`--bind ${gitDir} ${gitDir}`); + assert.ok(repository > project, ".git is a mount point of its own: it cannot be renamed away and back"); + for (const path of [join(gitDir, "config"), join(gitDir, "config.worktree"), join(gitDir, "worktrees")]) { + assert.ok(args.indexOf(`--ro-bind ${path} ${path}`) > repository, path); + } + } finally { + release(); + } + assert.equal(existsSync(join(gitDir, "config.worktree")), false, "placeholders removed"); + assert.equal(existsSync(join(gitDir, "worktrees")), false); + // An existing commondir is mounted read-only. + await writeFile(join(gitDir, "commondir"), ".\n"); + const placeholders = new Map([[join(gitDir, "config.worktree"), "file"], [join(gitDir, "worktrees"), "directory"]]); + const withCommon = bubblewrapArguments(pathsFor(w), { command: "/usr/bin/codex", args: [], cwd: w.project }, + (path) => kind(path) ?? placeholders.get(path) ?? null).join(" "); + assert.ok(withCommon.includes(`--ro-bind ${join(gitDir, "commondir")} ${join(gitDir, "commondir")}`)); +}); + +function run(w, script) { + const wrapped = new AgentIsolation({ userDataPath: w.userData, enabled: () => true, tempRoot: w.temp }) + .wrap({ sessionId: "s1", provider: "codex", cwd: w.project, command: "/bin/sh", args: ["-c", script], env: w.env }); + try { + const result = spawnSync(wrapped.command, wrapped.args, { cwd: w.project, env: wrapped.env, encoding: "utf8", timeout: 30_000 }); + return result.stdout.split("\n").filter(Boolean); + } finally { + wrapped.cleanup(); + } +} + +const WORKFLOW = [ + `git init -q . && echo a > a.txt && git add a.txt && git ${GIT_IDENTITY} commit -q -m one && echo GIT-COMMIT-ok`, + `git checkout -q -b x && echo b > b.txt && git add b.txt && git ${GIT_IDENTITY} commit -q -m two && git checkout -q - && git branch -q y && echo GIT-BRANCH-ok`, + `echo c >> a.txt && git ${GIT_IDENTITY} stash -q && git ${GIT_IDENTITY} stash pop -q && git ${GIT_IDENTITY} commit -q -am three && echo GIT-STASH-ok` +]; +const PROTECTED = [ + 'echo .. > .git/commondir 2>/dev/null || echo W-commondir-denied', + '(printf "[core]\\n\\thooksPath = /tmp\\n" > .git/config.worktree) 2>/dev/null || echo W-config-worktree-denied', + 'mkdir .git/worktrees 2>/dev/null || echo W-worktrees-denied', + 'git worktree add -q wt 2>/dev/null || echo GIT-WORKTREE-ADD-denied', + `git init -q deep/n && (echo .. > deep/n/.git/commondir) 2>/dev/null || echo W-nested-commondir-denied`, + 'mv .git moved 2>/dev/null || echo MV-git-denied', + 'mkdir -p .git/hooks .git/info && mv .git/hooks .git/h2 2>/dev/null || echo MV-hooks-denied', + 'mv .git/info .git/i2 2>/dev/null || echo MV-info-denied', + 'mv deep/n/.git deep/n/moved 2>/dev/null || echo MV-nested-git-denied' +]; + +test("seatbelt, for real: init, add, commit, checkout, branch and stash work; git's pointers and worktree records cannot be created", onMac, async (t) => { + const w = await world(t); + const lines = run(w, [...WORKFLOW, ...PROTECTED, 'git log --oneline | wc -l | tr -d " "'].join("; ")); + assert.deepEqual(lines, ["GIT-COMMIT-ok", "GIT-BRANCH-ok", "GIT-STASH-ok", "W-commondir-denied", "W-config-worktree-denied", "W-worktrees-denied", + "GIT-WORKTREE-ADD-denied", "W-nested-commondir-denied", "MV-git-denied", "MV-hooks-denied", "MV-info-denied", "MV-nested-git-denied", "2"]); + for (const path of [".git/commondir", ".git/config.worktree", ".git/worktrees", "deep/n/.git/commondir", "moved", ".git/h2", ".git/i2"]) { + assert.equal(existsSync(join(w.project, path)), false, path); + } +}); + +test("seatbelt, for real: an existing repository's records stay as they are; ordinary work continues", onMac, async (t) => { + const w = await world(t); + git(w, "init", "-q"); + await writeFile(join(w.project, "a.txt"), "a\n"); + git(w, "add", "a.txt"); + git(w, "-c", "user.email=a@b", "-c", "user.name=n", "commit", "-q", "-m", "one"); + git(w, "worktree", "add", "-q", join(w.base, "linked")); + await writeFile(join(w.project, ".git", "config.worktree"), "[user]\n\tname = kept\n"); + const record = join(w.project, ".git", "worktrees", "linked"); + const before = await readFile(join(record, "gitdir"), "utf8"); + const lines = run(w, [ + `echo b > b.txt && git add b.txt && git ${GIT_IDENTITY} commit -q -m two && git checkout -q -b x && echo GIT-ok`, + 'echo /tmp/x > .git/worktrees/linked/gitdir 2>/dev/null || echo W-record-denied', + 'echo .. > .git/worktrees/linked/commondir 2>/dev/null || echo W-record-commondir-denied', + 'rm -rf .git/worktrees 2>/dev/null; [ -d .git/worktrees/linked ] && echo RM-worktrees-denied', + 'echo x >> .git/config.worktree 2>/dev/null || echo W-config-worktree-denied' + ].join("; ")); + assert.deepEqual(lines, ["GIT-ok", "W-record-denied", "W-record-commondir-denied", "RM-worktrees-denied", "W-config-worktree-denied"]); + assert.equal(await readFile(join(record, "gitdir"), "utf8"), before); + assert.equal(await readFile(join(w.project, ".git", "config.worktree"), "utf8"), "[user]\n\tname = kept\n"); + assert.equal(git(w, "rev-parse", "--abbrev-ref", "HEAD").trim(), "x"); +}); diff --git a/tests/isolation-hardening.test.mjs b/tests/isolation-hardening.test.mjs new file mode 100644 index 000000000..06ca0aaac --- /dev/null +++ b/tests/isolation-hardening.test.mjs @@ -0,0 +1,333 @@ +import assert from "node:assert/strict"; +import { spawnSync } from "node:child_process"; +import { existsSync } from "node:fs"; +import { mkdir, mkdtemp, readFile, realpath, rm, writeFile } from "node:fs/promises"; +import { createServer } from "node:net"; +import { tmpdir } from "node:os"; +import { join } from "node:path"; +import test from "node:test"; +import { AgentControlService, subagentFolder } from "../src/main/services/AgentControlService.ts"; +import { EnvironmentRegistry } from "../src/main/services/EnvironmentRegistry.ts"; +import { AgentIsolation, SANDBOX_EXEC } from "../src/main/services/isolation/AgentIsolation.ts"; +import { contributedAccountHome, validateAccountHome } from "../src/main/services/isolation/accountHomeIsolation.ts"; +import { bubblewrapArguments } from "../src/main/services/isolation/bubblewrap.ts"; +import { isolationPaths } from "../src/main/services/isolation/isolationPaths.ts"; +import { TerminalManager } from "../src/main/services/TerminalManager.ts"; +import { availableRegistry, fakeSpawner } from "./helpers/terminal.mjs"; + +// Plugin data and gateway sockets together: the account home a launch plugin selected stays open, a CANVASTTY_*_ADDRESS +// that did not come from CanvasTTY's own gateways binds nothing, and a folder inside plugin data is a project only +// where CanvasTTY itself placed the launch. Any plugin ids: the host treats every plugin's data the same way. +const ACCOUNTS = "com.example.accounts"; +const ENVIRONMENTS = "com.example.env"; +const posixHost = { skip: process.platform === "win32" ? "bubblewrap and seatbelt paths are POSIX paths" : false }; +const mac = process.platform === "darwin" && existsSync(SANDBOX_EXEC); +const onMac = { skip: mac ? false : "macOS seatbelt (sandbox-exec) only" }; +const at = { x: 0, y: 0 }; +const BWRAP = "/usr/bin/bwrap"; +// The program an environment wraps the agent in; the spawner is fake, so it never runs. +const WRAPPER = process.execPath; + +/** A folder that holds CanvasTTY's userData (like a project at HOME), with an accounts plugin and an environment plugin. */ +async function world(t) { + // Short: Unix socket paths are limited to ~104 bytes, so /tmp rather than macOS's long per-user folder. + const base = await realpath(await mkdtemp(join(process.platform === "win32" ? tmpdir() : "/tmp", "ctty-hard-"))); + t.after(() => rm(base, { recursive: true, force: true })); + const home = join(base, "h"); + const userData = join(base, "u"); + const pluginData = join(userData, "plugin-data"); + const homes = join(pluginData, ACCOUNTS, "homes"); + const selected = join(homes, "acct-1"); + const sibling = join(homes, "acct-2"); + const worktrees = join(pluginData, ENVIRONMENTS, "worktrees"); + const worktree = join(worktrees, "w1"); + const otherWorktree = join(worktrees, "w2"); + const lifecycle = join(userData, "lifecycle", "runtime"); + const project = join(base, "p"); + const temp = join(base, "t"); + for (const dir of [home, join(home, ".ssh"), selected, sibling, worktree, otherWorktree, lifecycle, project, temp, + join(userData, "agent-control")]) await mkdir(dir, { recursive: true }); + await writeFile(join(selected, "auth.json"), "FAKE-SELECTED-TOKEN\n"); + await writeFile(join(sibling, "auth.json"), "FAKE-SIBLING-TOKEN\n"); + await writeFile(join(otherWorktree, "file.txt"), "other worktree\n"); + await writeFile(join(userData, "agent-control", "token-app"), "FAKE-APP-TOKEN\n"); + const env = { HOME: home, PATH: "/usr/bin:/bin" }; + return { base, home, userData, pluginData, homes, selected, sibling, worktrees, worktree, otherWorktree, lifecycle, project, temp, env }; +} + +const linux = (w) => new AgentIsolation({ userDataPath: w.userData, enabled: () => true, platform: "linux", bubblewrapPath: BWRAP, + bubblewrapProbe: () => null, tempRoot: w.temp, hostEnvironment: w.env }); + +/** The last mount whose target is `path` or one of its folders (a later mount on a folder hides earlier ones under it). */ +function mountFor(args, path) { + let found = null; + for (let i = 0; i < args.length; i++) { + const flag = args[i]; + if (flag === "--") break; + let target = null; + let mode = null; + if (flag === "--bind" || flag === "--dev-bind") { target = args[i + 2]; mode = "bind"; i += 2; } + else if (flag === "--ro-bind") { target = args[i + 2]; mode = "ro"; i += 2; } + else if (flag === "--tmpfs") { target = args[i + 1]; mode = "tmpfs"; i += 1; } + else if (flag === "--proc" || flag === "--chdir") { i += 1; continue; } + else continue; + if (path === target || path.startsWith(`${target.replace(/\/+$/u, "")}/`)) found = { target, mode }; + } + return found; +} + +const waitFor = async (predicate, timeoutMs = 5_000) => { + const deadline = Date.now() + timeoutMs; + while (Date.now() < deadline) { + if (await predicate()) return; + await new Promise((resolve) => setTimeout(resolve, 10)); + } + throw new Error("Condition was not met in time."); +}; + +/** A TerminalManager whose isolation layer is the real Linux one (bubblewrap arguments, the PTY spawner is fake). */ +function manager(t, w, { runtime, environments } = {}) { + const calls = []; + const terminals = new TerminalManager(() => undefined, availableRegistry(), undefined, runtime, true, fakeSpawner(calls)); + t.after(() => terminals.shutdown()); + const layer = linux(w); + // The launch environment comes from this process; the layer gets the fixture's HOME instead (it may put placeholder + // files into the CLI's folders there), and a CLI home only when a launch plugin set one inside the fixture. + const hermetic = (env) => { + const kept = Object.fromEntries(Object.entries(env).filter(([key]) => !/^(XDG_|CLAUDE_CONFIG_DIR$|GROK_HOME$)/u.test(key))); + return { ...kept, HOME: w.home, CODEX_HOME: env.CODEX_HOME?.startsWith(`${w.base}/`) ? env.CODEX_HOME : join(w.home, ".codex") }; + }; + terminals.configureIsolation({ containment: () => true, decide: (input) => layer.decide(input), + wrap: (launch) => layer.wrap({ ...launch, env: hermetic(launch.env) }) }); + if (environments) terminals.configureEnvironments(environments); + const card = (id) => terminals.list().find((session) => session.id === id); + return { terminals, calls, card }; +} + +test("together: the selected account home stays open, a plugin's CANVASTTY_*_ADDRESS binds nothing, the gateway folder is read-only", posixHost, async (t) => { + const w = await world(t); + const runtimeSocket = join(w.lifecycle, "r.sock"); + // A plugin aims a socket address at the accounts plugin's homes: before, its folder was bound writable after the + // tmpfs over plugin data, which reopened every account's sign-in. + const foreign = join(w.homes, "foo.sock"); + const runtime = { + prepareLaunch: () => ({ args: [], environment: { CANVASTTY_RUNTIME_ADDRESS: runtimeSocket, CANVASTTY_RUNTIME_CAPABILITY: "c" }, cleanup() {} }), + currentStatus: () => null + }; + const { terminals, calls, card } = manager(t, w, { runtime }); + // The launch pipeline already refuses CANVASTTY_* names from plugins; the layer must not depend on that. + terminals.configureLaunchPipeline({ + normalizeOptions: (_provider, options) => options, + unavailable: () => [], + forgetSession: async () => undefined, + prepare: async () => ({ ok: true, env: { CODEX_HOME: w.selected, CANVASTTY_FOO_ADDRESS: foreign }, args: [], secrets: [], + envSources: { CODEX_HOME: "Plugin Accounts", CANVASTTY_FOO_ADDRESS: "Plugin Accounts" }, thirdPartyModel: false, + accountHome: w.selected, cleanup: async () => undefined }) + }); + const created = terminals.create({ provider: "codex", profile: "acceptEdits", cwd: w.project, position: at, + launchOptions: { [ACCOUNTS]: { account: "acct-1" } } }); + await waitFor(() => calls.length === 1 || card(created.id).status === "failed"); + assert.equal(card(created.id).failureDetails, null); + const { command, args, options } = calls[0]; + assert.equal(command, BWRAP); + assert.equal(options.env.CODEX_HOME, w.selected); + assert.equal(options.env.CANVASTTY_FOO_ADDRESS, foreign, "the variable itself still reaches the agent"); + // The selected account home: bound back, writable (it is the CLI's own home). + assert.deepEqual(mountFor(args, join(w.selected, "auth.json")), { target: w.selected, mode: "bind" }); + // The plugin's address bound nothing: the other account and the homes folder stay under the tmpfs over plugin data. + assert.deepEqual(mountFor(args, join(w.sibling, "auth.json")), { target: w.pluginData, mode: "tmpfs" }); + assert.deepEqual(mountFor(args, foreign), { target: w.pluginData, mode: "tmpfs" }); + assert.ok(!args.some((arg, i) => (arg === "--bind" || arg === "--ro-bind") && args[i + 2] === w.homes), "the homes folder is never bound"); + // CanvasTTY's own gateway folder: read-only, never writable. + assert.deepEqual(mountFor(args, runtimeSocket), { target: w.lifecycle, mode: "ro" }); + assert.ok(!args.some((arg, i) => arg === "--bind" && args[i + 2] === w.lifecycle)); +}); + +test("a subagent cannot work in CanvasTTY's private data, even when its project (the home folder) contains it", async (t) => { + const w = await world(t); + const refused = /cannot work in .*: it is CanvasTTY's private data/u; + for (const requested of [w.homes, w.selected, join(w.pluginData, ACCOUNTS), w.pluginData, join(w.userData, "agent-control"), w.worktree]) { + assert.match(subagentFolder(w.base, w.base, requested, w.userData).error ?? "", refused, requested); + } + assert.deepEqual(subagentFolder(w.base, w.base, w.project, w.userData), { cwd: w.project }, "the rest of the project is fine"); + assert.deepEqual(subagentFolder(w.base, w.base, join(w.userData, "lifecycle"), w.userData), { cwd: join(w.userData, "lifecycle") }, + "userData outside its private data is the project's like any folder"); + // A project that is an environment's worktree: its own folders, never a whole plugin's data above it. + const sub = join(w.worktree, "src"); + await mkdir(sub); + assert.deepEqual(subagentFolder(w.worktree, w.worktree, "src", w.userData), { cwd: sub }); + assert.match(subagentFolder(join(w.pluginData, ENVIRONMENTS), w.worktree, w.worktree, w.userData).error ?? "", refused, + "a project that is a whole plugin's data does not open its folders"); + + // Through the service: an orchestrator the person started in the folder above userData. + const { terminals } = manager(t, w); + const control = new AgentControlService(terminals, { containment: () => true, userDataPath: w.userData }); + const parent = terminals.create({ provider: "codex", profile: "normal", cwd: w.base, position: at, role: "orchestrator" }); + assert.throws(() => control.spawn({ parentSessionId: parent.id, provider: "codex", cwd: w.homes }), refused); + const child = await control.spawn({ parentSessionId: parent.id, provider: "codex", cwd: w.project }); + assert.equal(child.cwd, w.project); +}); + +test("the layer opens a project inside plugin data only where CanvasTTY placed the launch", posixHost, async (t) => { + const w = await world(t); + const notPlaced = /inside CanvasTTY's private plugin data, and CanvasTTY did not place this launch there/u; + + // The exploit, past the subagent check: an agent's subagent asked for the account homes as its folder. + const { terminals, calls, card } = manager(t, w); + const parent = terminals.create({ provider: "codex", profile: "normal", cwd: w.base, position: at, role: "orchestrator" }); + const exploit = terminals.create({ provider: "codex", profile: "acceptEdits", cwd: w.homes, position: at, role: "subagent", + parentSessionId: parent.id }); + assert.equal(card(exploit.id).status, "failed"); + assert.match(card(exploit.id).failureDetails, notPlaced); + assert.equal(calls.filter((call) => call.command === BWRAP).length, 0, "nothing ran"); + + // The folder the person chose for a card, and its subagents below it: opened (the person's choice). + const chosen = terminals.create({ provider: "codex", profile: "acceptEdits", cwd: w.worktree, position: at, role: "orchestrator" }); + assert.equal(card(chosen.id).failureDetails, null); + const below = join(w.worktree, "src"); + await mkdir(below); + const nested = terminals.create({ provider: "codex", profile: "acceptEdits", cwd: below, position: at, role: "subagent", parentSessionId: chosen.id }); + assert.equal(card(nested.id).failureDetails, null); + const launched = calls.filter((call) => call.command === BWRAP); + assert.equal(launched.length, 2); + assert.deepEqual(mountFor(launched[0].args, join(w.worktree, "f")), { target: w.worktree, mode: "bind" }); + assert.deepEqual(mountFor(launched[0].args, join(w.otherWorktree, "file.txt")), { target: w.pluginData, mode: "tmpfs" }); + + // The pure check: no placement, a placement elsewhere, or one an environment of another plugin claims: refused. + const base = { provider: "codex", cwd: w.worktree, sessionTemp: join(w.temp, "s"), userDataPath: w.userData, sessionId: "s", env: w.env }; + assert.throws(() => isolationPaths(base), notPlaced); + assert.throws(() => isolationPaths({ ...base, placedProjects: [{ folder: w.base }] }), notPlaced, "a project above plugin data places nothing in it"); + assert.throws(() => isolationPaths({ ...base, placedProjects: [{ folder: w.worktree, pluginId: ACCOUNTS }] }), notPlaced, "another plugin's environment"); + assert.throws(() => isolationPaths({ ...base, placedProjects: [{ folder: w.worktree, pluginId: "../com.example.env" }] }), notPlaced); + assert.throws(() => isolationPaths({ ...base, cwd: w.homes, placedProjects: [{ folder: w.worktree, pluginId: ENVIRONMENTS }] }), notPlaced); + const paths = isolationPaths({ ...base, placedProjects: [{ folder: w.worktree, pluginId: ENVIRONMENTS }] }); + assert.deepEqual(paths.readableAgain.filter((path) => path.startsWith(w.pluginData)), [w.worktree]); +}); + +test("an environment's worktree inside its own plugin data still runs inside the layer; one in another plugin's data does not", posixHost, async (t) => { + const w = await world(t); + const environments = (cwd) => new EnvironmentRegistry({ + providers: () => [{ pluginId: ENVIRONMENTS, pluginName: "Env", serviceId: "env", secrets: false, + kinds: [{ kind: "worktree", label: "Worktree", keeps: { launch: true } }] }], + call: async (_pluginId, _serviceId, method, params) => { + const step = method.replace("canvastty.environment.", ""); + if (step === "prepare") return { ref: { id: "w1" }, label: "worktree w1", cwd }; + if (step === "wrap") return { command: WRAPPER, args: ["run", params.command, ...params.args], cwd: params.cwd }; + if (step === "resume") return { ok: true }; + return {}; + }, + secret: async () => null + }); + const create = (terminals) => terminals.create({ provider: "codex", profile: "acceptEdits", cwd: w.project, position: at, + environment: { pluginId: ENVIRONMENTS, kind: "worktree" } }); + + const own = manager(t, w, { environments: environments(w.worktree) }); + const placed = create(own.terminals); + await waitFor(() => own.calls.length === 1 || own.card(placed.id).status === "failed"); + assert.equal(own.card(placed.id).failureDetails, null); + assert.equal(own.card(placed.id).cwd, w.worktree); + const { command, args } = own.calls[0]; + assert.equal(command, BWRAP); + assert.deepEqual(mountFor(args, join(w.worktree, "new.txt")), { target: w.worktree, mode: "bind" }); + assert.deepEqual(mountFor(args, join(w.otherWorktree, "file.txt")), { target: w.pluginData, mode: "tmpfs" }); + assert.deepEqual(mountFor(args, join(w.selected, "auth.json")), { target: w.pluginData, mode: "tmpfs" }); + + // The environment plugin answers with the accounts plugin's homes: not its own data, so the launch is refused. + const foreign = manager(t, w, { environments: environments(w.homes) }); + const refused = create(foreign.terminals); + await waitFor(() => foreign.card(refused.id).status === "failed"); + assert.match(foreign.card(refused.id).failureDetails, /CanvasTTY did not place this launch there/u); + assert.equal(foreign.calls.length, 0); +}); + +test("account homes and socket folders: a trailing separator names the same home; a socket folder over a writable one is refused", posixHost, async (t) => { + const w = await world(t); + const dataDir = join(w.pluginData, ACCOUNTS); + assert.equal(contributedAccountHome("codex", dataDir, "CODEX_HOME", `${w.selected}/`), w.selected); + assert.equal(contributedAccountHome("codex", dataDir, "CODEX_HOME", `${w.selected}//`), w.selected); + assert.equal(validateAccountHome(w.userData, "codex", w.selected, { CODEX_HOME: `${w.selected}/` }), w.selected); + // Only the CLI's own home variable counts: never HOME or XDG_* (OpenCode's is OPENCODE_CONFIG_DIR). + for (const [provider, key] of [["opencode", "XDG_CONFIG_HOME"], ["opencode", "XDG_DATA_HOME"], ["codex", "HOME"]]) { + assert.equal(contributedAccountHome(provider, dataDir, key, w.selected), undefined, `${provider} ${key}`); + } + assert.equal(contributedAccountHome("opencode", dataDir, "OPENCODE_CONFIG_DIR", w.selected), w.selected); + + // A gateway socket folder that holds the writable project would make it read-only and undo its hooks tmpfs. + const work = join(w.base, "work"); + const project = join(work, "p"); + await mkdir(project, { recursive: true }); + const sessionTemp = join(w.temp, "s"); + const paths = isolationPaths({ provider: "codex", cwd: project, sessionTemp, env: w.env, userDataPath: w.userData, sessionId: "s", + gatewaySockets: [join(work, "gw.sock")] }); + const kinds = new Map([[project, "directory"], [work, "directory"], [sessionTemp, "directory"]]); + assert.throws(() => bubblewrapArguments(paths, { command: "/usr/bin/codex", args: [], cwd: project }, (path) => kinds.get(path) ?? null), + /gateway socket folder .*work contains .*p, which the agent writes/u); +}); + +test("seatbelt, for real: account home open, siblings and a foreign socket closed, gateway folder read-only, homes never a project", onMac, async (t) => { + const w = await world(t); + const runtimeSocket = join(w.lifecycle, "r.sock"); + const foreignFolder = join(w.base, "o"); + await mkdir(foreignFolder); + const foreignSocket = join(foreignFolder, "f.sock"); + const servers = [runtimeSocket, foreignSocket].map((path) => { + const server = createServer((socket) => socket.end()); + server.listen(path); + return server; + }); + t.after(() => servers.forEach((server) => server.close())); + await new Promise((resolve) => setTimeout(resolve, 100)); + const layer = new AgentIsolation({ userDataPath: w.userData, enabled: () => true, tempRoot: w.temp, hostEnvironment: w.env }); + const probe = (path, label) => `/usr/bin/python3 -c "import socket +s=socket.socket(socket.AF_UNIX) +try: + s.connect('${path}'); print('${label}-CONNECT-ok') +except Exception: print('${label}-CONNECT-denied')"`; + const wrapped = layer.wrap({ + sessionId: "together", + provider: "codex", + cwd: w.project, + command: "/bin/sh", + args: ["-c", [ + 'cat "$CODEX_HOME/auth.json" >/dev/null && echo OWN-READ-ok', + 'echo actor > "$CODEX_HOME/actor.txt" && echo OWN-WRITE-ok', + `cat ${JSON.stringify(join(w.sibling, "auth.json"))} >/dev/null 2>&1 || echo SIBLING-READ-denied`, + `ls ${JSON.stringify(w.homes)} >/dev/null 2>&1 || echo HOMES-LIST-denied`, + probe(runtimeSocket, "GATEWAY"), + probe(foreignSocket, "FOREIGN"), + `rm -f ${JSON.stringify(runtimeSocket)} 2>/dev/null; [ -S ${JSON.stringify(runtimeSocket)} ] && echo GATEWAY-RM-denied`, + `touch ${JSON.stringify(join(w.lifecycle, "planted"))} 2>/dev/null || echo GATEWAY-WRITE-denied` + ].join("; ")], + env: { ...w.env, CODEX_HOME: w.selected, CANVASTTY_RUNTIME_ADDRESS: runtimeSocket, CANVASTTY_FOO_ADDRESS: foreignSocket }, + profile: "acceptEdits", + accountHome: w.selected, + gatewaySockets: [runtimeSocket] + }); + try { + const result = spawnSync(wrapped.command, wrapped.args, { cwd: w.project, env: wrapped.env, encoding: "utf8", timeout: 20_000 }); + assert.equal(result.status, 0, result.stderr); + assert.deepEqual(result.stdout.split("\n").filter(Boolean), [ + "OWN-READ-ok", "OWN-WRITE-ok", "SIBLING-READ-denied", "HOMES-LIST-denied", + "GATEWAY-CONNECT-ok", "FOREIGN-CONNECT-denied", "GATEWAY-RM-denied", "GATEWAY-WRITE-denied" + ]); + assert.equal(await readFile(join(w.selected, "actor.txt"), "utf8"), "actor\n"); + assert.equal(existsSync(join(w.lifecycle, "planted")), false); + } finally { + wrapped.cleanup(); + } + // The exploit's folder as a project (a subagent's cwd, not placed by CanvasTTY): the layer is never set up. + assert.throws(() => layer.wrap({ sessionId: "exploit", provider: "codex", cwd: w.homes, command: "/bin/sh", args: ["-c", "true"], + env: w.env, profile: "acceptEdits" }), /did not place this launch there/u); + // An environment's worktree placed by its own plugin: writable there, nothing else of plugin data readable. + const placed = layer.wrap({ sessionId: "worktree", provider: "codex", cwd: w.worktree, command: "/bin/sh", args: ["-c", [ + "echo x > new.txt && echo WORKTREE-WRITE-ok", + `cat ${JSON.stringify(join(w.otherWorktree, "file.txt"))} >/dev/null 2>&1 || echo OTHER-WORKTREE-denied`, + `cat ${JSON.stringify(join(w.selected, "auth.json"))} >/dev/null 2>&1 || echo ACCOUNT-denied` + ].join("; ")], env: w.env, profile: "acceptEdits", placedProjects: [{ folder: w.worktree, pluginId: ENVIRONMENTS }] }); + try { + const result = spawnSync(placed.command, placed.args, { cwd: w.worktree, env: placed.env, encoding: "utf8", timeout: 20_000 }); + assert.deepEqual(result.stdout.split("\n").filter(Boolean), ["WORKTREE-WRITE-ok", "OTHER-WORKTREE-denied", "ACCOUNT-denied"]); + } finally { + placed.cleanup(); + } +}); diff --git a/tests/isolation-linux-host.test.mjs b/tests/isolation-linux-host.test.mjs index a9ac1b77b..085e65a46 100644 --- a/tests/isolation-linux-host.test.mjs +++ b/tests/isolation-linux-host.test.mjs @@ -84,6 +84,26 @@ test("bubblewrap: a missing CLI permission file inside its writable folder is pr assert.equal(existsSync(join(w.home, ".claude", "settings.json")), false); }); +test("bubblewrap: a CANVASTTY_*_ADDRESS into HOME binds nothing and un-hides nothing; a core gateway folder over a hidden path is refused", posixHost, async (t) => { + const w = await freshHome(t); + await mkdir(join(w.home, ".ssh")); + await writeFile(join(w.home, ".ssh", "id_test"), "FAKE-PRIVATE-KEY"); + const lifecycle = join(w.userData, "lifecycle", "runtime"); + await mkdir(lifecycle, { recursive: true }); + // The person's or a plugin's variable: in the environment only, not a gateway CanvasTTY set up. + const wrapped = linux(w).wrap({ sessionId: "s1", provider: "codex", cwd: w.project, command: "/bin/sh", args: ["-c", "true"], + env: { ...w.env, CANVASTTY_FOO_ADDRESS: join(w.home, "foo.sock") }, gatewaySockets: [join(lifecycle, "r.sock")] }); + try { + assert.equal(mountFor(wrapped.args, w.home)?.target, "/", "HOME is not bound"); + assert.equal(mountFor(wrapped.args, join(w.home, ".ssh", "id_test"))?.mode, "tmpfs", "its keys stay hidden"); + assert.equal(persistsWrite(wrapped.args, join(w.home, "notes.txt")), false); + assert.deepEqual(mountFor(wrapped.args, join(lifecycle, "r.sock")), { target: lifecycle, mode: "ro" }, "the core gateway folder, read-only"); + } finally { wrapped.cleanup(); } + // A core gateway whose folder holds something hidden is refused rather than mounted over it. + assert.throws(() => linux(w).wrap({ sessionId: "s2", provider: "codex", cwd: w.project, command: "/bin/sh", args: ["-c", "true"], env: w.env, + gatewaySockets: [join(w.home, "gw.sock")] }), /gateway socket folder .* contains .*\.ssh, which the agent may not read/u); +}); + test("bubblewrap: a fresh HOME gets the CLI's own state folders, so its first run can write them", posixHost, async (t) => { const w = await freshHome(t); const wrapped = wrapIn(linux(w), w, "opencode"); diff --git a/tests/isolation-userdata-readonly.test.mjs b/tests/isolation-userdata-readonly.test.mjs new file mode 100644 index 000000000..a250408cd --- /dev/null +++ b/tests/isolation-userdata-readonly.test.mjs @@ -0,0 +1,131 @@ +import assert from "node:assert/strict"; +import { spawnSync } from "node:child_process"; +import { existsSync } from "node:fs"; +import { mkdir, mkdtemp, readFile, realpath, rm, writeFile } from "node:fs/promises"; +import { tmpdir } from "node:os"; +import { join } from "node:path"; +import test from "node:test"; +import { AgentIsolation } from "../src/main/services/isolation/AgentIsolation.ts"; +import { isolationPaths } from "../src/main/services/isolation/isolationPaths.ts"; +import { seatbeltProfile } from "../src/main/services/isolation/seatbelt.ts"; +import { bubblewrapArguments } from "../src/main/services/isolation/bubblewrap.ts"; + +const mac = process.platform === "darwin" && existsSync("/usr/bin/sandbox-exec"); +const onMac = { skip: mac ? false : "macOS seatbelt (sandbox-exec) only" }; + +/** A fake HOME that holds CanvasTTY's data folder (its settings, a plugin, an account home, a project the app keeps). */ +async function world(t) { + const base = await realpath(await mkdtemp(join(process.platform === "win32" ? tmpdir() : "/tmp", "ctty-ud-test-"))); + t.after(() => rm(base, { recursive: true, force: true })); + const home = join(base, "h"); + const userData = join(home, ".config", "canvastty"); + const temp = join(base, "t"); + const accountHome = join(userData, "account-homes", "a1"); + const appProject = join(userData, "projects", "p"); + for (const dir of [temp, join(home, "src"), join(userData, "plugins", "x"), accountHome, appProject, join(userData, "agent-control")]) { + await mkdir(dir, { recursive: true }); + } + await writeFile(join(userData, "settings.json"), "{\"agentIsolation\":\"on\"}\n"); + await writeFile(join(userData, "plugins", "x", "index.js"), "// plugin\n"); + await writeFile(join(userData, "agent-control", "token-app"), "APP-TOKEN"); + const env = { HOME: home, PATH: "/usr/bin:/bin:/usr/sbin:/sbin", LANG: "en_US.UTF-8" }; + return { base, home, userData, temp, accountHome, appProject, env }; +} + +function pathsFor(w, cwd, env = {}) { + return isolationPaths({ provider: "codex", cwd, sessionTemp: join(w.temp, "s"), env: { ...w.env, ...env }, userDataPath: w.userData, sessionId: "s" }); +} + +test("the paths: a project that contains CanvasTTY's data folder leaves that folder read-only, its account home writable", async (t) => { + const w = await world(t); + const paths = pathsFor(w, w.home, { CODEX_HOME: w.accountHome }); + assert.ok(paths.writable.includes(w.home), "the project stays writable"); + assert.ok(paths.readOnlyFolders.includes(w.userData), "CanvasTTY's data folder is read-only"); + const profile = seatbeltProfile(paths); + const deny = profile.indexOf(`(deny file-write* (subpath "${w.userData}")`); + assert.ok(deny > profile.indexOf(`(subpath "${w.home}")`), "denied after the project is allowed"); + const again = profile.indexOf(`(allow file-write* (subpath "${w.accountHome}"))`); + assert.ok(again > deny, "the CLI's own account home is writable again"); + // Its parent inside the project cannot be renamed away (and back) or replaced. + assert.deepEqual(paths.fixedFolders, [join(w.home, ".config")]); + assert.ok(profile.includes(`(deny file-write-unlink (literal "${join(w.home, ".config")}"))`)); + // A project the app keeps inside its data folder (Even G2's default), or any other project: nothing changes. + assert.deepEqual(pathsFor(w, w.appProject).readOnlyFolders, []); + assert.deepEqual(pathsFor(w, w.appProject).fixedFolders, []); + assert.ok(pathsFor(w, w.appProject).writable.includes(w.appProject)); + assert.deepEqual(pathsFor(w, join(w.home, "src")).readOnlyFolders, []); + // A read-only (plan) project at HOME has nothing to take back. + assert.deepEqual(isolationPaths({ provider: "codex", cwd: w.home, sessionTemp: join(w.temp, "s"), env: w.env, userDataPath: w.userData, + sessionId: "s", readOnlyProject: true }).readOnlyFolders, []); +}); + +test("bubblewrap: CanvasTTY's data folder is mounted read-only over a project at HOME, its account home bound writable again", async (t) => { + const w = await world(t); + const paths = pathsFor(w, w.home, { CODEX_HOME: w.accountHome }); + const kinds = new Map([[w.home, "directory"], [join(w.home, ".config"), "directory"], [w.userData, "directory"], [w.accountHome, "directory"], [join(w.userData, "agent-control"), "directory"]]); + const args = bubblewrapArguments(paths, { command: "/usr/bin/codex", args: [], cwd: w.home }, + // The CLI's permission settings exist (LinuxHostPaths puts a placeholder where one is missing). + (path) => kinds.get(path) ?? (paths.protectedWrites.includes(path) && !path.endsWith(".lock") ? "file" : null)).join(" "); + const project = args.indexOf(`--bind ${w.home} ${w.home}`); + const readOnly = args.indexOf(`--ro-bind ${w.userData} ${w.userData}`); + assert.ok(project >= 0 && readOnly > project, "read-only over the writable project"); + const parent = args.indexOf(`--bind ${join(w.home, ".config")} ${join(w.home, ".config")}`); + assert.ok(parent > project && parent < readOnly, "its parent is a mount point (cannot be renamed), mounted before it"); + assert.ok(args.indexOf(`--bind ${w.accountHome} ${w.accountHome}`, readOnly) > readOnly, "the account home is writable again"); + assert.ok(args.indexOf(`--tmpfs ${join(w.userData, "agent-control")}`) > readOnly, "private data still hidden"); + // A project inside the data folder is bound writable, nothing over it. + const inside = bubblewrapArguments(pathsFor(w, w.appProject), { command: "/usr/bin/codex", args: [], cwd: w.appProject }, + (path) => path === w.appProject || path === w.userData ? "directory" : null).join(" "); + assert.ok(inside.includes(`--bind ${w.appProject} ${w.appProject}`)); + assert.ok(!inside.includes(`--ro-bind ${w.userData} ${w.userData}`)); +}); + +function run(w, cwd, script, env = {}) { + const wrapped = new AgentIsolation({ userDataPath: w.userData, enabled: () => true, tempRoot: w.temp }) + .wrap({ sessionId: "s1", provider: "codex", cwd, command: "/bin/sh", args: ["-c", script], env: { ...w.env, ...env } }); + try { + const result = spawnSync(wrapped.command, wrapped.args, { cwd, env: wrapped.env, encoding: "utf8", timeout: 20_000 }); + return result.stdout.split("\n").filter(Boolean); + } finally { + wrapped.cleanup(); + } +} + +test("seatbelt, for real: a project at HOME cannot change CanvasTTY's settings or plugins; ordinary work still succeeds", onMac, async (t) => { + const w = await world(t); + const ud = w.userData; + const lines = run(w, w.home, [ + 'echo ok > notes.txt && echo W-project-ok', + 'mkdir -p src/deep && echo ok > src/deep/f && echo W-subfolder-ok', + `echo '{}' > "${ud}/settings.json" 2>/dev/null || echo W-settings-denied`, + `echo x > "${ud}/settings.json.tmp" 2>/dev/null || echo W-settings-tmp-denied`, + `echo x > "${ud}/plugins/x/index.js" 2>/dev/null || echo W-plugin-denied`, + `mkdir "${ud}/plugins/y" 2>/dev/null || echo W-new-plugin-denied`, + `rm -rf "${ud}/plugins" 2>/dev/null; [ -f "${ud}/plugins/x/index.js" ] && echo RM-plugins-denied`, + `mv "${ud}" "$HOME/moved" 2>/dev/null || echo MV-userdata-denied`, + 'mv "$HOME/.config" "$HOME/moved" 2>/dev/null || echo MV-parent-denied', + 'rm -rf "$HOME/.config" 2>/dev/null; [ -d "$HOME/.config" ] && echo RM-parent-denied', + 'mkdir "$HOME/.config/other-app" && echo x > "$HOME/.config/other-app/f" && echo W-parent-entry-ok', + `echo x > "${ud}/projects/p/f" 2>/dev/null || echo W-app-project-denied`, + `cat "${ud}/settings.json" >/dev/null && echo R-settings-ok`, + `cat "${ud}/agent-control/token-app" 2>/dev/null || echo R-token-denied`, + 'echo s > "$CODEX_HOME/auth.json" && echo W-account-home-ok' + ].join("; "), { CODEX_HOME: w.accountHome }); + assert.deepEqual(lines, ["W-project-ok", "W-subfolder-ok", "W-settings-denied", "W-settings-tmp-denied", "W-plugin-denied", "W-new-plugin-denied", + "RM-plugins-denied", "MV-userdata-denied", "MV-parent-denied", "RM-parent-denied", "W-parent-entry-ok", "W-app-project-denied", "R-settings-ok", "R-token-denied", "W-account-home-ok"]); + assert.equal(await readFile(join(ud, "settings.json"), "utf8"), "{\"agentIsolation\":\"on\"}\n"); + assert.equal(await readFile(join(ud, "plugins", "x", "index.js"), "utf8"), "// plugin\n"); + assert.equal(existsSync(join(ud, "plugins", "y")), false); + assert.equal(existsSync(join(w.home, "moved")), false); + assert.equal(await readFile(join(w.home, "notes.txt"), "utf8"), "ok\n"); +}); + +test("seatbelt, for real: a project CanvasTTY keeps inside its data folder stays writable", onMac, async (t) => { + const w = await world(t); + const lines = run(w, w.appProject, [ + 'echo ok > f && mkdir -p sub && echo ok > sub/g && echo W-project-ok', + `echo '{}' > "${w.userData}/settings.json" 2>/dev/null || echo W-settings-denied` + ].join("; ")); + assert.deepEqual(lines, ["W-project-ok", "W-settings-denied"]); + assert.equal(await readFile(join(w.appProject, "sub", "g"), "utf8"), "ok\n"); +}); diff --git a/tests/launch-cleanup-order.test.mjs b/tests/launch-cleanup-order.test.mjs new file mode 100644 index 000000000..f18e5c115 --- /dev/null +++ b/tests/launch-cleanup-order.test.mjs @@ -0,0 +1,211 @@ +import assert from "node:assert/strict"; +import { mkdtemp, rm, stat } from "node:fs/promises"; +import { tmpdir } from "node:os"; +import { join } from "node:path"; +import test from "node:test"; +import { LaunchPipeline } from "../src/main/services/LaunchPipeline.ts"; +import { TerminalManager } from "../src/main/services/TerminalManager.ts"; +import { availableRegistry, fakeSpawner } from "./helpers/terminal.mjs"; + +function deferred() { + let resolve; + const promise = new Promise((done) => { resolve = done; }); + return { promise, resolve }; +} + +function cleanupPipeline(runs, serviceCalls = []) { + return new LaunchPipeline({ + contributors: () => [{ + pluginId: "fixture.cleanup", + pluginName: "Cleanup fixture", + serviceId: "launch", + secrets: false, + launch: { fields: [{ key: "enabled", label: "Enabled", kind: "boolean", default: true }] } + }], + call: async (...args) => { + serviceCalls.push(args); + return { env: { FIXTURE_FILE: "{launchFiles}/run.txt" }, files: [{ relPath: "run.txt", content: "temporary launch data" }] }; + }, + secret: async () => null, + runsRoot: runs, + timeoutMs: 2000 + }); +} + +const createCard = (terminals, cwd) => terminals.create({ + provider: "codex", + profile: "normal", + cwd, + position: { x: 0, y: 0 }, + launchOptions: { "fixture.cleanup": { enabled: true } } +}); + +/** + * One card with a launch plugin. Each run's cleanup is held until the test releases it, and every step of the + * child (run folder) and parent (card folder) removals is recorded in `events`. + */ +async function fixture(t, { rejectChild = false } = {}) { + const root = await mkdtemp(join(tmpdir(), "ctty-cleanup-order-")); + const runs = join(root, "runs"); + const rows = []; + const calls = []; + const events = []; + const spawned = [deferred(), deferred()]; + const forgotten = deferred(); + const pipeline = cleanupPipeline(runs); + const prepare = pipeline.prepare.bind(pipeline); + const forget = pipeline.forgetSession.bind(pipeline); + pipeline.prepare = async (context) => { + const contribution = await prepare(context); + assert.equal(contribution.ok, true); + const cleanup = contribution.cleanup; + const index = rows.length; + const row = { started: deferred(), release: deferred(), calls: 0 }; + rows.push(row); + contribution.cleanup = async () => { + row.calls++; + events.push(`child${index}:start`); + row.started.resolve(); + try { + await row.release.promise; + if (rejectChild) throw new Error("fixture child cleanup failure"); + await cleanup(); + } finally { + events.push(`child${index}:settled`); + } + }; + return contribution; + }; + let forgetCalls = 0; + pipeline.forgetSession = (id, runDirectories) => { + forgetCalls++; + events.push("parent:start"); + const pending = forget(id, runDirectories).then(() => { events.push("parent:done"); }); + forgotten.resolve(pending); + return pending; + }; + const fake = fakeSpawner(calls); + const terminals = new TerminalManager(() => undefined, availableRegistry(), undefined, undefined, true, (...args) => { + const pty = fake(...args); + spawned[calls.length - 1]?.resolve(); + return pty; + }); + terminals.configureLaunchPipeline(pipeline); + t.after(async () => { + for (const row of rows) row.release.resolve(); + terminals.disposeAll(); + await forgotten.promise; + await rm(root, { recursive: true, force: true }); + }); + const session = createCard(terminals, root); + await spawned[0].promise; + return { root, runs, rows, calls, events, spawned, forgotten, terminals, session, forgetCalls: () => forgetCalls }; +} + +for (const exited of [false, true]) { + for (const rejectChild of [false, true]) { + test(`parent launch cleanup waits for ${exited ? "previous exit" : "active disposal"} cleanup (${rejectChild ? "rejection" : "success"})`, { timeout: 5000 }, async (t) => { + const f = await fixture(t, { rejectChild }); + if (exited) f.calls[0].process.emitExit(1); + f.terminals.dispose(f.session.id); + f.terminals.dispose(f.session.id); + await f.rows[0].started.promise; + assert.equal(f.forgetCalls(), 0, "parent removal must not overlap a held child removal"); + assert.deepEqual(f.events, ["child0:start"]); + + f.rows[0].release.resolve(); + await f.forgotten.promise; + assert.deepEqual(f.events, ["child0:start", "child0:settled", "parent:start", "parent:done"]); + assert.equal(f.rows[0].calls, 1); + assert.equal(f.forgetCalls(), 1, "disposal remains idempotent"); + await assert.rejects(stat(join(f.runs, f.session.id)), { code: "ENOENT" }, "parent still cleaned after child failure"); + }); + } +} + +test("restart retains every pending run cleanup and invokes each captured callback once before forgetting the card", { timeout: 5000 }, async (t) => { + const f = await fixture(t); + f.calls[0].process.emitExit(1); + await f.rows[0].started.promise; + f.terminals.restart(f.session.id); + await f.spawned[1].promise; + assert.equal(f.rows.length, 2); + + f.terminals.dispose(f.session.id); + assert.equal(f.rows[1].calls, 0, "the newer cleanup queues behind the old launch callback"); + assert.equal(f.forgetCalls(), 0); + // A late notification from the old PTY must neither consume nor overwrite the new launch cleanup. + f.calls[0].process.emitExit(1); + f.rows[0].release.resolve(); + await f.rows[1].started.promise; + assert.equal(f.forgetCalls(), 0, "parent waits for both run directories"); + + f.rows[1].release.resolve(); + await f.forgotten.promise; + assert.deepEqual(f.events, ["child0:start", "child0:settled", "child1:start", "child1:settled", "parent:start", "parent:done"]); + assert.deepEqual(f.rows.map((row) => row.calls), [1, 1]); + assert.equal(f.forgetCalls(), 1); + await assert.rejects(stat(join(f.runs, f.session.id)), { code: "ENOENT" }); +}); + +test("a launch for a card closed in the same tick never asks its plugins or spawns", { timeout: 5000 }, async (t) => { + const root = await mkdtemp(join(tmpdir(), "ctty-cleanup-order-")); + const serviceCalls = []; + const calls = []; + const pipeline = cleanupPipeline(join(root, "runs"), serviceCalls); + const forgotten = []; + const forget = pipeline.forgetSession.bind(pipeline); + pipeline.forgetSession = (...args) => { + const pending = forget(...args); + forgotten.push(pending); + return pending; + }; + const terminals = new TerminalManager(() => undefined, availableRegistry(), undefined, undefined, true, fakeSpawner(calls)); + terminals.configureLaunchPipeline(pipeline); + t.after(async () => { + terminals.disposeAll(); + await rm(root, { recursive: true, force: true }); + }); + + const session = createCard(terminals, root); + const internal = terminals.sessions.get(session.id); + const tasks = [...internal.launchTasks]; + terminals.dispose(session.id); + await Promise.allSettled(tasks); + await Promise.allSettled(forgotten); + + assert.equal(serviceCalls.length, 0, "no plugin prepares a launch for a card that is already closed"); + assert.equal(calls.length, 0); + await assert.rejects(stat(join(root, "runs", session.id)), { code: "ENOENT" }); +}); + +test("a failed launch-file removal at close is reported with the card id", { timeout: 5000 }, async (t) => { + const root = await mkdtemp(join(tmpdir(), "ctty-cleanup-order-")); + const calls = []; + const pipeline = cleanupPipeline(join(root, "runs")); + const failure = new Error("fixture forget failure"); + const forgotten = deferred(); + pipeline.forgetSession = async () => { + forgotten.resolve(); + throw failure; + }; + const warnings = []; + t.mock.method(console, "warn", (...args) => { warnings.push(args); }); + const terminals = new TerminalManager(() => undefined, availableRegistry(), undefined, undefined, true, fakeSpawner(calls)); + terminals.configureLaunchPipeline(pipeline); + t.after(async () => { + terminals.disposeAll(); + await rm(root, { recursive: true, force: true }); + }); + + const session = createCard(terminals, root); + await Promise.allSettled([...terminals.sessions.get(session.id).launchTasks]); + terminals.dispose(session.id); + await forgotten.promise; + for (let i = 0; i < 20 && !warnings.length; i++) await new Promise((done) => setImmediate(done)); + + const warning = warnings.find(([message]) => String(message).includes(session.id)); + assert.ok(warning, "the close reports which card's launch files stayed behind"); + assert.ok(warning[1] instanceof AggregateError); + assert.deepEqual(warning[1].errors, [failure]); +}); diff --git a/tests/launch-contributors.test.mjs b/tests/launch-contributors.test.mjs index df2491a1e..8a53e7ee8 100644 --- a/tests/launch-contributors.test.mjs +++ b/tests/launch-contributors.test.mjs @@ -310,11 +310,11 @@ test("secret env comes from the plugin's own secrets, reaches the child, and is calls[0].print(`env: PROVIDER_API_KEY=${SECRET} PLAIN=visible\n`); const control = new AgentControlService(manager); - const observed = control.observe(created.id).output; + const observed = (await control.observe(created.id)).output; assert.match(observed, /PROVIDER_API_KEY= PLAIN=visible/u); assert.doesNotMatch(observed, new RegExp(SECRET, "u")); calls[0].exit(1); - assert.doesNotMatch(control.result(created.id).output, new RegExp(SECRET, "u")); + assert.doesNotMatch((await control.result(created.id)).output, new RegExp(SECRET, "u")); assert.doesNotMatch(card(manager, created.id).failureDetails ?? "", new RegExp(SECRET, "u")); await manager.shutdown(); diff --git a/tests/launch-file-lifecycle.test.mjs b/tests/launch-file-lifecycle.test.mjs new file mode 100644 index 000000000..7131f2ff5 --- /dev/null +++ b/tests/launch-file-lifecycle.test.mjs @@ -0,0 +1,282 @@ +import assert from "node:assert/strict"; +import { mkdtemp, readdir, rm, stat } from "node:fs/promises"; +import { tmpdir } from "node:os"; +import { basename, join } from "node:path"; +import test from "node:test"; +import { EnvironmentRegistry } from "../src/main/services/EnvironmentRegistry.ts"; +import { LaunchPipeline } from "../src/main/services/LaunchPipeline.ts"; +import { TerminalManager } from "../src/main/services/TerminalManager.ts"; +import { persistedTerminalSession } from "../src/main/services/TerminalSessionStore.ts"; +import { availableRegistry, fakeSpawner } from "./helpers/terminal.mjs"; + +function deferred() { + let resolve; + const promise = new Promise((done) => { resolve = done; }); + return { promise, resolve }; +} + +async function fixture(t, { spawnFailure = false } = {}) { + const root = await mkdtemp(join(tmpdir(), "ctty-launch-files-")); + const runs = join(root, "runs"); + const calls = []; + const pipelines = []; + const gates = []; + const disposals = []; + const fake = fakeSpawner(calls); + const terminals = new TerminalManager(() => undefined, availableRegistry(), undefined, undefined, true, (...args) => { + if (spawnFailure) throw new Error("fixture PTY spawn failure"); + return fake(...args); + }); + // Closing a card starts its run-file cleanup in the background; the tests await that same promise. + const forget = terminals.forgetLaunchFiles.bind(terminals); + terminals.forgetLaunchFiles = (session) => { + const pending = forget(session); + disposals.push(pending); + return pending; + }; + const gate = () => { + const result = deferred(); + gates.push(result); + return result; + }; + + function pipeline({ holdPrepare = false, holdForget = false, failCleanup = false } = {}) { + const started = deferred(); + const release = gate(); + const forgetStarted = deferred(); + const forgetRelease = gate(); + const events = []; + const rows = []; + if (!holdPrepare) release.resolve(); + if (!holdForget) forgetRelease.resolve(); + const value = new LaunchPipeline({ + contributors: () => [{ + pluginId: "fixture.files", + pluginName: "Files", + serviceId: "launch", + secrets: false, + launch: { fields: [{ key: "enabled", label: "Enabled", kind: "boolean" }] } + }], + call: async () => { + started.resolve(); + await release.promise; + return { env: { FIXTURE_FILE: "{launchFiles}/run.txt" }, files: [{ relPath: "run.txt", content: "temporary" }] }; + }, + secret: async () => null, + runsRoot: runs, + timeoutMs: 4000 + }); + const prepare = value.prepare.bind(value); + const remove = value.forgetSession.bind(value); + value.prepare = async (context) => { + const result = await prepare(context); + assert.equal(result.ok, true); + const row = { file: result.env.FIXTURE_FILE, runDirectory: result.runDirectory, cleanupCalls: 0 }; + const cleanup = result.cleanup; + rows.push(row); + result.cleanup = async () => { + row.cleanupCalls++; + events.push("child"); + if (failCleanup) throw new Error("fixture cleanup failure"); + await cleanup(); + }; + return result; + }; + value.forgetSession = async (id, runDirectories) => { + events.push("parent:start"); + forgetStarted.resolve(); + await forgetRelease.promise; + await remove(id, runDirectories); + events.push("parent:done"); + }; + const result = { value, started, release, forgetStarted, forgetRelease, events, rows }; + pipelines.push(result); + return result; + } + + const create = (extra = {}) => terminals.create({ + provider: "codex", + profile: "normal", + cwd: root, + position: { x: 0, y: 0 }, + launchOptions: { "fixture.files": { enabled: true } }, + ...extra + }); + const drain = async () => { await Promise.allSettled(disposals); }; + t.after(async () => { + for (const item of gates) item.resolve(); + terminals.disposeAll(); + await drain(); + await rm(root, { recursive: true, force: true }); + }); + return { root, runs, calls, terminals, pipeline, create, gate, drain, disposals }; +} + +for (const failCleanup of [false, true]) { + test(`closing during preparation drains its late files before parent deletion (${failCleanup ? "failed" : "successful"} child cleanup)`, { timeout: 5000 }, async (t) => { + const f = await fixture(t); + const p = f.pipeline({ holdPrepare: true, failCleanup }); + f.terminals.configureLaunchPipeline(p.value); + const session = f.create(); + await p.started.promise; + f.terminals.dispose(session.id); + assert.deepEqual(p.events, []); + + p.release.resolve(); + await f.drain(); + assert.deepEqual(p.events, ["child", "parent:start", "parent:done"]); + assert.equal(p.rows[0].cleanupCalls, 1); + assert.equal(f.calls.length, 0); + // A failed child cleanup is retried by the parent step, which still removes only this card's own run folder. + await assert.rejects(stat(join(f.runs, session.id)), { code: "ENOENT" }); + }); +} + +test("restart tracks both pending preparations and sequentially forgets their original pipelines after reconfiguration", { timeout: 5000 }, async (t) => { + const f = await fixture(t); + const a = f.pipeline({ holdPrepare: true, holdForget: true }); + const b = f.pipeline({ holdPrepare: true }); + const c = f.pipeline(); + f.terminals.configureLaunchPipeline(a.value); + const session = f.create(); + await a.started.promise; + // Force a restart while its original plugin response is still in flight, as restoration can do. + const internal = f.terminals.sessions.get(session.id); + internal.metadata.status = "failed"; + internal.metadata.exitCode = 1; + f.terminals.configureLaunchPipeline(b.value); + f.terminals.restart(session.id); + await b.started.promise; + f.terminals.configureLaunchPipeline(c.value); + f.terminals.dispose(session.id); + + a.release.resolve(); + await Promise.resolve(); + assert.deepEqual(a.events, [], "parent cannot precede pending preparation"); + b.release.resolve(); + await a.forgetStarted.promise; + assert.deepEqual(a.events, ["child", "parent:start"]); + assert.deepEqual(b.events, ["child"], "same parent cannot be removed concurrently by another pipeline"); + assert.deepEqual(c.events, [], "current configuration is not the owner"); + + a.forgetRelease.resolve(); + await f.drain(); + assert.deepEqual(b.events, ["child", "parent:start", "parent:done"]); + assert.deepEqual([a.rows[0].cleanupCalls, b.rows[0].cleanupCalls], [1, 1]); + assert.equal(f.calls.length, 0); + await assert.rejects(stat(join(f.runs, session.id)), { code: "ENOENT" }); +}); + +for (const outcome of ["success", "refusal", "throw"]) { + test(`close while environment wrapping ${outcome} releases planned capability and files once`, { timeout: 5000 }, async (t) => { + const f = await fixture(t); + const p = f.pipeline(); + f.terminals.configureLaunchPipeline(p.value); + const wrapping = deferred(); + const release = f.gate(); + let plannedCleanups = 0; + const plan = f.terminals.planSpawn.bind(f.terminals); + f.terminals.planSpawn = (...args) => { + const result = plan(...args); + assert.ok(!("failure" in result)); + const cleanup = result.cleanup; + result.cleanup = () => { + plannedCleanups++; + cleanup(); + }; + return result; + }; + const registry = new EnvironmentRegistry({ + providers: () => [{ + pluginId: "fixture.env", + pluginName: "Environment", + serviceId: "env", + secrets: false, + kinds: [{ kind: "test", label: "Test", fields: [] }] + }], + call: async (_plugin, _service, method) => method.endsWith(".prepare") ? { ref: { id: "one" }, label: "Test" } : {}, + secret: async () => null + }); + registry.wrap = async (_environment, input) => { + wrapping.resolve(); + await release.promise; + if (outcome === "throw") throw new Error("fixture wrapper threw"); + if (outcome === "refusal") return { ok: false, reason: "fixture refused" }; + return { ok: true, ...input.launch, secrets: [] }; + }; + f.terminals.configureEnvironments(registry); + const session = f.create({ environment: { pluginId: "fixture.env", kind: "test" } }); + await wrapping.promise; + await stat(p.rows[0].file); + f.terminals.dispose(session.id); + assert.deepEqual(p.events, []); + + release.resolve(); + await f.drain(); + assert.equal(plannedCleanups, 1); + assert.equal(p.rows[0].cleanupCalls, 1); + assert.equal(f.calls.length, 0); + assert.deepEqual(p.events, ["child", "parent:start", "parent:done"]); + await assert.rejects(stat(join(f.runs, session.id)), { code: "ENOENT" }); + }); +} + +test("spawn refusal releases unadopted files and capabilities before eventual parent cleanup", { timeout: 5000 }, async (t) => { + const f = await fixture(t, { spawnFailure: true }); + const p = f.pipeline(); + f.terminals.configureLaunchPipeline(p.value); + let cleanups = 0; + const plan = f.terminals.planSpawn.bind(f.terminals); + f.terminals.planSpawn = (...args) => { + const result = plan(...args); + const cleanup = result.cleanup; + result.cleanup = () => { + cleanups++; + cleanup(); + }; + return result; + }; + const session = f.create(); + const internal = f.terminals.sessions.get(session.id); + await Promise.all([...internal.launchTasks]); + assert.equal(internal.metadata.status, "failed"); + assert.equal(cleanups, 1); + assert.equal(p.rows[0].cleanupCalls, 1); + assert.deepEqual(p.events, ["child"]); + + f.terminals.dispose(session.id); + await f.drain(); + assert.deepEqual(p.events, ["child", "parent:start", "parent:done"]); + await assert.rejects(stat(join(f.runs, session.id)), { code: "ENOENT" }); +}); + +test("a delayed close keeps the run files of a card brought back under the same id", { timeout: 5000 }, async (t) => { + const f = await fixture(t); + const p = f.pipeline({ holdForget: true }); + f.terminals.configureLaunchPipeline(p.value); + const session = f.create(); + const first = f.terminals.sessions.get(session.id); + await Promise.all([...first.launchTasks]); + assert.equal(f.calls.length, 1); + const record = persistedTerminalSession(first.metadata, first.threadId, first.extras); + const firstRun = p.rows[0].runDirectory; + + // Close the card; its folder removal is held, as a slow disk or a busy Windows file would hold it. + f.terminals.dispose(session.id); + await p.forgetStarted.promise; + + // The same card comes back (restore under its saved id) and prepares a new run before the old removal finishes. + await f.terminals.restoreSessionRecords([{ ...record, lastState: "running" }], "continue"); + const second = f.terminals.sessions.get(session.id); + assert.ok(second && second !== first); + await Promise.all([...(second.launchTasks ?? [])]); + assert.equal(p.rows.length, 2); + const secondRun = p.rows[1].runDirectory; + assert.notEqual(secondRun, firstRun); + + p.forgetRelease.resolve(); + await f.drain(); + await assert.rejects(stat(firstRun), { code: "ENOENT" }, "the closed card's own run folder is removed"); + await stat(p.rows[1].file); + assert.deepEqual(await readdir(join(f.runs, session.id)), [basename(secondRun)], "the newer card's run folder is kept"); +}); diff --git a/tests/launch-readiness.test.mjs b/tests/launch-readiness.test.mjs index b1aa64656..00716d505 100644 --- a/tests/launch-readiness.test.mjs +++ b/tests/launch-readiness.test.mjs @@ -14,9 +14,9 @@ import { controlRequest } from "../scripts/canvastty-control.mjs"; import { AgentControlService } from "../src/main/services/AgentControlService.ts"; import { LaunchPipeline } from "../src/main/services/LaunchPipeline.ts"; import { PluginSessions } from "../src/main/services/PluginSessions.ts"; -import { TerminalManager } from "../src/main/services/TerminalManager.ts"; +import { INPUT_READY_HOOK_GRACE_MS, TerminalManager } from "../src/main/services/TerminalManager.ts"; import { AgentControlGateway } from "../src/main/services/agent-control/AgentControlGateway.ts"; -import { ScopedOrchestrationHandler } from "../src/main/services/agent-browser/OrchestrationTools.ts"; +import { AGENT_TOOL_DELIVERY_MS, ScopedOrchestrationHandler } from "../src/main/services/agent-browser/OrchestrationTools.ts"; const at = { x: 0, y: 0 }; const cwd = process.cwd(); @@ -209,3 +209,236 @@ test("the control CLI says a starting session is not ready instead of claiming t assert.equal(f.calls.length, 2); assert.deepEqual(f.calls[1].written, [], "the refused send was not queued"); }); + +// OpenCode drops every key that arrives while its full-screen view starts, so a prompt typed when its PTY exists is +// lost and the agent sits at an empty prompt. Its prompt goes in only once its input field is up: a lifecycle hook, +// or the control sequence that shows its text cursor on the alternate screen (never its human-readable screen text). + +const ALT_SCREEN_HIDDEN_CURSOR = "\x1b[?1049h\x1b[?2004h\x1b[?25l"; +const CURSOR_SHOWN = "\x1b[1 q\x1b[?25h"; + +/** + * A slow-starting OpenCode stand-in: it draws its start-up for `startupMs` with the cursor hidden and drops every key + * written meanwhile (as OpenCode does), then shows its input field. `accepted` holds what reached that field. + */ +function slowOpenCode({ startupMs = 120, ready = true } = {}) { + const runs = []; + const spawn = () => { + let emit = () => undefined; + const exits = []; + const run = { written: [], accepted: [], dropped: [], ready: false, exit: (code) => exits.forEach((listener) => listener({ exitCode: code })) }; + runs.push(run); + setTimeout(() => emit(ALT_SCREEN_HIDDEN_CURSOR), 5); + setTimeout(() => emit(" Ask anything… \"Fix broken tests\" tab agents ctrl+p commands"), startupMs / 2); + if (ready) setTimeout(() => { run.ready = true; emit(CURSOR_SHOWN); }, startupMs); + return { + pid: 70_000 + runs.length, process: "opencode", resize() {}, kill() {}, pause() {}, resume() {}, + write(text) { run.written.push(text); (run.ready ? run.accepted : run.dropped).push(text); }, + onData(listener) { emit = listener; return { dispose() {} }; }, + onExit(listener) { exits.push(listener); return { dispose() {} }; } + }; + }; + return { runs, spawn }; +} + +function slowOpenCodeOrchestration(t, options) { + const cli = slowOpenCode(options); + let orchestratorStarted = false; + const terminals = new TerminalManager(() => {}, clis(), undefined, undefined, true, (...args) => { + if (!orchestratorStarted) { orchestratorStarted = true; return spawner([])(...args); } + return cli.spawn(...args); + }); + t.after(() => terminals.disposeAll()); + const control = new AgentControlService(terminals); + const handler = new ScopedOrchestrationHandler(control); + const orchestrator = terminals.create({ provider: "codex", profile: "normal", cwd, position: at, role: "orchestrator" }); + return { cli, terminals, control, execute: (tool, args) => handler.execute(orchestrator.id, { tool, arguments: args }) }; +} + +test("spawn_agent: a slow-starting OpenCode gets its first prompt once its input field is up, exactly once", async (t) => { + const f = slowOpenCodeOrchestration(t, { startupMs: 150 }); + const spawned = f.execute("spawn_agent", { provider: "opencode", cwd, prompt: "do part A" }); + await new Promise((resolve) => setTimeout(resolve, 100)); + assert.equal(await state(spawned), "pending", "the PTY exists and the home screen text is drawn, but the field is not up"); + assert.deepEqual(f.cli.runs[0].written, []); + const result = await spawned; + assert.equal(typeof result.sessionId, "string"); + assert.deepEqual(f.cli.runs[0].accepted, ["do part A\r"], "the prompt reached the input field"); + assert.deepEqual(f.cli.runs[0].dropped, [], "nothing was typed while OpenCode started"); + + // send_to_agent to the running card goes in at once. + assert.deepEqual(await f.execute("send_to_agent", { sessionId: result.sessionId, prompt: "now part B" }), { sessionId: result.sessionId, sent: true }); + assert.deepEqual(f.cli.runs[0].accepted, ["do part A\r", "now part B\r"]); +}); + +test("spawn_agent: an OpenCode that exits before its input field is up fails the call with the reason", async (t) => { + const f = slowOpenCodeOrchestration(t, { ready: false }); + const spawned = f.execute("spawn_agent", { provider: "opencode", cwd, prompt: "never typed" }); + await tick(); + f.cli.runs[0].exit(1); + await assert.rejects(spawned, (error) => { + assert.match(error.bridgeError?.message ?? "", /was not delivered: The session/u); + return true; + }); + assert.deepEqual(f.cli.runs[0].written, []); +}); + +test("OpenCode that never shows its input field gets no prompt and a bounded reason", async (t) => { + const cli = slowOpenCode({ ready: false }); + const terminals = new TerminalManager(() => {}, clis(), undefined, undefined, true, cli.spawn); + t.after(() => terminals.disposeAll()); + const session = terminals.create({ provider: "opencode", profile: "normal", cwd, position: at }); + const started = Date.now(); + const result = await terminals.deliverInput(session.id, "never\r", 200); + assert.equal(result.delivered, false); + assert.match(result.reason, /did not show its input field within 200 ms; nothing was typed/u); + assert.ok(Date.now() - started < 1_000); + assert.deepEqual(cli.runs[0].written, [], "its home screen text alone is not readiness"); +}); + +test("OpenCode readiness: a cursor shown before the alternate screen does not count; one cut across chunks does", async (t) => { + const calls = []; + const terminals = new TerminalManager(() => {}, clis(), undefined, undefined, true, (command, args, options) => { + const pty = spawner(calls)(command, args, options); + return { ...pty, onData(listener) { calls.at(-1).emit = listener; return { dispose() {} }; } }; + }); + t.after(() => terminals.disposeAll()); + const session = terminals.create({ provider: "opencode", profile: "normal", cwd, position: at }); + const pending = terminals.deliverInput(session.id, "split\r", 2_000); + calls[0].emit("\x1b[?25h"); + calls[0].emit("\x1b[?104"); + calls[0].emit("9h\x1b[?25l drawing"); + await tick(); + assert.deepEqual(calls[0].written, []); + calls[0].emit("\x1b[?2"); + await tick(); + assert.deepEqual(calls[0].written, []); + calls[0].emit("5h"); + assert.deepEqual(await pending, { delivered: true }); + assert.deepEqual(calls[0].written, ["split\r"]); +}); + +test("OpenCode readiness: a lifecycle hook counts only after a grace, and sends made while it starts arrive in order", async (t) => { + const calls = []; + const terminals = new TerminalManager(() => {}, clis(), undefined, undefined, true, spawner(calls)); + t.after(() => terminals.disposeAll()); + const session = terminals.create({ provider: "opencode", profile: "normal", cwd, position: at }); + const first = terminals.deliverInput(session.id, "first\r", INPUT_READY_HOOK_GRACE_MS + 2_000); + const second = terminals.deliverInput(session.id, "second\r", INPUT_READY_HOOK_GRACE_MS + 2_000); + await tick(); + assert.deepEqual(calls[0].written, []); + const hookAt = Date.now(); + terminals.applyProviderSignal(session.id, { kind: "lifecycle", state: "idle", event: "session.created" }); + await new Promise((resolve) => setTimeout(resolve, 200)); + assert.deepEqual(calls[0].written, [], "a hook alone is not the input field"); + assert.deepEqual([await first, await second], [{ delivered: true }, { delivered: true }]); + assert.ok(Date.now() - hookAt >= INPUT_READY_HOOK_GRACE_MS - 20, "typed only after the grace"); + assert.deepEqual(calls[0].written, ["first\r", "second\r"]); +}); + +test("OpenCode readiness: a resumed card whose plugin reports before its input field is up gets the text on the cursor signal", async (t) => { + const calls = []; + const terminals = new TerminalManager(() => {}, clis(), undefined, undefined, true, (command, args, options) => { + const pty = spawner(calls)(command, args, options); + return { ...pty, onData(listener) { calls.at(-1).emit = listener; return { dispose() {} }; } }; + }); + t.after(() => terminals.disposeAll()); + const session = terminals.create({ provider: "opencode", profile: "normal", cwd, position: at }); + calls[0].emit("\x1b[?1049h\x1b[?25l"); + // The plugin reports idle on session.updated about 0.6 s before typed keys reach the prompt. + terminals.applyProviderSignal(session.id, { kind: "lifecycle", state: "idle", event: "session.updated" }); + const sent = terminals.deliverInput(session.id, "resumed\r", 5_000); + await new Promise((resolve) => setTimeout(resolve, 300)); + assert.deepEqual(calls[0].written, [], "not typed on the hook"); + const shownAt = Date.now(); + calls[0].emit("\x1b[?25h"); + assert.deepEqual(await sent, { delivered: true }); + assert.ok(Date.now() - shownAt < INPUT_READY_HOOK_GRACE_MS / 2, "typed on the cursor signal, not after the grace"); + assert.deepEqual(calls[0].written, ["resumed\r"]); +}); + +test("OpenCode readiness: a send cancelled while it waits behind another settles at once and is never typed", async (t) => { + const calls = []; + const terminals = new TerminalManager(() => {}, clis(), undefined, undefined, true, (command, args, options) => { + const pty = spawner(calls)(command, args, options); + return { ...pty, onData(listener) { calls.at(-1).emit = listener; return { dispose() {} }; } }; + }); + t.after(() => terminals.disposeAll()); + const session = terminals.create({ provider: "opencode", profile: "normal", cwd, position: at }); + const first = terminals.deliverInput(session.id, "first\r", 5_000); + const controller = new AbortController(); + const queued = terminals.deliverInput(session.id, "queued\r", 5_000, controller.signal); + const third = terminals.deliverInput(session.id, "third\r", 5_000); + await tick(); + const abortedAt = Date.now(); + controller.abort(); + assert.deepEqual(await queued, { delivered: false, reason: "The delivery was cancelled." }); + assert.ok(Date.now() - abortedAt < 200, "settled without waiting for the send ahead of it"); + calls[0].emit("\x1b[?1049h\x1b[?25h"); + assert.deepEqual([await first, await third], [{ delivered: true }, { delivered: true }]); + assert.deepEqual(calls[0].written, ["first\r", "third\r"]); +}); + +test("OpenCode readiness is not awaited on Windows, whose pseudo console may not pass the cursor sequences", async (t) => { + const platform = Object.getOwnPropertyDescriptor(process, "platform"); + const calls = []; + const terminals = new TerminalManager(() => {}, clis(), undefined, undefined, true, spawner(calls)); + t.after(() => terminals.disposeAll()); + const session = terminals.create({ provider: "opencode", profile: "normal", cwd, position: at }); + Object.defineProperty(process, "platform", { ...platform, value: "win32" }); + try { + assert.deepEqual(await terminals.deliverInput(session.id, "windows\r", 200), { delivered: true }); + } finally { + Object.defineProperty(process, "platform", platform); + } + assert.deepEqual(calls[0].written, ["windows\r"]); +}); + +test("spawn_agent and send_to_agent wait for delivery within a budget under the MCP client's deadline", async (t) => { + assert.ok(AGENT_TOOL_DELIVERY_MS <= 45_000); + const f = slowOpenCodeOrchestration(t, { ready: false }); + const waits = []; + const { control } = f; + const spawn = control.spawn.bind(control); + control.spawn = (request, signal, promptWaitMs) => { waits.push(["spawn", promptWaitMs]); return spawn(request, signal, 150); }; + const send = control.send.bind(control); + control.send = (id, text, submit, signal, waitMs) => { waits.push(["send", waitMs]); return send(id, text, submit, signal, 150); }; + await assert.rejects(f.execute("spawn_agent", { provider: "opencode", cwd, prompt: "late" }), (error) => { + assert.match(error.bridgeError?.message ?? "", /did not show its input field within 150 ms; nothing was typed/u); + return true; + }); + const [child] = f.terminals.listMetadata().filter((session) => session.role === "subagent"); + await assert.rejects(f.execute("send_to_agent", { sessionId: child.id, prompt: "again" })); + assert.equal(waits[0][0], "spawn"); + assert.ok(waits[0][1] > 0 && waits[0][1] <= AGENT_TOOL_DELIVERY_MS, String(waits[0][1])); + assert.deepEqual(waits[1], ["send", AGENT_TOOL_DELIVERY_MS]); +}); + +test("OpenCode readiness: cancellation and a restart while it starts never deliver the queued text", async (t) => { + const cli = slowOpenCode({ startupMs: 80 }); + const terminals = new TerminalManager(() => {}, clis(), undefined, undefined, true, cli.spawn); + t.after(() => terminals.disposeAll()); + const session = terminals.create({ provider: "opencode", profile: "normal", cwd, position: at }); + const controller = new AbortController(); + const canceled = terminals.deliverInput(session.id, "cancel\r", 2_000, controller.signal); + const replaced = terminals.deliverInput(session.id, "old launch\r", 2_000); + controller.abort(); + assert.deepEqual(await canceled, { delivered: false, reason: "The delivery was cancelled." }); + cli.runs[0].exit(0); + terminals.restart(session.id); + assert.equal((await replaced).delivered, false); + await new Promise((resolve) => setTimeout(resolve, 150)); + assert.ok(cli.runs[1].ready, "the new launch became ready"); + assert.ok(cli.runs.every((run) => run.written.length === 0), "neither launch got the text"); +}); + +test("Claude and Codex still get input as soon as their PTY exists", async (t) => { + const calls = []; + const terminals = new TerminalManager(() => {}, clis(), undefined, undefined, true, spawner(calls)); + t.after(() => terminals.disposeAll()); + for (const provider of ["claude", "codex"]) { + const session = terminals.create({ provider, profile: "normal", cwd, position: at }); + assert.deepEqual(await terminals.deliverInput(session.id, `${provider} task\r`, 200), { delivered: true }); + } + assert.deepEqual(calls.map((call) => call.written), [["claude task\r"], ["codex task\r"]]); +}); diff --git a/tests/material-handoff-service.test.mjs b/tests/material-handoff-service.test.mjs index 141e452a4..8ce26e6e3 100644 --- a/tests/material-handoff-service.test.mjs +++ b/tests/material-handoff-service.test.mjs @@ -979,3 +979,19 @@ test("a delivered handoff says when the outcome could not be saved", async () => await rm(blocked, { recursive: true }); }); }); + +test("a handoff waits for the agent's CLI to take input instead of failing at once", async () => { + await withHandoffs(async ({ handoffs, terminal, draft }) => { + terminal.add({ provider: "claude", id: "s1" }); + terminal.tui = claudeTui(terminal); + const waits = []; + const deliverInput = terminal.deliverInput.bind(terminal); + terminal.deliverInput = async (id, data, waitMs) => { + waits.push(waitMs); + return deliverInput(id, data); + }; + assert.equal((await handoffs.send(draft("s1"))).handoff.delivery.state, "submitted"); + assert.ok(waits.length >= 2, "the paste and its Enter"); + assert.ok(waits.every((waitMs) => waitMs >= 10_000), `waits: ${waits.join(", ")}`); + }); +}); diff --git a/tests/opencode-result.test.mjs b/tests/opencode-result.test.mjs index 1355f842a..a6f77225e 100644 --- a/tests/opencode-result.test.mjs +++ b/tests/opencode-result.test.mjs @@ -7,10 +7,12 @@ import test from "node:test"; import { AGENT_RUNTIME_ENV, CAPTURE_RESULT_ENV, MAX_RESULT_CHARS } from "../src/agent-runtime/runtime-protocol.mjs"; import { finalAnswer } from "../src/agent-runtime/opencode-final-answer.mjs"; import { RuntimeGateway } from "../src/main/services/agent-runtime/RuntimeGateway.ts"; +import { applyRuntimeSignal } from "../src/main/services/agent-runtime/applyRuntimeSignal.ts"; +import { LaunchPipeline } from "../src/main/services/LaunchPipeline.ts"; import { TerminalManager } from "../src/main/services/TerminalManager.ts"; import { AgentControlService } from "../src/main/services/AgentControlService.ts"; import { ScopedOrchestrationHandler } from "../src/main/services/agent-browser/OrchestrationTools.ts"; -import { availableRegistry, fakeSpawner } from "./helpers/terminal.mjs"; +import { availableRegistry, fakeSpawner, inputReadySpawner, openCodeInputReady } from "./helpers/terminal.mjs"; const POSIX = { skip: process.platform === "win32" ? "Unix socket test; the Windows pipe transport has its own suite." : false }; @@ -55,9 +57,9 @@ await hooks.event({ event: { type: "session.status", properties: { sessionID: "s await hooks.event({ event: { type: "session.idle", properties: { sessionID: "ses_root" } } }); `; -function runFakeOpenCode(capability, answer, captureResult) { +function runFakeOpenCode(capability, answer, captureResult, script = FAKE_OPENCODE, extraEnv = {}, onCapture = null) { const env = { - ...process.env, + ...process.env, ...extraEnv, PLUGIN_URL: new URL("../src/agent-runtime/opencode-plugin.mjs", import.meta.url).href, FAKE_ANSWER: JSON.stringify(answer), [AGENT_RUNTIME_ENV.address]: capability.address, @@ -67,35 +69,43 @@ function runFakeOpenCode(capability, answer, captureResult) { }; if (captureResult) env[CAPTURE_RESULT_ENV] = "1"; else delete env[CAPTURE_RESULT_ENV]; - const child = spawn(process.execPath, ["--input-type=module", "-e", FAKE_OPENCODE], { env, stdio: ["ignore", "ignore", "pipe"] }); + const child = spawn(process.execPath, ["--input-type=module", "-e", script], { env, stdio: ["pipe", "pipe", "pipe"] }); + const timeout = setTimeout(() => child.kill(), 5_000); let stderr = ""; child.stderr.on("data", (chunk) => { stderr += chunk; }); - return new Promise((resolve) => child.on("close", (code) => resolve({ code, stderr }))); + return new Promise((resolve, reject) => { + // The script writes to stdout when its SDK read starts and waits on stdin until onCapture has run. + if (onCapture) child.stdout.once("data", () => { + Promise.resolve().then(onCapture).then(() => child.stdin.end("continue"), (error) => { child.kill(); reject(error); }); + }); + child.on("error", (error) => { clearTimeout(timeout); reject(error); }); + child.on("close", (code) => { clearTimeout(timeout); resolve({ code, stderr }); }); + }); } -async function setup(t) { +async function setupTerminals(t) { const root = await realpath(await mkdtemp(join(tmpdir(), "ctty-ocr-"))); t.after(() => rm(root, { recursive: true, force: true })); const calls = []; - const terminals = new TerminalManager(() => undefined, availableRegistry(), undefined, undefined, true, fakeSpawner(calls)); + const terminals = new TerminalManager(() => undefined, availableRegistry(), undefined, undefined, true, inputReadySpawner(calls)); t.after(() => terminals.disposeAll()); - const runtimeDirectory = await mkdtemp(join(tmpdir(), "ctty-rt-")); - t.after(() => rm(runtimeDirectory, { recursive: true, force: true })); - // What main/index.ts does with each runtime signal. - const gateway = new RuntimeGateway({ - runtimeDirectory, - onSignal: (id, signal) => { - terminals.applyProviderSignal(id, { kind: "lifecycle", state: signal.state }); - if (signal.result) terminals.recordAnswer(id, signal.result); - } - }); - await gateway.start(); - t.after(() => gateway.close()); const orchestrator = terminals.create({ provider: "codex", cwd: root, profile: "normal", position: { x: 0, y: 0 }, role: "orchestrator" }); const control = new AgentControlService(terminals, { waitTiming: { checkMs: 5, settleMs: 300, quietMs: 2_000 } }); const handler = new ScopedOrchestrationHandler(control); const child = await handler.execute(orchestrator.id, { id: "1", tool: "spawn_agent", arguments: { provider: "opencode", cwd: root } }); - return { root, calls, terminals, gateway, handler, orchestrator, childId: child.sessionId }; + return { root, calls, terminals, control, handler, orchestrator, childId: child.sessionId }; +} + +async function setup(t) { + const fixture = await setupTerminals(t); + const { terminals } = fixture; + const runtimeDirectory = await mkdtemp(join(tmpdir(), "ctty-rt-")); + t.after(() => rm(runtimeDirectory, { recursive: true, force: true })); + // The same glue main/index.ts runs for each runtime signal. + const gateway = new RuntimeGateway({ runtimeDirectory, onSignal: (id, signal) => applyRuntimeSignal(terminals, id, signal) }); + await gateway.start(); + t.after(() => gateway.close()); + return { ...fixture, gateway }; } test("an OpenCode subagent's final reply reaches wait_for_agent and get_agent_result, masked", POSIX, async (t) => { @@ -147,11 +157,14 @@ test("an ordinary OpenCode card (not a subagent) never keeps an answer", async ( test("wait_for_agent ignores an idle before the prompt's turn and returns that turn's answer", async (t) => { const root = await realpath(await mkdtemp(join(tmpdir(), "ctty-ocr-"))); t.after(() => rm(root, { recursive: true, force: true })); - const terminals = new TerminalManager(() => undefined, availableRegistry(), undefined, undefined, true, fakeSpawner([])); + const calls = []; + const terminals = new TerminalManager(() => undefined, availableRegistry(), undefined, undefined, true, fakeSpawner(calls)); t.after(() => terminals.disposeAll()); const orchestrator = terminals.create({ provider: "codex", cwd: root, profile: "normal", position: { x: 0, y: 0 }, role: "orchestrator" }); const handler = new ScopedOrchestrationHandler(new AgentControlService(terminals, { waitTiming: { checkMs: 5, settleMs: 10, quietMs: 60_000 } })); - const child = await handler.execute(orchestrator.id, { id: "1", tool: "spawn_agent", arguments: { provider: "opencode", cwd: root, prompt: "do part A" } }); + const spawning = handler.execute(orchestrator.id, { id: "1", tool: "spawn_agent", arguments: { provider: "opencode", cwd: root, prompt: "do part A" } }); + await openCodeInputReady(calls, 1); + const child = await spawning; const wait = (timeoutSeconds) => handler.execute(orchestrator.id, { id: "w", tool: "wait_for_agent", arguments: { sessionId: child.sessionId, timeoutSeconds } }); // OpenCode's session.created reports idle before the prompt's turn starts. terminals.applyProviderSignal(child.sessionId, { kind: "lifecycle", state: "idle", event: "session.created" }); @@ -167,3 +180,318 @@ test("wait_for_agent ignores an idle before the prompt's turn and returns that t assert.equal(done.reason, "idle"); assert.deepEqual(done.answer, { text: "part A done", truncated: false }); }); + +test("OpenCode turn correlation rejects preceding answers and continuations across submitted prompts", async (t) => { + const { terminals, childId } = await setupTerminals(t); + const apply = (state, event, turnId, answer) => { + const current = terminals.applyProviderSignal(childId, { kind: "lifecycle", state, event, requestId: turnId }); + if (current && answer) terminals.recordAnswer(childId, { text: answer, truncated: false }, { turnId }); + return current; + }; + assert.equal(apply("working", "session.status:busy", "turn1"), true); + assert.equal(apply("idle", "session.idle", "turn1", "old answer"), true); + assert.equal(terminals.answer(childId).text, "old answer"); + assert.deepEqual(await terminals.deliverInput(childId, "new task\r"), { delivered: true }); + assert.equal(terminals.answer(childId), null, "a submitted prompt drops the previous answer"); + assert.equal(apply("idle", "session.idle", "turn1", "delayed old answer"), false); + assert.equal(apply("working", "permission.replied", "turn1"), false, "an old continuation cannot acknowledge the new prompt"); + assert.equal(terminals.getMetadata(childId).status, "working", "but it still moves the status of the turn that runs"); + terminals.recordAnswer(childId, { text: "uncorrelated old answer", truncated: false }); + assert.equal(terminals.answer(childId), null); + assert.equal(terminals.turnProgress(childId).turnStartedSincePrompt, false); + assert.equal(apply("working", "session.status:busy", "turn2"), true); + assert.equal(apply("idle", "session.idle", "turn1", "late old after new start"), false); + assert.equal(terminals.getMetadata(childId).status, "working", "a late signal of an earlier turn changes nothing"); + assert.equal(apply("needs_approval", "permission.asked", "turn2"), true); + const generation = terminals.sessions.get(childId).inputGeneration ?? 0; + terminals.input(childId, "y\r"); + assert.equal(terminals.sessions.get(childId).inputGeneration ?? 0, generation, "an approval answer is not a new task"); + assert.equal(apply("working", "permission.replied", "turn2"), true); + assert.equal(apply("working", "session.status:retry", "turn2"), true); + assert.equal(apply("idle", "session.idle", "turn2", "new answer"), true); + assert.equal(terminals.answer(childId).text, "new answer"); + apply("working", "session.status:busy", "turn3"); + await terminals.deliverInput(childId, "interrupt current work\r"); + const starts = terminals.sessions.get(childId).turnStarts; + assert.equal(apply("working", "session.status:busy", "turn4"), true); + assert.equal(terminals.sessions.get(childId).turnStarts, starts + 1, "a new turn id is a new turn even while the previous one was working"); + assert.equal(terminals.turnProgress(childId).turnStartedSincePrompt, true); + assert.equal(apply("idle", "session.idle", "turn3", "interrupted answer"), false); + assert.equal(terminals.answer(childId), null); +}); + +test("with status hooks off the plugin reports nothing and the wait says the answer cannot be captured", POSIX, async (t) => { + const { terminals, gateway, handler, orchestrator, childId } = await setup(t); + terminals.setLifecycleHooksEnabled(false); + const signals = []; + const apply = terminals.applyProviderSignal.bind(terminals); + terminals.applyProviderSignal = (id, signal, ...args) => { signals.push(signal); return apply(id, signal, ...args); }; + // A lease that exists for another reason (the decision hook) still carries no lifecycle report and no answer. + const capability = gateway.registerSession(childId, "opencode", true); + const run = await runFakeOpenCode(capability, "never captured", true, FAKE_OPENCODE, { CANVASTTY_LIFECYCLE_HOOKS_ENABLED: "0" }); + assert.equal(run.code, 0, run.stderr); + assert.deepEqual(signals, []); + assert.equal(terminals.getMetadata(childId).status, "unavailable"); + assert.equal(terminals.answer(childId), null); + const waited = await handler.execute(orchestrator.id, { id: "w", tool: "wait_for_agent", arguments: { sessionId: childId, timeoutSeconds: 5 } }); + assert.equal(waited.reason, "quiet"); + assert.equal(waited.answer, undefined); + assert.equal(waited.answerUnavailable.reason, "status_hooks_off"); + assert.match(waited.answerUnavailable.message, /status hooks are off[\s\S]*Read its output/u); + const result = await handler.execute(orchestrator.id, { id: "r", tool: "get_agent_result", arguments: { sessionId: childId } }); + assert.equal(result.answer, undefined); + assert.deepEqual(result.answerUnavailable, waited.answerUnavailable); + // Turning them on again applies to agents started or restarted afterwards, not to this launch. + terminals.setLifecycleHooksEnabled(true); + assert.equal(terminals.applyProviderSignal(childId, { kind: "lifecycle", state: "idle", event: "session.idle", requestId: "t" }), false); + const later = await handler.execute(orchestrator.id, { id: "r2", tool: "get_agent_result", arguments: { sessionId: childId } }); + assert.equal(later.answerUnavailable?.reason, "status_hooks_off"); +}); + +test("OpenCode async final answer keeps its ending turn and drops a read superseded by a new turn", POSIX, async (t) => { + const { terminals, gateway, childId } = await setup(t); + const capability = gateway.registerSession(childId, "opencode", true); + const signals = []; + const original = terminals.applyProviderSignal.bind(terminals); + terminals.applyProviderSignal = (id, signal, ...args) => { signals.push(signal); return original(id, signal, ...args); }; + const script = ` + const { CanvasTTYLifecycle } = await import(process.env.PLUGIN_URL); + let resolveOld, started; const capturing = new Promise((done) => started = done); + let reads = 0; + const client = { session: { messages: async () => { + if (reads++ === 0) { started(); return new Promise((done) => resolveOld = done); } + return { data: [{ info: { role: "assistant" }, parts: [{ type: "text", text: "second answer" }] }] }; + } } }; + const hooks = await CanvasTTYLifecycle({ client }); + const event = (type, rest = {}) => hooks.event({ event: { type, properties: { sessionID: "root", ...rest } } }); + await event("session.created", { info: { id: "root" } }); + await event("session.status", { status: { type: "busy" } }); + await event("session.status", { status: { type: "idle" } }); + const old = event("session.idle"); await capturing; + await event("session.status", { status: { type: "busy" } }); + resolveOld({ data: [{ info: { role: "assistant" }, parts: [{ type: "text", text: "old delayed answer" }] }] }); + await old; + await event("permission.asked"); await event("permission.replied"); + await event("session.status", { status: { type: "retry" } }); + await event("session.status", { status: { type: "idle" } }); + await event("session.idle"); + `; + const run = await runFakeOpenCode(capability, "unused", true, script); + assert.equal(run.code, 0, run.stderr); + const starts = signals.filter((signal) => signal.event === "session.status:busy"); + assert.equal(starts.length, 2); + assert.notEqual(starts[0].requestId, starts[1].requestId, "each turn has its own id"); + assert.equal(signals.filter((signal) => signal.event === "session.status:idle").length, 0, "no idle before the answer was read"); + const ends = signals.filter((signal) => signal.event === "session.idle"); + assert.equal(ends.length, 1, "a superseded answer read reports neither idle nor an answer"); + assert.equal(ends[0].requestId, starts[1].requestId); + assert.equal(signals.find((signal) => signal.event === "permission.replied").requestId, starts[1].requestId); + assert.equal(signals.find((signal) => signal.event === "session.status:retry").requestId, starts[1].requestId); + assert.equal(terminals.answer(childId).text, "second answer"); +}); + +test("an answer read that finishes after the next prompt was submitted is not that prompt's answer", POSIX, async (t) => { + const { terminals, gateway, handler, orchestrator, childId } = await setup(t); + const capability = gateway.registerSession(childId, "opencode", true); + const script = ` + const { CanvasTTYLifecycle } = await import(process.env.PLUGIN_URL); + const client = { session: { messages: async () => { + process.stdout.write("capturing"); + await new Promise((done) => process.stdin.once("data", done)); + return { data: [{ info: { role: "assistant" }, parts: [{ type: "text", text: "late preceding answer" }] }] }; + } } }; + const hooks = await CanvasTTYLifecycle({ client }); + const event = (type, rest = {}) => hooks.event({ event: { type, properties: { sessionID: "root", ...rest } } }); + await event("session.created", { info: { id: "root" } }); + await event("session.status", { status: { type: "busy" } }); + await event("session.idle"); + `; + const run = await runFakeOpenCode(capability, "unused", true, script, {}, () => terminals.deliverInput(childId, "next task\r")); + assert.equal(run.code, 0, run.stderr); + const result = await handler.execute(orchestrator.id, { id: "poll", tool: "get_agent_result", arguments: { sessionId: childId } }); + assert.equal(result.answer, undefined); + assert.equal(terminals.turnProgress(childId).turnStartedSincePrompt, false, "the late idle does not acknowledge the next prompt"); +}); + +test("approval continuations and repeated working events stay in the same turn", async (t) => { + const { terminals, childId } = await setupTerminals(t); + const apply = (state, event, turnId) => terminals.applyProviderSignal(childId, { kind: "lifecycle", state, event, requestId: turnId }); + assert.equal(apply("working", "session.status:busy", "turn1"), true); + const starts = terminals.sessions.get(childId).turnStarts; + assert.equal(apply("working", "session.status:busy", "turn1"), true); + assert.equal(apply("working", "session.status:retry", "turn1"), true); + assert.equal(terminals.sessions.get(childId).turnStarts, starts, "a repeated working or retry is not another start"); + assert.equal(apply("needs_approval", "permission.asked", "turn1"), true); + assert.equal(terminals.getMetadata(childId).status, "needs_approval"); + const generation = terminals.sessions.get(childId).inputGeneration ?? 0; + await terminals.deliverInput(childId, "y\r"); + assert.equal(terminals.sessions.get(childId).inputGeneration ?? 0, generation, "an approval answer keeps the current turn"); + assert.equal(apply("working", "permission.replied", "turn1"), true); + assert.equal(terminals.turnProgress(childId).turnStartedSincePrompt, true, "the continuation acknowledges the approval"); + const continued = terminals.sessions.get(childId).turnStarts; + assert.equal(apply("working", "session.status:retry", "turn1"), true); + assert.equal(terminals.sessions.get(childId).turnStarts, continued); + assert.equal(apply("idle", "session.idle", "turn1"), true); + terminals.recordAnswer(childId, { text: "answer after approval", truncated: false }, { turnId: "turn1" }); + assert.equal(terminals.answer(childId).text, "answer after approval"); + assert.equal(terminals.getMetadata(childId).status, "idle"); + terminals.input(childId, "next task\r"); + assert.equal(terminals.sessions.get(childId).inputGeneration, generation + 1); + assert.equal(apply("working", "permission.replied", "turn1"), false); +}); + +test("a result-capturing OpenCode does not report idle while its final answer is still being read", POSIX, async (t) => { + const { terminals, gateway, handler, orchestrator, childId } = await setup(t); + const capability = gateway.registerSession(childId, "opencode", true); + const script = ` + const { CanvasTTYLifecycle } = await import(process.env.PLUGIN_URL); + const client = { session: { messages: async () => { + process.stdout.write("capturing"); await new Promise((done) => process.stdin.once("data", done)); + return { data: [{ info: { role: "assistant" }, parts: [{ type: "text", text: "captured after delay" }] }] }; + } } }; + const hooks = await CanvasTTYLifecycle({ client }); + const event = (type, rest = {}) => hooks.event({ event: { type, properties: { sessionID: "root", ...rest } } }); + await event("session.created", { info: { id: "root" } }); + await event("session.status", { status: { type: "busy" } }); + await event("session.status", { status: { type: "idle" } }); + await event("session.idle"); + `; + const run = await runFakeOpenCode(capability, "unused", true, script, {}, () => { + assert.equal(terminals.getMetadata(childId).status, "working"); + assert.equal(terminals.answer(childId), null); + }); + assert.equal(run.code, 0, run.stderr); + const waited = await handler.execute(orchestrator.id, { id: "w", tool: "wait_for_agent", arguments: { sessionId: childId, timeoutSeconds: 5 } }); + assert.equal(waited.reason, "idle"); + assert.equal(waited.answer.text, "captured after delay"); +}); + +for (const failure of ["error", "timeout"]) test(`a captured OpenCode turn still ends, without an answer, after an SDK read ${failure}`, POSIX, async (t) => { + const { terminals, gateway, childId } = await setup(t); + const accepted = []; + const apply = terminals.applyProviderSignal.bind(terminals); + terminals.applyProviderSignal = (id, signal, ...args) => { accepted.push(signal); return apply(id, signal, ...args); }; + const capability = gateway.registerSession(childId, "opencode", true); + const script = ` + const { CanvasTTYLifecycle } = await import(process.env.PLUGIN_URL); + const client = { session: { messages: async () => { ${failure === "error" ? 'throw new Error("fixture failed");' : "return new Promise(() => {});"} } } }; + const hooks = await CanvasTTYLifecycle({ client }); + const event = (type, rest = {}) => hooks.event({ event: { type, properties: { sessionID: "root", ...rest } } }); + await event("session.created", { info: { id: "root" } }); await event("session.status", { status: { type: "busy" } }); + await event("session.status", { status: { type: "idle" } }); await event("session.idle"); + `; + const run = await runFakeOpenCode(capability, "unused", true, script); + assert.equal(run.code, 0, run.stderr); + assert.equal(accepted.filter((signal) => signal.event === "session.status:idle").length, 0); + assert.equal(accepted.filter((signal) => signal.event === "session.idle").length, 1, "the turn end is still reported"); + assert.equal(terminals.answer(childId), null); + assert.equal(terminals.getMetadata(childId).status, "idle"); +}); + +/** Feeds one runtime signal through the glue main uses, as the gateway would hand it over. */ +const signalFrom = (terminals, id) => (state, event, turnId, answer) => applyRuntimeSignal(terminals, id, + { state, event, turnId, ...(answer ? { result: { text: answer, truncated: false } } : {}) }); + +test("a person's Enter in the middle of a turn still lets that turn end the wait, without an answer", async (t) => { + const { terminals, handler, orchestrator, childId } = await setupTerminals(t); + const signal = signalFrom(terminals, childId); + assert.deepEqual(await terminals.deliverInput(childId, "orchestrator task\r"), { delivered: true }); + signal("working", "session.status:busy", "turn1"); + // The person presses Enter in the subagent's card while it works: the turn's answer is no longer for that prompt. + terminals.input(childId, "\r"); + const waiting = handler.execute(orchestrator.id, { id: "w", tool: "wait_for_agent", arguments: { sessionId: childId, timeoutSeconds: 5 } }); + setTimeout(() => signal("idle", "session.idle", "turn1", "answer to the earlier prompt"), 20); + const waited = await waiting; + assert.equal(waited.reason, "idle", "the wait does not hang until its timeout"); + assert.equal(waited.status, "idle"); + assert.equal(waited.answer, undefined); + assert.equal(waited.answerUnavailable, undefined, "status hooks are on: nothing says the answer cannot be captured"); +}); + +test("an answer that arrives as the agent exits is kept when it belongs to the current turn", async (t) => { + const { calls, terminals, handler, orchestrator, childId } = await setupTerminals(t); + const signal = signalFrom(terminals, childId); + await terminals.deliverInput(childId, "last task\r"); + signal("working", "session.status:busy", "turn1"); + calls.at(-1).process.emitExit(0); + assert.equal(terminals.getMetadata(childId).status, "done"); + signal("idle", "session.idle", "turn1", "final words"); + assert.equal(terminals.getMetadata(childId).status, "done", "the exit stays"); + const result = await handler.execute(orchestrator.id, { id: "r", tool: "get_agent_result", arguments: { sessionId: childId } }); + assert.equal(result.state, "done"); + assert.equal(result.answer?.text, "final words"); + // A late answer of an earlier turn still does not count after the exit. + signal("idle", "session.idle", "turn0", "stale"); + assert.equal(terminals.answer(childId)?.text, "final words"); +}); + +test("an Enter with nothing typed in an idle card keeps the stored answer", async (t) => { + const { terminals, childId } = await setupTerminals(t); + const signal = signalFrom(terminals, childId); + await terminals.deliverInput(childId, "task\r"); + signal("working", "session.status:busy", "turn1"); + signal("idle", "session.idle", "turn1", "kept answer"); + const generation = terminals.sessions.get(childId).inputGeneration; + terminals.input(childId, "\r"); + assert.equal(terminals.answer(childId)?.text, "kept answer", "an empty Enter submits nothing"); + assert.equal(terminals.sessions.get(childId).inputGeneration, generation); + // Typed key by key, as a person types: the Enter after it submits a new task. + for (const key of ["n", "e", "x", "t"]) terminals.input(childId, key); + terminals.input(childId, "\r"); + assert.equal(terminals.answer(childId), null); + assert.equal(terminals.sessions.get(childId).inputGeneration, generation + 1); +}); + +test("every input write goes through writeInput, which keeps the turn and answer bookkeeping", async (t) => { + const { terminals, childId } = await setupTerminals(t); + const signal = signalFrom(terminals, childId); + const writes = []; + const writeInput = terminals.writeInput; + assert.equal(typeof writeInput, "function", "one write path for person and agent input"); + terminals.writeInput = function (...args) { writes.push(args[3]); return writeInput.apply(this, args); }; + await terminals.deliverInput(childId, "task\r"); + terminals.input(childId, "x"); + assert.deepEqual(writes, ["task\r", "x"], "input() and deliverInput write through it"); + terminals.writeInput = writeInput; + signal("working", "session.status:busy", "turn1"); + signal("idle", "session.idle", "turn1", "kept answer"); + // A writer that held input and replays it later (outside input()) uses the same method: its Enter submits a task. + const session = terminals.sessions.get(childId); + const generation = session.inputGeneration; + for (const key of ["n", "e", "x", "t", "\r"]) assert.equal(terminals.writeInput(childId, session, session.process, key), true); + assert.equal(terminals.answer(childId), null, "the replayed prompt drops the earlier answer"); + assert.equal(session.inputGeneration, generation + 1); + // Never into a process the card no longer runs. + assert.equal(terminals.writeInput(childId, session, { write: () => assert.fail("stale process") }, "late\r"), false); + assert.equal(session.inputGeneration, generation + 1); +}); + +for (const [atCreate, beforeSpawn] of [[false, true], [true, false]]) { + test(`a deferred launch takes status hooks as they are when it is planned (${atCreate ? "on" : "off"} at create, ${beforeSpawn ? "on" : "off"} at launch)`, async (t) => { + const root = await realpath(await mkdtemp(join(tmpdir(), "ctty-ocr-deferred-"))); + t.after(() => rm(root, { recursive: true, force: true })); + let release; + const prepared = new Promise((resolve) => { release = resolve; }); + const pipeline = new LaunchPipeline({ + contributors: () => [{ pluginId: "fixture.hold", pluginName: "Hold", serviceId: "launch", secrets: false, + launch: { fields: [{ key: "on", label: "On", kind: "boolean", default: true }] } }], + call: async () => { await prepared; return {}; }, + secret: async () => null, + runsRoot: join(root, "runs"), + timeoutMs: 4_000 + }); + const calls = []; + const terminals = new TerminalManager(() => undefined, availableRegistry(), undefined, undefined, atCreate, fakeSpawner(calls)); + t.after(() => { release(); terminals.disposeAll(); }); + terminals.configureLaunchPipeline(pipeline); + const card = terminals.create({ provider: "opencode", cwd: root, profile: "normal", position: { x: 0, y: 0 }, + launchOptions: { "fixture.hold": { on: true } } }, { captureResult: true }); + terminals.setLifecycleHooksEnabled(beforeSpawn); + release(); + const deadline = Date.now() + 4_000; + while (calls.length === 0 && Date.now() < deadline) await new Promise((resolve) => setTimeout(resolve, 5)); + assert.equal(calls.length, 1, "the launch spawned"); + assert.equal(terminals.answerUnavailable(card.id), beforeSpawn ? null : "status_hooks_off"); + assert.equal(terminals.applyProviderSignal(card.id, { kind: "lifecycle", state: "working", event: "session.status:busy", requestId: "t1" }), beforeSpawn); + }); +} diff --git a/tests/orchestration-gateway.test.mjs b/tests/orchestration-gateway.test.mjs index e2f0a524a..ace1c16c2 100644 --- a/tests/orchestration-gateway.test.mjs +++ b/tests/orchestration-gateway.test.mjs @@ -370,6 +370,7 @@ test("a spawn_agent canceled while it was starting closes the agent it created", const control = { status: () => ({ role: "orchestrator", provider: "codex" }), profileFor: () => ({ profile: "normal", inherited: true }), + subagentLaunchOptions: (_parent, _provider, launchOptions) => ({ launchOptions, sources: {} }), spawn: async () => { controller.abort(); return { id: "child-1", provider: "codex", status: "running", title: "worker" }; diff --git a/tests/orchestration-providers-wait.test.mjs b/tests/orchestration-providers-wait.test.mjs index ebe189141..2513b9878 100644 --- a/tests/orchestration-providers-wait.test.mjs +++ b/tests/orchestration-providers-wait.test.mjs @@ -2,6 +2,8 @@ import assert from "node:assert/strict"; import test from "node:test"; import { AGENT_PROVIDER_IDS, + CLIENT_SAFE_AGENT_WAIT_SECONDS, + DEFAULT_AGENT_WAIT_SECONDS, MAX_AGENT_WAIT_SECONDS, ORCHESTRATION_TOOL_DEFINITIONS, ORCHESTRATION_TOOL_NAMES, @@ -22,10 +24,14 @@ test("the catalog lists list_providers first and wait_for_agent with a bounded t assert.deepEqual(definition("list_providers").inputSchema, { type: "object", properties: {}, required: [], additionalProperties: false }); const wait = definition("wait_for_agent").inputSchema; assert.deepEqual(wait.required, ["sessionId"]); - assert.equal(wait.properties.timeoutSeconds.maximum, 600); - assert.equal(MAX_AGENT_WAIT_SECONDS, 600); - assert.equal(validateOrchestrationArguments("wait_for_agent", { sessionId: "a", timeoutSeconds: 600 }).ok, true); - assert.match(validateOrchestrationArguments("wait_for_agent", { sessionId: "a", timeoutSeconds: 601 }).error, /above the maximum/u); + assert.equal(wait.properties.timeoutSeconds.maximum, 100); + assert.equal(MAX_AGENT_WAIT_SECONDS, 100); + assert.ok(DEFAULT_AGENT_WAIT_SECONDS <= CLIENT_SAFE_AGENT_WAIT_SECONDS && CLIENT_SAFE_AGENT_WAIT_SECONDS < 60, + "a call without timeoutSeconds, and every call from a 60-second MCP client, ends before that client gives up"); + assert.equal(validateOrchestrationArguments("wait_for_agent", { sessionId: "a", timeoutSeconds: 100 }).ok, true); + assert.match(validateOrchestrationArguments("wait_for_agent", { sessionId: "a", timeoutSeconds: 101 }).error, /above the maximum/u); + assert.match(definition("wait_for_agent").description, /at most 100; unless this session is Claude Code, a call returns after at most 50 seconds/u); + assert.match(definition("wait_for_agent").description, /A timeout means it is still running: read output as its progress and call wait_for_agent again/u); assert.equal(validateOrchestrationArguments("list_providers", {}).ok, true); assert.match(validateOrchestrationArguments("list_providers", { x: 1 }).error, /Unexpected argument/u); assert.match(definition("list_providers").description, /Never search the filesystem/u); @@ -217,7 +223,7 @@ test("wait_for_agent stops at once when the call is canceled", async () => { const controller = new AbortController(); const started = Date.now(); setTimeout(() => controller.abort(), 20); - await assert.rejects(call(handler, "wait_for_agent", { sessionId: "child", timeoutSeconds: 600 }, controller.signal), + await assert.rejects(call(handler, "wait_for_agent", { sessionId: "child", timeoutSeconds: 100 }, controller.signal), (error) => error.bridgeError?.code === "CANCELED"); assert.ok(Date.now() - started < 1_000); const aborted = new AbortController(); @@ -261,7 +267,7 @@ test("after a prompt, an idle before the turn starts does not end wait_for_agent const { tmpdir } = await import("node:os"); const { join } = await import("node:path"); const { TerminalManager } = await import("../src/main/services/TerminalManager.ts"); - const { availableRegistry, fakeSpawner } = await import("./helpers/terminal.mjs"); + const { availableRegistry, fakeSpawner, openCodeInputReady } = await import("./helpers/terminal.mjs"); const root = await realpath(await mkdtemp(join(tmpdir(), "ctty-turn-"))); t.after(() => rm(root, { recursive: true, force: true })); const calls = []; @@ -277,7 +283,9 @@ test("after a prompt, an idle before the turn starts does not end wait_for_agent terminals.applyProviderSignal(idleChild.sessionId, { kind: "lifecycle", state: "idle" }); assert.equal((await handler.execute(orchestrator.id, { id: "i", tool: "wait_for_agent", arguments: { sessionId: idleChild.sessionId } })).reason, "idle"); - const child = await handler.execute(orchestrator.id, { id: "1", tool: "spawn_agent", arguments: { provider: "opencode", cwd: root, prompt: "do part A" } }); + const spawning = handler.execute(orchestrator.id, { id: "1", tool: "spawn_agent", arguments: { provider: "opencode", cwd: root, prompt: "do part A" } }); + await openCodeInputReady(calls, 2); + const child = await spawning; // The CLI's startup idle arrives after the prompt was typed but before its turn starts. terminals.applyProviderSignal(child.sessionId, { kind: "lifecycle", state: "idle" }); const early = await wait(1); @@ -303,3 +311,166 @@ test("the skill tells the orchestrator to pass a model the person names", async const skill = await readFile(new URL("../agent/orchestrator/SKILL.md", import.meta.url), "utf8"); assert.match(skill, /If the person names a model, pass it as `model`/u); }); + +test("wait_for_agent returns before the orchestrator's MCP client gives up, as a timeout that says the agent still runs", async () => { + for (const [provider, requested, expectedMs] of [ + ["opencode", 100, 50_000], ["codex", 100, 50_000], ["opencode", undefined, 50_000], ["opencode", 20, 20_000], ["claude", 100, 100_000] + ]) { + const { terminals, control, handler } = setup(); + terminals.sessions.get("orch").provider = provider; + let received; + const real = control.waitFor.bind(control); + control.waitFor = (id, request) => { received = request.timeoutMs; return real(id, { ...request, timeoutMs: 30 }); }; + terminals.sessions.get("child").buffer = "step 3 of 5"; + const result = await call(handler, "wait_for_agent", { sessionId: "child", ...(requested ? { timeoutSeconds: requested } : {}) }); + assert.equal(received, expectedMs, `${provider} asking for ${requested ?? "the default"}`); + assert.deepEqual([result.reason, result.status, result.output], ["timeout", "working", "step 3 of 5"]); + assert.match(result.message, /Still running\. Read output for its progress and call wait_for_agent again\./u); + } + const { terminals, handler } = setup(); + terminals.sessions.get("child").status = "idle"; + assert.equal((await call(handler, "wait_for_agent", { sessionId: "child" })).message, undefined, "only a timeout says to call again"); +}); + +/** A recorded OpenCode-style session: alternate screen, absolute cursor moves, colors, a spinner and in-place redraws. */ +function openCodeRedraws() { + const e = "\u001b"; + const at = (row, column) => `${e}[${row};${column}H`; + const parts = [ + `${e}[?1049h${e}[?25l${e}[?2004h${e}]0;opencode${e}\\${e}[2J${e}[H`, + `${at(1, 1)}${e}[48;2;24;24;37m${e}[38;2;205;214;244m┃ opencode build gpt-5.5${e}[K${e}[0m`, + `${at(3, 3)}${e}[1mYou${e}[22m`, + `${at(4, 3)}${e}[38;5;252mFix the failing Button test${e}[39m`, + `${at(6, 3)}${e}[38;5;245mThinking${e}[0m` + ]; + // The spinner turns in place many times while the model works. + for (let turn = 0; turn < 120; turn++) parts.push(`${at(6, 12)}${"⠋⠙⠹⠸⠼⠴⠦⠧⠇⠏"[turn % 10]}${at(40, 1)}${e}[7m working ${turn}s ${e}[27m${e}[K`); + parts.push( + `${at(6, 1)}${e}[2K${at(6, 3)}${e}[38;2;166;227;161mThe Button test passes now.${e}[0m`, + `${at(7, 3)}Used key sk-secret-value for the request.`, + `${at(8, 3)}${e}[38;5;111m│${e}[0m src/Button.tsx ${e}[32m+3${e}[0m ${e}[31m-1${e}[0m`, + `${at(40, 1)}${e}[2K${e}[7m ctrl+p commands ${e}[27m` + ); + return parts.join(""); +} + +test("wait_for_agent and get_agent_result return a TUI's screen as plain text, not its escape sequences", async () => { + const { terminals, handler } = setup(); + const child = terminals.sessions.get("child"); + setTimeout(() => { + child.buffer = openCodeRedraws(); + child.outputOffset = child.buffer.length; + child.status = "idle"; + }, 20); + const waited = await call(handler, "wait_for_agent", { sessionId: "child", timeoutSeconds: 5 }); + const result = await call(handler, "get_agent_result", { sessionId: "child" }); + for (const output of [waited.output, result.output]) { + assert.equal(/[\u0000-\u0008\u000b-\u001f\u007f]/u.test(output), false, "no control characters or escape sequences"); + assert.deepEqual(output.split("\n"), [ + "┃ opencode build gpt-5.5", + "", + " You", + " Fix the failing Button test", + "", + " The Button test passes now.", + " Used key [redacted] for the request.", + " │ src/Button.tsx +3 -1", + ...Array.from({ length: 31 }, () => ""), + " ctrl+p commands" + ]); + assert.ok(!output.includes("sk-secret") && !output.includes("Thinking") && !output.includes("working"), "masked; only what the screen shows now"); + } +}); + +test("the plain-text output stays bounded and masked for a long line-based transcript", async () => { + const { terminals, control } = setup(); + const child = terminals.sessions.get("child"); + child.buffer = Array.from({ length: 4_000 }, (_value, index) => `\u001b[32mline ${index}\u001b[0m with key sk-secret-value\r\n`).join(""); + child.outputOffset = child.buffer.length; + child.status = "idle"; + const result = await control.result("child"); + assert.ok(result.output.length <= 8_192); + assert.ok(result.output.endsWith("line 3999 with key [redacted]"), result.output.slice(-80)); + assert.ok(!result.output.includes("\u001b") && !result.output.includes("sk-secret")); +}); + +/** Invisible SGR resets that pad a stream to an exact length: any length from 6 up is a sum of 3s and 4s. */ +function sgrPadding(length) { + const e = "\u001b"; + let fours = 0; + while ((length - 4 * fours) % 3 !== 0) fours++; + return `${e}[0m`.repeat(fours) + `${e}[m`.repeat((length - 4 * fours) / 3); +} + +test("a scrollback window that starts inside an escape sequence leaves no fragment of it on the screen", async () => { + const { terminals, handler } = setup(); + const child = terminals.sessions.get("child"); + const e = "\u001b"; + const at = (row, column) => `${e}[${row};${column}H`; + const start = "18;9Hof specialists who had internalized the machine."; + const redraws = []; + for (let turn = 0; turn < 900; turn++) redraws.push(`${at(40, 1)}${e}[7m working ${turn}s ${e}[27m${e}[K`); + const end = `${at(3, 3)}The first response was the assembler.`; + const body = start + redraws.join("") + end; + // The last 65,536 characters (the window the screen is rendered from) begin right after the ESC [ of a cursor move. + child.buffer = `${"earlier output\r\n".repeat(50)}${e}[` + body + sgrPadding(65_536 - body.length); + child.outputOffset = child.buffer.length; + child.status = "idle"; + const waited = await call(handler, "wait_for_agent", { sessionId: "child", timeoutSeconds: 5 }); + const result = await call(handler, "get_agent_result", { sessionId: "child" }); + for (const output of [waited.output, result.output]) { + assert.ok(!output.includes("18;9H"), output.slice(0, 80)); + assert.ok(output.includes(" The first response was the assembler."), output); + } +}); + +test("exitLines are the exited TUI's screen as plain text, not its redraws joined together", async () => { + const { terminals, handler } = setup(); + const child = terminals.sessions.get("child"); + const e = "\u001b"; + const at = (row, column) => `${e}[${row};${column}H`; + // Drawn once at the start and never again: still on the screen when the CLI exits, more than 16 KiB later. + const early = `${at(3, 3)}${e}[2mThinking${e}[0m${at(8, 3)}${e}[38;5;111m▣ Build · glm-5.3-flash${e}[0m`; + const turns = 1_300; + const progress = []; + for (let turn = 0; turn < turns; turn++) progress.push(`${at(2, 3)}${e}[38;2;166;227;161m${"■".repeat(turn % 7)}${"⬝".repeat(7 - (turn % 7))}${e}[0m`); + // A TUI redraws only the cells that changed: here a line written once, then its typos fixed in place. + const body = early + progress.join("") + + `${at(4, 3)}As explicit memoy address, and if a single instructon was` + + `${at(4, 19)}ry address, and if a single instruction was` + + `${at(6, 3)}${e}[31mError: the model stopped answering.${e}[0m`; + assert.ok(body.length > 16_384 && body.length < 65_000, String(body.length)); + // The rendered window (the last 65,536 characters) begins with the `m` of a color sequence whose start was cut off. + child.buffer = `${"x".repeat(100)}${e}[38;2;1;2;3` + "m" + body + sgrPadding(65_536 - 1 - body.length); + child.outputOffset = child.buffer.length; + child.exitCode = 1; + child.status = "failed"; + const waited = await call(handler, "wait_for_agent", { sessionId: "child", timeoutSeconds: 5 }); + const observed = await call(handler, "observe_agent", { sessionId: "child" }); + const result = await call(handler, "get_agent_result", { sessionId: "child" }); + assert.equal(waited.reason, "failed"); + const last = (turns - 1) % 7; + assert.deepEqual(waited.exitLines.split("\n"), [ + ` ${"■".repeat(last)}${"⬝".repeat(7 - last)}`, + " Thinking", + " As explicit memory address, and if a single instruction was", + " Error: the model stopped answering.", + " ▣ Build · glm-5.3-flash" + ]); + assert.equal(observed.exitLines, waited.exitLines); + assert.equal(result.exitLines, waited.exitLines); +}); + +test("observe_agent returns the screen as plain text too, within its maxChars", async () => { + const { terminals, handler } = setup(); + const child = terminals.sessions.get("child"); + child.buffer = openCodeRedraws(); + child.outputOffset = child.buffer.length; + child.status = "working"; + const observed = await call(handler, "observe_agent", { sessionId: "child" }); + assert.equal(/[\u0000-\u0008\u000b-\u001f\u007f]/u.test(observed.output), false, "no ESC or other control bytes"); + assert.ok(observed.output.includes(" The Button test passes now.") && observed.output.includes("[redacted]")); + assert.ok(!observed.output.includes("sk-secret") && !observed.output.includes("working")); + const short = await call(handler, "observe_agent", { sessionId: "child", maxChars: 256 }); + assert.ok(short.output.length <= 256 && short.output.endsWith(" ctrl+p commands"), short.output); +}); diff --git a/tests/permission-gate-fail-closed.test.mjs b/tests/permission-gate-fail-closed.test.mjs index a786939de..d4d57c576 100644 --- a/tests/permission-gate-fail-closed.test.mjs +++ b/tests/permission-gate-fail-closed.test.mjs @@ -14,7 +14,7 @@ import { join } from "node:path"; import test from "node:test"; import { fileURLToPath } from "node:url"; import { - AGENT_RUNTIME_ENV, DECISION_FAIL_CLOSED_ENV, OPENCODE_DECISIONS_ENV, RUNTIME_PROTOCOL_VERSION, permissionGateTimings + AGENT_RUNTIME_ENV, DECISION_FAIL_CLOSED_ENV, OPENCODE_DECISIONS_ENV, OPENCODE_DECISION_READS_ENV, RUNTIME_PROTOCOL_VERSION, permissionGateTimings } from "../src/agent-runtime/runtime-protocol.mjs"; import { FAIL_CLOSED_MESSAGE, parseDecision } from "../src/agent-runtime/permission-gate.mjs"; import { createOpenCodeDecisions } from "../src/agent-runtime/opencode-decisions.mjs"; @@ -262,6 +262,18 @@ test("launch: the decision hook carries the fail-closed flag; a launch without i assert.equal(launch.decisions, false); assert.equal(JSON.stringify(launch.args).includes(DECISION_FAIL_CLOSED_ENV), false); assert.equal(JSON.stringify(launch.args).includes("permission-gate"), false); + // Base protection off, a decision plugin applies: the fail-closed hook covers shells and writes, never reads. + const pluginOnly = new DecisionHooks({ baseProtection: () => false, services: () => [{ pluginId: "p", pluginName: "P", serviceId: "s", mayAllow: false }], + call: async () => null, session: () => null }); + const pluginBridge = new AgentRuntimeBridge(gateway, { ...options, coreHooksEnabled: true, wantsDecisions: (provider) => pluginOnly.wanted(provider), + decisionReads: () => pluginOnly.checksReads() }); + const pluginLaunch = pluginBridge.prepareLaunch({ terminalSessionId: "b", provider: "claude", cwd: root }); + assert.equal(pluginLaunch.decisions, true); + assert.equal(JSON.parse(pluginLaunch.args[1]).hooks.PreToolUse[0].matcher, "Bash|Write|Edit|MultiEdit|NotebookEdit"); + const on = new DecisionHooks({ baseProtection: () => true, services: () => [], call: async () => null, session: () => null }); + assert.equal(on.checksReads(), true); + assert.equal(new DecisionHooks({ baseProtection: () => { throw new Error("settings"); }, services: () => [], call: async () => null, session: () => null }).checksReads(), true, + "a settings read that fails counts as on"); }); test("OpenCode: with decisions on, a check that cannot be made fails the tool call; ordinary answers are unchanged", async () => { @@ -282,7 +294,12 @@ test("OpenCode: with decisions on, a check that cannot be made fails the tool ca await createOpenCodeDecisions({ env, send: async () => ({ behavior, message: "", unavailable: false }) }).guard(...call); } // Tools that are not checked are never sent and never fail. - await createOpenCodeDecisions({ env, send: async () => null }).guard({ tool: "read", callID: "c2" }, { args: { filePath: "/etc/hosts" } }); + await createOpenCodeDecisions({ env, send: async () => null }).guard({ tool: "webfetch", callID: "c2" }, { args: { url: "https://example.com" } }); + // Base protection off at launch: a read-only file tool is not checked, so it never fails closed. + await createOpenCodeDecisions({ env, send: async () => null }).guard({ tool: "read", callID: "c3" }, { args: { filePath: "/etc/hosts" } }); + // Base protection on at launch: it is checked (for CanvasTTY's private data), so it fails closed like the others. + await assert.rejects(createOpenCodeDecisions({ env: { ...env, [OPENCODE_DECISION_READS_ENV]: "1" }, send: async () => null }) + .guard({ tool: "read", callID: "c4" }, { args: { filePath: "/etc/hosts" } }), (error) => error.message === FAIL_CLOSED_MESSAGE); }); test("the gate script stays importable without side effects (no socket, no stdin read)", async () => { diff --git a/tests/plugin-launch-choices.test.mjs b/tests/plugin-launch-choices.test.mjs index 884ea5bb6..1f024e90b 100644 --- a/tests/plugin-launch-choices.test.mjs +++ b/tests/plugin-launch-choices.test.mjs @@ -158,7 +158,9 @@ test("spawn_agent takes plugin launch options and hands them to the launch", () get: (id) => (id === "orch" ? parent : undefined), getMetadata: (id) => (id === "orch" ? parent : null), listMetadata: () => [parent], - create: (request) => { created.push(request); return { id: "child", ...request, status: "starting", title: "c" }; } + create: (request) => { created.push(request); return { id: "child", ...request, status: "starting", title: "c" }; }, + launchOptionsOf: () => undefined, + modelSetByOf: () => [] }; const control = new AgentControlService(terminals); control.spawn({ parentSessionId: "orch", provider: "codex", cwd: process.cwd(), launchOptions: { "canvastty-accounts": { account: "ollama" } } }); diff --git a/tests/runtime-result-transport.test.mjs b/tests/runtime-result-transport.test.mjs new file mode 100644 index 000000000..010f56428 --- /dev/null +++ b/tests/runtime-result-transport.test.mjs @@ -0,0 +1,129 @@ +import assert from "node:assert/strict"; +import { spawn } from "node:child_process"; +import { realpathSync } from "node:fs"; +import { mkdtemp, rm } from "node:fs/promises"; +import { tmpdir } from "node:os"; +import { join } from "node:path"; +import test from "node:test"; +import { fileURLToPath } from "node:url"; +import { AGENT_RUNTIME_ENV, CAPTURE_RESULT_ENV } from "../src/agent-runtime/runtime-protocol.mjs"; +import { AgentControlService } from "../src/main/services/AgentControlService.ts"; +import { AgentRuntimeBridge } from "../src/main/services/agent-runtime/AgentRuntimeBridge.ts"; +import { RuntimeGateway } from "../src/main/services/agent-runtime/RuntimeGateway.ts"; +import { applyRuntimeSignal } from "../src/main/services/agent-runtime/applyRuntimeSignal.ts"; +import { TerminalManager } from "../src/main/services/TerminalManager.ts"; +import { availableRegistry, inputReadySpawner } from "./helpers/terminal.mjs"; + +const POSIX = { timeout: 15_000, skip: process.platform === "win32" ? "Real Unix socket roundtrip; the Windows pipe transport has its own suite." : false }; +const helper = fileURLToPath(new URL("../src/agent-runtime/hook-helper.mjs", import.meta.url)); +const plugin = fileURLToPath(new URL("../src/agent-runtime/opencode-plugin.mjs", import.meta.url)); + +/** Runs a hook or plugin process the way the agent CLI would, with the launch's environment. */ +function run(args, env, input = "") { + const child = spawn(process.execPath, args, { env: { ...process.env, ...env }, stdio: ["pipe", "ignore", "pipe"] }); + const timeout = setTimeout(() => child.kill(), 5_000); + let stderr = ""; + child.stderr.on("data", (chunk) => { stderr += chunk; }); + child.stdin.end(input); + return new Promise((resolve, reject) => { + child.on("error", reject); + child.on("close", (code) => { clearTimeout(timeout); code === 0 ? resolve() : reject(new Error(stderr || `exit ${code}`)); }); + }); +} + +const openCodeTurn = (answer) => ` + const { CanvasTTYLifecycle } = await import(${JSON.stringify(new URL("../src/agent-runtime/opencode-plugin.mjs", import.meta.url).href)}); + const hooks = await CanvasTTYLifecycle({ client: { session: { messages: async () => ({ data: [{ info: { role: "assistant" }, parts: [{ type: "text", text: ${JSON.stringify(answer)} }] }] }) } } }); + const event = (type, rest = {}) => hooks.event({ event: { type, properties: { sessionID: "worker-root", ...rest } } }); + await event("session.created", { info: { id: "worker-root" } }); + await event("session.status", { status: { type: "busy" } }); + await event("session.status", { status: { type: "idle" } }); + await event("session.idle"); +`; + +/** A real gateway, bridge and hook helper/plugin, with fake CLI PTYs; signals go through main's own glue. */ +async function fixture(t, statusHooks) { + const root = realpathSync(await mkdtemp(join(tmpdir(), "ctty-rt-"))); + const calls = []; + let terminals; + const gateway = new RuntimeGateway({ runtimeDirectory: root, onSignal: (id, signal) => applyRuntimeSignal(terminals, id, signal) }); + await gateway.start(); + const bridge = new AgentRuntimeBridge(gateway, { + runtimeDirectory: join(root, "configs"), helper: { command: process.execPath, args: [helper] }, openCodePluginPath: plugin, + coreHooksEnabled: statusHooks, environment: {}, + kimiHomeDirectory: join(root, "kimi"), hermesHomeDirectory: join(root, "hermes"), grokHomeDirectory: join(root, "grok") + }); + terminals = new TerminalManager(() => undefined, availableRegistry(), undefined, bridge, statusHooks, inputReadySpawner(calls)); + t.after(async () => { + terminals.disposeAll(); + await gateway.close(); + await rm(root, { recursive: true, force: true }); + }); + const control = new AgentControlService(terminals, { waitTiming: { checkMs: 5, settleMs: 0, quietMs: 200 } }); + const parent = terminals.create({ provider: "codex", profile: "normal", cwd: root, role: "orchestrator", position: { x: 0, y: 0 } }); + // Status hooks are one switch: the bridge (hooks, leases) and the cards (status) follow it together, as in main. + const setStatusHooks = (enabled) => { + bridge.setCoreHooksEnabled(enabled); + terminals.setLifecycleHooksEnabled(enabled); + }; + return { root, calls, gateway, bridge, terminals, control, parent, setStatusHooks }; +} + +async function finishTurn(provider, env, answer) { + if (provider === "codex") { + await run([helper, "working", "UserPromptSubmit"], env, JSON.stringify({ turn_id: "fixture-turn" })); + await run([helper, "idle", "Stop"], env, JSON.stringify({ turn_id: "fixture-turn", last_assistant_message: answer })); + } else { + await run(["--input-type=module", "-e", openCodeTurn(answer)], env); + } +} + +for (const provider of ["codex", "opencode"]) { + test(`${provider} subagent's answer reaches wait_for_agent through its hooks while status hooks are on`, POSIX, async (t) => { + const f = await fixture(t, true); + const worker = await f.control.spawn({ parentSessionId: f.parent.id, provider, cwd: f.root }); + const env = f.calls.at(-1).options.env; + assert.equal(env[CAPTURE_RESULT_ENV], "1"); + assert.ok(env[AGENT_RUNTIME_ENV.capabilityToken]); + await f.terminals.deliverInput(worker.id, "worker task\r"); + await finishTurn(provider, env, "worker final answer"); + const waited = await f.control.waitFor(worker.id, { timeoutMs: 2_000 }); + assert.equal(waited.reason, "idle"); + assert.equal(waited.answer?.text, "worker final answer"); + assert.equal(waited.answerUnavailable, undefined); + + // Turning status hooks off revokes the lease: this launch has no hooks any more, and the result says so. + f.setStatusHooks(false); + assert.equal(f.gateway.currentStatus(worker.id), null); + assert.equal((await f.control.result(worker.id)).answer?.text, "worker final answer", "an answer captured earlier stays"); + await f.terminals.deliverInput(worker.id, "next task\r"); + const after = await f.control.result(worker.id); + assert.equal(after.answer, undefined); + assert.equal(after.answerUnavailable?.reason, "status_hooks_off"); + }); + + test(`${provider} subagent started with status hooks off gets no hooks, and the wait says its answer cannot be captured`, POSIX, async (t) => { + const f = await fixture(t, false); + const worker = await f.control.spawn({ parentSessionId: f.parent.id, provider, cwd: f.root }); + const launch = f.calls.at(-1); + assert.equal(launch.options.env[AGENT_RUNTIME_ENV.capabilityToken], undefined, "no runtime lease"); + assert.deepEqual(launch.args.filter((arg) => typeof arg === "string" && /^hooks\./u.test(arg)), [], "no lifecycle hooks"); + await f.terminals.deliverInput(worker.id, "worker task\r"); + const waited = await f.control.waitFor(worker.id, { timeoutMs: 2_000 }); + assert.equal(waited.reason, "quiet"); + assert.equal(waited.status, "unavailable"); + assert.equal(waited.answer, undefined); + assert.equal(waited.answerUnavailable?.reason, "status_hooks_off"); + assert.match(waited.answerUnavailable.message, /status hooks are off/u); + + // Turning them on applies to agents started or restarted afterwards. + f.setStatusHooks(true); + assert.equal((await f.control.result(worker.id)).answerUnavailable?.reason, "status_hooks_off"); + launch.process.emitExit(0); + f.terminals.restart(worker.id); + const restarted = f.calls.at(-1); + assert.notEqual(restarted, launch); + assert.ok(restarted.options.env[AGENT_RUNTIME_ENV.capabilityToken], "the restarted launch has its hooks"); + assert.equal((await f.control.result(worker.id)).answerUnavailable, undefined); + }); +} diff --git a/tests/secret-redaction-overlap-property.test.mjs b/tests/secret-redaction-overlap-property.test.mjs new file mode 100644 index 000000000..babb593f6 --- /dev/null +++ b/tests/secret-redaction-overlap-property.test.mjs @@ -0,0 +1,114 @@ +/** + * Property: registered secrets that overlap one another, sit next to one another or are wrapped by the terminal + * never leave a character visible, in whole-text and tail masking alike. + * Secrets use only lower-case letters; the surrounding text never does, so any lower-case letter left outside a + * marker is part of a secret that leaked. + */ +import assert from "node:assert/strict"; +import test from "node:test"; +import { SecretRedactionRegistry } from "../src/main/services/safety/SecretRedaction.ts"; + +function prng(seed) { + let state = seed >>> 0; + return () => { + state = (state + 0x6d2b79f5) >>> 0; + let t = state; + t = Math.imul(t ^ (t >>> 15), t | 1); + t ^= t + Math.imul(t ^ (t >>> 7), t | 61); + return ((t ^ (t >>> 14)) >>> 0) / 4294967296; + }; +} + +const FILLER = "0123456789 .,;:!?()[]#%"; +const WRAP = ["\n", "\n ", "\n│ ", " ", "\r\n"]; + +function scenario(random) { + const int = (n) => Math.floor(random() * n); + const pick = (items) => items[int(items.length)]; + // A small alphabet makes accidental overlaps between secrets frequent; derived secrets force them. + const alphabet = "abcdef".slice(0, 2 + int(5)); + const word = (length) => Array.from({ length }, () => pick([...alphabet])).join(""); + const secrets = []; + for (let count = 1 + int(5); secrets.length < count;) { + const roll = random(); + if (secrets.length && roll < 0.35) { + // Shares a suffix of an earlier secret as its prefix: the two overlap when written back to back. + const base = pick(secrets); + secrets.push(base.slice(base.length - 2 - int(Math.max(1, base.length - 3))) + word(4 + int(8))); + } else if (secrets.length && roll < 0.5) { + const base = pick(secrets); + secrets.push(word(1 + int(3)) + base.slice(0, Math.max(6, base.length - int(4)))); + } else secrets.push(word(8 + int(16))); + } + const valid = secrets.filter((secret) => secret.length >= 8); + const filler = () => Array.from({ length: int(30) }, () => pick([...FILLER])).join(""); + const wrapped = (secret) => { + if (random() < 0.6) return secret; + let out = ""; + for (let i = 0; i < secret.length; i++) out += secret[i] + (i < secret.length - 1 && random() < 0.15 ? pick(WRAP) : ""); + return out; + }; + let text = ""; + for (let pieces = 1 + int(12); pieces > 0; pieces--) { + text += filler(); + const first = pick(valid); + const second = pick(valid); + const roll = random(); + if (roll < 0.4) { + // Overlap: second starts inside first wherever their texts agree. + let joined = first + second; + for (let k = Math.min(first.length, second.length) - 1; k > 0; k--) { + if (first.endsWith(second.slice(0, k))) { joined = first + second.slice(k); break; } + } + text += wrapped(joined); + } else if (roll < 0.7) text += wrapped(first) + wrapped(second); // adjacent + else text += wrapped(first); + } + text += filler(); + return { secrets: valid, text }; +} + +const leaked = (masked) => masked.replace(//gu, "").match(/[a-z]+/gu); + +test("property: no character of overlapping, adjacent or wrapped secrets survives masking", () => { + for (let seed = 1; seed <= 1500; seed++) { + const random = prng(seed); + const { secrets, text } = scenario(random); + const registry = new SecretRedactionRegistry(); + registry.add("vault", secrets); + const masked = registry.redact(text); + assert.equal(leaked(masked), null, `seed ${seed}: ${JSON.stringify({ secrets, text, masked })}`); + assert.equal(registry.redact(masked), masked, `seed ${seed}: idempotent`); + const maxChars = 1 + Math.floor(random() * text.length); + // The tail is the whole masked text cut to its last maxChars, so it can start inside a marker. + assert.equal(registry.redactTail(text, maxChars), masked.slice(-maxChars), `seed ${seed}: tail`); + } +}); + +test("property: tail masking equals masking the whole text and cutting it", () => { + // Line ends next to tokens, header schemes, separators and wrapped runs: the cases a tail window's line start + // must never split. The tail must equal whole-text masking cut to the same length. + const anthropic = ["sk", "ant", ""].join("-"); + const fragments = [ + "plain words", "line 42 output", "x".repeat(30), "Authorization:", "Authorization: Bearer", "Bearer", "Basic", "token", + `${anthropic}abcdefghij0123456789`, "abcd1234", "abcdefgh", "Q7vX2mK9pL4sT8wZ", "1nB6cR3yH5jD0fGa", "MY_TOKEN", "MY_TOKEN =", + "= value123", ": secretvalue", '{"token": "', 'abc def"}', "?token=abc", "https://user:pw@host", "|", "│", " ", "\t", + "PASSWORD=", "hunter22", "界界界", "ünïcödé", "->", "=>", ",", "." + ]; + const breaks = ["\n", "\r\n", "\r", "\n ", "\n│ ", " ", ""]; + const registry = new SecretRedactionRegistry(); + registry.add("vault", ["registered-secret-value-1", "another_registered_value_22"]); + for (let seed = 1; seed <= 400; seed++) { + const random = prng(seed); + const pick = (items) => items[Math.floor(random() * items.length)]; + let text = ""; + // Longer than the held values' reach plus the tail margin, so both cut where they would in a terminal. + for (let pieces = 2_500 + Math.floor(random() * 1_000); pieces > 0; pieces--) { + text += random() < 0.05 ? pick(["registered-secret-value-1", "another_registered_value_22", "registered-secret-\nvalue-1"]) : pick(fragments); + text += pick(breaks); + } + const whole = registry.redact(text); + const maxChars = 1 + Math.floor(random() * 4_000); + assert.equal(registry.redactTail(text, maxChars), whole.slice(-maxChars), `seed ${seed}: tail`); + } +}); diff --git a/tests/secret-redaction.test.mjs b/tests/secret-redaction.test.mjs index f3b74d5c2..8400979d9 100644 --- a/tests/secret-redaction.test.mjs +++ b/tests/secret-redaction.test.mjs @@ -87,6 +87,50 @@ test("masking stays linear on long runs of name-like text", () => { } }); +test("registry: known-value matching avoids text-sized typed arrays", () => { + const registry = new SecretRedactionRegistry(); + const key = "custom-secret-value-not-a-token"; + registry.add("vault", [key]); + const plain = `${"x".repeat(1_000_000)}${key}`; + const wrappedKey = `${key.slice(0, 12)}\n│ ${key.slice(12)}`; + const wrapped = `${"x".repeat(1_000_000)}${wrappedKey}`; + const NativeInt32Array = globalThis.Int32Array; + const NativeUint32Array = globalThis.Uint32Array; + let largestAllocation = 0; + const track = (Native) => new Proxy(Native, { + construct(target, args, newTarget) { + largestAllocation = Math.max(largestAllocation, Number(args[0]) || 0); + return Reflect.construct(target, args, newTarget); + } + }); + try { + globalThis.Int32Array = track(NativeInt32Array); + globalThis.Uint32Array = track(NativeUint32Array); + assert.equal(registry.redact(plain), `${"x".repeat(1_000_000)}`); + assert.equal(registry.redact(wrapped), `${"x".repeat(1_000_000)}`); + } finally { + globalThis.Int32Array = NativeInt32Array; + globalThis.Uint32Array = NativeUint32Array; + } + assert.ok(largestAllocation < 256, `allocated a typed array with ${largestAllocation} elements for a short held value`); +}); + +test("high-entropy checks do not build per-character match arrays for long ordinary runs", () => { + const ordinary = "x".repeat(1_000_000); + const originalMatch = String.prototype.match; + const largeMatchCalls = []; + try { + String.prototype.match = function (pattern) { + if (this.length >= ordinary.length) largeMatchCalls.push(pattern.source); + return originalMatch.call(this, pattern); + }; + assert.equal(redactCredentials(ordinary), ordinary); + } finally { + String.prototype.match = originalMatch; + } + assert.deepEqual(largeMatchCalls, [], "entropy classification counts characters without collecting regex matches"); +}); + /** What OpenCode printed in a live run: the config with its key, wrapped by the terminal inside a box. */ function boxedDump(key) { const config = `{"provider":{"x":{"options":{"baseURL":"https://api.example.test/v1","apiKey":"${key}"}}}}`; @@ -115,6 +159,75 @@ test("registry: a held value is removed also when wrapping split it across box l assert.equal(registry.redact(`x ${launch}`), `x ${launch}`); }); +test("registry: overlapping registered secrets are masked as one complete range", () => { + const registry = new SecretRedactionRegistry(); + registry.add("vault", ["abcdefgh", "abcdefghi", "defghijklmnop"]); + assert.equal(registry.redact("abc\n│ defghi tail"), " tail", "wrapped match at one start chooses the longer value"); + assert.equal(registry.redact("abcdefghi tail"), " tail", "unwrapped matches at one start choose the longer value"); + assert.equal(registry.redact("abcdefghi"), "", "wrap-free input preserves longest-match precedence"); + assert.equal(registry.redact("abcdefghijklmnop tail"), " tail", "a later overlapping secret cannot expose its suffix"); +}); + +test("registry: exact streams retain an overlap after an earlier match skips intervening hits", () => { + const registry = new SecretRedactionRegistry(); + const overlapping = "abcabcabcabc"; + registry.add("vault", ["xyabcabc", overlapping]); + assert.equal(registry.redact(`xy${"abc".repeat(8)}`), ""); +}); + +test("registry: every terminal wrap character is removed from the search view", () => { + const registry = new SecretRedactionRegistry(); + const value = "wrap-character-equivalence-key"; + registry.add("vault", [value]); + const wrapCharacters = [ + ...Array.from({ length: 5 }, (_, index) => String.fromCharCode(0x09 + index)), " ", "\u00a0", "\u1680", + ...Array.from({ length: 11 }, (_, index) => String.fromCharCode(0x2000 + index)), + "\u2028", "\u2029", "\u202f", "\u205f", "\u3000", "\ufeff", + ...Array.from({ length: 0x80 }, (_, index) => String.fromCharCode(0x2500 + index)) + ]; + const gaps = Array.from({ length: value.length - 1 }, () => []); + for (let index = 0; index < wrapCharacters.length; index++) { + gaps[Math.floor(index * gaps.length / wrapCharacters.length)].push(wrapCharacters[index]); + } + let wrapped = ""; + for (let index = 0; index < value.length; index++) wrapped += value[index] + (gaps[index]?.join("") ?? ""); + assert.equal(registry.redact(wrapped), ""); +}); + +test("registry: an invalid longer wrapped candidate does not hide a valid shorter candidate", () => { + const registry = new SecretRedactionRegistry(); + const shorter = "custom-secret-value-not-a-token"; + const longer = `${shorter}-tail`; + registry.add("vault", [shorter, longer]); + const text = `${shorter}${" ".repeat(65)}-tail`; + assert.equal(registry.redact(text), `${" ".repeat(65)}-tail`); +}); + +test("registry: exact-only wrapped values work when no bare candidate is present", () => { + const registry = new SecretRedactionRegistry(); + registry.add("vault", ["abc def", "not-present-secret"]); + assert.equal(registry.redact("abc def"), ""); +}); + +test("registry: wrapped overlaps survive the shared offset horizon and UTF-16 tail cut", () => { + const registry = new SecretRedactionRegistry(); + registry.add("vault", ["a".repeat(8), "a".repeat(10)]); + const tail = 4_000; + const base = scrollback(80_000); + const sample = `😀${Array.from({ length: 12 }, () => "a").join("\n│ ")}`; + // Place the first match three UTF-16 code units before the masked output's tail cut. + const removed = 12 + 11 * 3 - "".length; + const at = base.length + sample.length - removed - tail - 3 - "😀".length; + const text = `${base.slice(0, at)}${sample}${base.slice(at)}`; + const masked = registry.redact(text); + const marker = masked.indexOf(""); + assert.equal(marker, masked.length - tail - 3); + assert.equal(masked.slice(marker - 2, marker), "😀"); + assert.ok(masked.includes("")); + assert.ok(!masked.includes("\n│ a"), "every overlapping wrapped suffix is masked"); + assert.equal(registry.redactTail(text, tail), masked.slice(-tail)); +}); + test("registry: a value added again stays masked when an owner is full; only the least recently added one goes", (t) => { const warnings = []; t.mock.method(console, "warn", (message) => warnings.push(String(message))); @@ -145,6 +258,68 @@ test("registry: generic shapes without a held value, including keys the terminal assert.match(random, /plain words/u); }); +test("registry: held structural words cannot disable generic credential masking", () => { + const registry = new SecretRedactionRegistry(); + const privateKeyWords = ["PRIVATE", "KEY"].join(" "); + const passwordLabel = ["pass", "word"].join(""); + const authorizationLabel = ["Author", "ization"].join(""); + const embeddedToken = sk(run("R", 30)); + const fullRegisteredValue = `credential-envelope-${embeddedToken}-private-suffix`; + registry.add("vault", [privateKeyWords, passwordLabel, authorizationLabel, fullRegisteredValue]); + + const pem = `-----BEGIN RSA ${privateKeyWords}-----\n${run("M", 64)}\n-----END RSA ${privateKeyWords}-----`; + assert.equal(registry.redact(pem), "", "a held header fragment must not hide the PEM structure"); + + const assignment = `${passwordLabel}: ${"ordinary-secret-value-123"}`; + const redactedAssignment = registry.redact(assignment); + assert.equal(redactedAssignment.includes("ordinary-secret-value-123"), false, "the assignment value remains hidden after its label is masked"); + assert.match(redactedAssignment, //u); + + const json = `{"${passwordLabel}":"${"ordinary-json-secret-456"}"}`; + const redactedJson = registry.redact(json); + assert.equal(redactedJson.includes("ordinary-json-secret-456"), false, "the JSON value remains hidden after its key is masked"); + + const authorization = `${authorizationLabel}: Basic ${"opaqueCredentialValue123"}`; + const redactedAuthorization = registry.redact(authorization); + assert.equal(redactedAuthorization.includes("opaqueCredentialValue123"), false, "the authorization value remains hidden after its header is masked"); + assert.match(redactedAuthorization, //u); + + assert.equal(registry.redact(fullRegisteredValue), "", "an inner generic token cannot split a longer registered value"); +}); + +test("registry: a name joined to a held random value keeps its name", () => { + const registry = new SecretRedactionRegistry(); + registry.add("plugin:p.keys", [mixed]); + assert.equal(registry.redact(`env: PROVIDER_API_KEY=${mixed} PLAIN=visible`), "env: PROVIDER_API_KEY= PLAIN=visible"); + assert.equal(registry.redact(`token_${mixed}_tail`), "token__tail"); + assert.equal(registry.redact(`x ${mixed}${["aB3dE5gH", "iJ7kL9mN", "oP1qR2sT", "uV4wX6yZ"].join("")} y`), "x y", "a random run next to the value is still masked"); +}); + +test("registry: empty-registry fast path preserves overlapping JSON and generic redaction", () => { + const empty = new SecretRedactionRegistry(); + const rangePath = new SecretRedactionRegistry(); + rangePath.add("vault", ["unrelated-held-fixture-value-32"]); + const privateKeyWords = ["PRIVATE", "KEY"].join(" "); + const pem = (body) => `-----BEGIN RSA ${privateKeyWords}-----\n${body}\n-----END RSA ${privateKeyWords}-----`; + const apiKey = ["api", "Key"].join(""); + const authorization = ["Author", "ization"].join(""); + const basic = ["B", "asic"].join(""); + const url = `https://u:pw@fixture.example/?access_token=${run("Q", 12)}`; + const samples = [ + [`{"${apiKey}":"${pem(run("M", 24))}"}`, `{"${apiKey}":""}`], + [`{"${authorization}":"${authorization}: ${basic} ${"opaqueCredentialValue123"}"}`, `{"${authorization}":""}`], + [`{"token":"${url}"}`, `{"token":""}`], + [pem(`{"${apiKey}":"${"ordinary-json-secret-456"}"}`), ""] + ]; + + for (const [sample, expected] of samples) { + const fastPath = empty.redact(sample); + assert.equal(fastPath, rangePath.redact(sample), sample.slice(0, 32)); + assert.equal(fastPath, expected, sample.slice(0, 32)); + assert.equal(fastPath.includes("opaqueCredentialValue123") || fastPath.includes("ordinary-json-secret-456"), false); + } +}); + test("the vault hands every value it reads or writes to the registry; agent-readable text is masked", async (t) => { const directory = await mkdtemp(join(tmpdir(), "canvastty-redaction-vault-")); t.after(() => rm(directory, { recursive: true, force: true })); @@ -163,7 +338,7 @@ test("the vault hands every value it reads or writes to the registry; agent-read readBuffer: () => ({ buffer: `env OPENAI=${value}\n${secrets.github}\n` }), redactSecrets: (text) => manager.redactSecrets(text) }); - const observed = control.observe("s1").output; + const observed = (await control.observe("s1")).output; assert.equal(observed.includes(value) || observed.includes(secrets.github), false); assert.match(observed, /env OPENAI=/u); }); @@ -204,17 +379,17 @@ test("observe_agent and get_agent_result mask the whole buffer before cutting th const control = new AgentControlService(f.terminals); // The 300-character observation starts 12 characters before the secret ends. f.print(`${"a".repeat(1_000)}${PLAIN_SECRET}${"b".repeat(288)}`); - const observed = control.observe(f.card.id, 300).output; + const observed = (await control.observe(f.card.id, 300)).output; assert.ok(observed.length <= 300, "the bound is kept"); assert.deepEqual(fragments(observed), []); // get_agent_result's 8 192-character tail, cut inside the secret the same way. f.print(`${PLAIN_SECRET}${"c".repeat(8_192 - 10)}`); - const result = control.result(f.card.id).output; + const result = (await control.result(f.card.id)).output; assert.ok(result.length <= 8_192); assert.deepEqual(fragments(result), []); // Wrapped secrets are still masked whole. f.print(`\r\n${PLAIN_SECRET.slice(0, 15)}\r\n${PLAIN_SECRET.slice(15)}\r\n`); - assert.deepEqual(fragments(control.observe(f.card.id, 8_192).output), []); + assert.deepEqual(fragments((await control.observe(f.card.id, 8_192)).output), []); }); test("plugin screen text and failure details mask the whole buffer before cutting", async (t) => { @@ -318,8 +493,8 @@ test("observe_agent, get_agent_result and the plugin screen mask a bounded windo f.print(scrollback(250_000)); f.print(`${PLAIN_SECRET}${"g".repeat(200)}`); const control = new AgentControlService(f.terminals); - assert.deepEqual(fragments(control.observe(f.card.id).output), []); - assert.deepEqual(fragments(control.result(f.card.id).output), []); + assert.deepEqual(fragments((await control.observe(f.card.id)).output), []); + assert.deepEqual(fragments((await control.result(f.card.id)).output), []); const { PluginSessions } = await import("../src/main/services/PluginSessions.ts"); const screens = []; const sessions = new PluginSessions({ terminals: f.terminals, notify: (_p, _s, _m, event) => { if (event.screen !== undefined) screens.push(event.screen); return true; } }); @@ -330,7 +505,8 @@ test("observe_agent, get_agent_result and the plugin screen mask a bounded windo await new Promise((resolve) => setTimeout(resolve, 20)); assert.ok(screens.length > 0); assert.deepEqual(fragments(screens.join("\n")), []); - assert.ok(masked.length >= 3 && Math.max(...masked) < 40_000, `masked ${masked} characters per call`); + // get_agent_result renders its plain text from a 64 KiB masked window (plus the tail margin); never the whole scrollback. + assert.ok(masked.length >= 3 && Math.max(...masked) < 90_000, `masked ${masked} characters per call`); }); /** @@ -382,6 +558,23 @@ test("registry: held values of 4k, 16k and 64k characters are masked whole, also } }); +test("registry: maximum-length repetitive values mask dense overlapping occurrences", () => { + const registry = new SecretRedactionRegistry(); + const value = "a".repeat(65_536); + registry.add("vault", [value]); + const text = "a".repeat(150_000); + assert.equal(registry.redact(text), ""); +}); + +test("registry: long native-prefix skips preserve UTF-16 matches", () => { + const registry = new SecretRedactionRegistry(); + const value = `😀${"x".repeat(127)}`; + const distinctPrefix = `abcdefgH${"x".repeat(130)}`; + registry.add("vault", [value, distinctPrefix]); + assert.equal(registry.redact(`before 🛡️${value} after`), "before 🛡️ after"); + assert.equal(registry.redact(distinctPrefix), ""); +}); + test("registry: a long held value never breaks masking of anything else", () => { const registry = new SecretRedactionRegistry(); registry.add("plugin:p.big", [longSecret(4_000)]); diff --git a/tests/shutdown-steps.test.mjs b/tests/shutdown-steps.test.mjs new file mode 100644 index 000000000..fd0c46dd5 --- /dev/null +++ b/tests/shutdown-steps.test.mjs @@ -0,0 +1,76 @@ +import assert from "node:assert/strict"; +import test from "node:test"; +import { runShutdownSteps } from "../src/main/services/shutdownSteps.ts"; + +test("a failed shutdown step does not skip the steps after it", async () => { + const calls = [], warnings = []; + const failure = new Error("fixture close failed"); + await assert.doesNotReject(runShutdownSteps([ + { name: "Even G2", run: async () => { calls.push("Even G2"); throw failure; } }, + { name: "sync step", run: () => { calls.push("sync step"); throw failure; } }, + ...["terminal processes", "browser", "plugin services", "plugins", "materials"].map(name => ({ name, run: async () => { calls.push(name); } })) + ], (message, error) => warnings.push({ message, error }))); + assert.deepEqual(calls, ["Even G2", "sync step", "terminal processes", "browser", "plugin services", "plugins", "materials"]); + assert.equal(warnings.length, 2); + assert.equal(warnings[0].error, failure); + assert.match(warnings[0].message, /Even G2 failed/u); +}); + +test("shutdown steps run in order and each waits for the previous one", async () => { + const events = []; + let release; + const held = new Promise(resolve => { release = resolve; }); + const done = runShutdownSteps([ + { name: "first", run: async () => { events.push("first:start"); await held; events.push("first:done"); } }, + { name: "second", run: () => { events.push("second"); } }, + { name: "absent service", run: () => undefined } + ], () => assert.fail("no step failed")); + await Promise.resolve(); + assert.deepEqual(events, ["first:start"]); + release(); + await done; + assert.deepEqual(events, ["first:start", "first:done", "second"]); +}); + +test("quitting still hangs up the terminals when the Even G2 companion fails to close", async () => { + const { readFile } = await import("node:fs/promises"); + const { stripTypeScriptTypes } = await import("node:module"); + const { runInNewContext } = await import("node:vm"); + const main = (await readFile(new URL("../src/main/index.ts", import.meta.url), "utf8")).replaceAll("\r\n", "\n"); + const start = main.indexOf("async function shutdownServices"); + const source = main.slice(start, main.indexOf("\n}\n", start) + 2); + const calls = []; + // The usage-history save finishes only after the limits service aborts its requests. + let limitsAborted; + const aborted = new Promise(resolve => { limitsAborted = resolve; }); + const service = (name, method, fail = false) => ({ [method]: async () => { calls.push(name); if (fail) throw new Error(`fixture ${name} failed`); } }); + const warn = console.warn; + console.warn = () => undefined; + try { + const shutdownServices = runInNewContext(`${stripTypeScriptTypes(source)}; shutdownServices`, { + runShutdownSteps, console, clearTimeout, clearInterval, Promise, Error, + diagnostics: { record: () => undefined, flush: async () => { calls.push("diagnostics"); } }, + mascotManager: { dispose: () => { calls.push("mascots"); } }, + agentChatHistory: { dispose: () => { calls.push("chat history"); } }, + agentControl: service("agent control", "close"), + updateTimer: null, updateInterval: null, browserRequests: new Map(), + evenG2: service("Even G2", "close", true), + terminalManager: { shutdown: async () => { calls.push("terminals"); }, waitForProcessExits: async () => 0 }, + usageHistory: { dispose: () => { calls.push("usage history"); return aborted.then(() => { calls.push("usage history saved"); }); } }, + limitsService: { dispose: () => { calls.push("limits"); limitsAborted(); } }, + agentGateway: service("agent gateway", "close"), + runtimeGateway: service("runtime gateway", "close"), + browserService: service("browser", "dispose"), + pluginServices: service("plugin services", "dispose"), + pluginManager: service("plugins", "dispose"), + handoffService: service("handoff", "dispose"), + materialService: service("materials", "dispose") + }); + await shutdownServices(); + } finally { + console.warn = warn; + } + assert.deepEqual(calls, ["mascots", "chat history", "agent control", "Even G2", "terminals", "usage history", "limits", + "usage history saved", "agent gateway", "runtime gateway", "browser", "plugin services", "plugins", "handoff", "materials", + "diagnostics"]); +}); diff --git a/tests/subagent-model-account.test.mjs b/tests/subagent-model-account.test.mjs new file mode 100644 index 000000000..d6e91c98f --- /dev/null +++ b/tests/subagent-model-account.test.mjs @@ -0,0 +1,312 @@ +import assert from "node:assert/strict"; +import { mkdtemp, rm } from "node:fs/promises"; +import { tmpdir } from "node:os"; +import { join } from "node:path"; +import test from "node:test"; +import { AgentControlService } from "../src/main/services/AgentControlService.ts"; +import { ScopedOrchestrationHandler } from "../src/main/services/agent-browser/OrchestrationTools.ts"; +import { LaunchPipeline, modelSettingArgument } from "../src/main/services/LaunchPipeline.ts"; +import { TerminalManager } from "../src/main/services/TerminalManager.ts"; +import { TerminalSessionStore } from "../src/main/services/TerminalSessionStore.ts"; +import { availableRegistry, inputReadySpawner } from "./helpers/terminal.mjs"; + +// A subagent spawned without launch options once ran on its CLI's own default model (for OpenCode the hosted OpenCode +// Zen model), while its orchestrator ran on a model the person picked through a plugin (a model account). It now keeps +// the orchestrator's options for every plugin the spawn does not name; naming a plugin is an explicit choice, and a +// second model flag next to a plugin's own is refused instead of starting a CLI that crashes. + +const at = { x: 0, y: 0 }; +// Any plugin id: the core knows no plugin by name. +const MODELS = "example-model-accounts"; +const EXTRAS = "example-extras"; +const OTHER = "example-other-models"; +const BARE = "example-no-fields"; + +async function setup(t, { delegable = true, store } = {}) { + const root = await mkdtemp(join(tmpdir(), "ctty-sub-account-")); + t.after(() => rm(root, { recursive: true, force: true })); + const calls = []; + const prepared = []; + // What the person can change while the orchestrator runs: turn a plugin off, or a plugin update changes its fields. + const state = { gone: new Set(), extrasFields: [{ key: "flavor", label: "Flavor", kind: "text" }] }; + const terminals = new TerminalManager(() => undefined, availableRegistry(), undefined, undefined, true, inputReadySpawner(calls)); + t.after(() => store ? terminals.shutdown() : terminals.disposeAll()); + terminals.configureLaunchPipeline(new LaunchPipeline({ + contributors: () => [ + { pluginId: MODELS, pluginName: "Model accounts", serviceId: "accounts", secrets: false, + launch: { fields: [{ key: "account", label: "Model account", kind: "text" }], ...(delegable ? { delegable: true } : {}) } }, + // Options that do not decide the model. + { pluginId: EXTRAS, pluginName: "Extras", serviceId: "extras", secrets: false, + launch: { fields: state.extrasFields } }, + // A plugin without fields: choosing it in the launcher stores {}. + { pluginId: BARE, pluginName: "No fields", serviceId: "bare", secrets: false, launch: { fields: [] } }, + // A second plugin that also sets the model. + { pluginId: OTHER, pluginName: "Other models", serviceId: "other", secrets: false, + launch: { fields: [{ key: "pick", label: "Model", kind: "text" }], delegable: true } } + ].filter((contributor) => !state.gone.has(contributor.pluginId)), + call: async (pluginId, _service, method, params) => { + if (method !== "canvastty.launch.prepare") return null; + prepared.push({ pluginId, sessionId: params.sessionId, chosen: params.chosen, options: params.options }); + if (pluginId === OTHER && params.chosen) return { env: {}, secretEnv: {}, args: [`--model=other/${params.options.pick}`], files: [] }; + return pluginId === MODELS && params.chosen && params.options.account && params.options.account !== "none" + ? { env: {}, secretEnv: {}, args: ["--model", `acct_${params.options.account}/glm-5.3-flash`], files: [] } + : { env: {}, secretEnv: {}, args: [], files: [] }; + }, + secret: async () => null, runsRoot: join(root, "runs"), timeoutMs: 2000 + })); + if (store) { + terminals.configureSessionPersistence(new TerminalSessionStore(store), "continue"); + await terminals.restorePersistedSessions(); + } + const control = new AgentControlService(terminals); + const handler = new ScopedOrchestrationHandler(control, null, + { cli: () => "available", limits: () => null, models: () => ({ models: ["zai/glm-5.3"], checkedAt: Date.now() }), checkModel: async () => null }, + {}); + const orchestrator = (launchOptions, provider = "opencode") => terminals.create({ provider, profile: "normal", cwd: root, position: at, role: "orchestrator", + ...(launchOptions ? { launchOptions } : {}) }); + const spawn = (parent, args) => handler.execute(parent.id, { id: `c-${Math.random()}`, tool: "spawn_agent", arguments: { provider: "opencode", cwd: root, prompt: "Small task", ...args } }); + const settle = () => new Promise((resolve) => setTimeout(resolve, 50)); + return { root, calls, prepared, state, terminals, control, handler, orchestrator, spawn, settle }; +} + +test("a subagent spawned without launch options runs on its orchestrator's model account, never on the CLI default", async (t) => { + const s = await setup(t); + const parent = s.orchestrator({ [MODELS]: { account: "glm-flash" } }); + await s.settle(); + assert.deepEqual(s.terminals.modelSetByOf(parent.id), [MODELS]); + const child = await s.spawn(parent, {}); + await s.settle(); + assert.deepEqual(child.servedBy, { provider: "opencode", model: null, launchOptions: { [MODELS]: "inherited" } }); + const childPrepare = s.prepared.find((entry) => entry.sessionId === child.sessionId); + assert.deepEqual(childPrepare, { pluginId: MODELS, sessionId: child.sessionId, chosen: true, options: { account: "glm-flash" } }); + const launched = s.calls.at(-1).args; + assert.ok(launched.includes("acct_glm-flash/glm-5.3-flash"), launched.join(" ")); + assert.equal(launched.filter((arg) => arg === "--model").length, 1); + assert.deepEqual(s.terminals.launchOptionsOf(child.sessionId), { [MODELS]: { account: "glm-flash" } }); +}); + +test("naming the plugin is an explicit choice; without any options the CLI's default serves the subagent", async (t) => { + const s = await setup(t); + const parent = s.orchestrator({ [MODELS]: { account: "glm-flash" } }); + await s.settle(); + const own = await s.spawn(parent, { launchOptions: { [MODELS]: { account: "none" } } }); + await s.settle(); + assert.deepEqual(own.servedBy.launchOptions, { [MODELS]: "explicit" }); + assert.deepEqual(s.terminals.launchOptionsOf(own.sessionId), { [MODELS]: { account: "none" } }); + assert.ok(!s.calls.at(-1).args.some((arg) => arg.startsWith("acct_")), "explicit none: no account model"); + assert.deepEqual(s.terminals.modelSetByOf(own.sessionId), []); + + const bare = await setup(t); + const plain = bare.orchestrator(); + const child = await bare.spawn(plain, {}); + await bare.settle(); + assert.deepEqual(child.servedBy, { provider: "opencode", model: null, launchOptions: {} }); + assert.equal(bare.terminals.launchOptionsOf(child.sessionId), undefined); +}); + +test("another value needs the plugin's delegable declaration; the orchestrator's own values do not", async (t) => { + const s = await setup(t, { delegable: false }); + const parent = s.orchestrator({ [MODELS]: { account: "glm-flash" }, [EXTRAS]: { flavor: "mild" } }); + await s.settle(); + const inherited = await s.spawn(parent, {}); + assert.deepEqual(inherited.servedBy.launchOptions, { [MODELS]: "inherited", [EXTRAS]: "inherited" }); + assert.deepEqual(s.terminals.launchOptionsOf(inherited.sessionId), { [MODELS]: { account: "glm-flash" }, [EXTRAS]: { flavor: "mild" } }); + const same = await s.spawn(parent, { launchOptions: { [MODELS]: { account: "glm-flash" } } }); + assert.deepEqual(same.servedBy.launchOptions, { [MODELS]: "explicit", [EXTRAS]: "inherited" }); + await assert.rejects(s.spawn(parent, { launchOptions: { [MODELS]: { account: "other-paid" } } }), + /Model accounts has not declared its launch options safe for an orchestrator to choose/u); +}); + +test("a subagent of another CLI cannot silently drop the options that set the orchestrator's model", async (t) => { + const s = await setup(t); + const parent = s.orchestrator({ [MODELS]: { account: "glm-flash" }, [EXTRAS]: { flavor: "mild" } }); + await s.settle(); + const refused = new RegExp(`model is set by its ${MODELS} launch options, which the person chose for opencode; a codex subagent would not get that model\\. Spawn a opencode subagent instead\\.`, "u"); + await assert.rejects(s.spawn(parent, { provider: "codex" }), refused); + assert.throws(() => s.control.spawn({ parentSessionId: parent.id, provider: "codex", cwd: s.root }), refused); + // Naming the plugin for the other CLI is explicit; options that do not decide the model stay with their CLI. + const explicit = await s.spawn(parent, { provider: "codex", launchOptions: { [MODELS]: { account: "none" } } }); + assert.deepEqual(explicit.servedBy.launchOptions, { [MODELS]: "explicit" }); + assert.deepEqual(s.terminals.launchOptionsOf(explicit.sessionId), { [MODELS]: { account: "none" } }); +}); + +test("AgentControlService.spawn applies the same inheritance for every caller", async (t) => { + const s = await setup(t); + const parent = s.orchestrator({ [MODELS]: { account: "glm-flash" } }); + await s.settle(); + const child = await s.control.spawn({ parentSessionId: parent.id, provider: "opencode", cwd: s.root }); + await s.settle(); + assert.deepEqual(s.terminals.launchOptionsOf(child.id), { [MODELS]: { account: "glm-flash" } }); + assert.deepEqual(s.terminals.modelSetByOf(child.id), [MODELS]); +}); + +test("a subagent with an explicit model next to launch options that set the model is refused instead of a crashing double --model", async (t) => { + const s = await setup(t); + const parent = s.orchestrator({ [EXTRAS]: { flavor: "mild" } }); + await s.settle(); + const spawned = s.calls.length; + const child = await s.control.spawn({ parentSessionId: parent.id, provider: "opencode", cwd: s.root, model: "zai/glm-5.3", + launchOptions: { [MODELS]: { account: "glm-flash" } } }); + await s.settle(); + const failed = s.terminals.getMetadata(child.id); + assert.equal(failed.status, "failed"); + assert.match(failed.failureDetails, /Model accounts sets the model of this launch, and the launch also asks for model zai\/glm-5\.3/u); + assert.equal(s.calls.length, spawned, "no CLI was started with two --model flags"); + + // Options that do not set the model take an explicit model as before. + const plain = await s.control.spawn({ parentSessionId: parent.id, provider: "opencode", cwd: s.root, model: "zai/glm-5.3" }); + await s.settle(); + assert.notEqual(s.terminals.getMetadata(plain.id).status, "failed"); + assert.equal(s.calls.at(-1).args.filter((arg) => arg === "--model").length, 1); + + // A subagent that would inherit them is refused before anything starts. + const accounts = s.orchestrator({ [MODELS]: { account: "glm-flash" } }); + await s.settle(); + await assert.rejects(s.spawn(accounts, { model: "zai/glm-5.3" }), + new RegExp(`inherits its orchestrator's ${MODELS} launch options, which set its model; it cannot also take model zai/glm-5\\.3`, "u")); +}); + +test("a subagent whose launch options come from two plugins that both set the model is refused", async (t) => { + const s = await setup(t); + const parent = s.orchestrator(); + await s.settle(); + const spawned = s.calls.length; + const child = await s.control.spawn({ parentSessionId: parent.id, provider: "opencode", cwd: s.root, + launchOptions: { [MODELS]: { account: "glm-flash" }, [OTHER]: { pick: "x" } } }); + await s.settle(); + const failed = s.terminals.getMetadata(child.id); + assert.equal(failed.status, "failed"); + assert.match(failed.failureDetails, /Model accounts and Other models both set the model of this launch/u); + assert.equal(s.calls.length, spawned); +}); + +test("a person's own card keeps starting with a model and a plugin that sets the model, also when it is restored", async (t) => { + const s = await setup(t); + const card = s.terminals.create({ provider: "opencode", profile: "normal", cwd: s.root, position: at, model: "zai/glm-5.3", + launchOptions: { [MODELS]: { account: "glm-flash" } } }); + await s.settle(); + assert.notEqual(s.terminals.getMetadata(card.id).status, "failed"); + assert.ok(s.calls.at(-1).args.includes("acct_glm-flash/glm-5.3-flash")); + + const store = await mkdtemp(join(tmpdir(), "ctty-sub-account-store-")); + t.after(() => rm(store, { recursive: true, force: true })); + await new TerminalSessionStore(store).replace([{ + id: "person-card", provider: "opencode", profile: "normal", role: "agent", title: "Agent", titleCustomized: false, + cwd: s.root, position: at, size: { width: 700, height: 430 }, lastState: "running", restore: true, + model: "zai/glm-5.3", options: { [MODELS]: { account: "glm-flash" } } + }]); + const restored = await setup(t, { store }); + await restored.settle(); + const metadata = restored.terminals.getMetadata("person-card"); + assert.notEqual(metadata.status, "failed", metadata.failureDetails ?? ""); + assert.equal(restored.calls.length, 1); + assert.ok(restored.calls[0].args.includes("acct_glm-flash/glm-5.3-flash")); + assert.deepEqual(restored.terminals.modelSetByOf("person-card"), [MODELS]); +}); + +test("a subagent of another CLI is never checked against options the person chose for its orchestrator's CLI", async (t) => { + const s = await setup(t, { delegable: false }); + // EXTRAS is not delegable and does not set the model: an opencode orchestrator's choice is not a codex subagent's. + const parent = s.orchestrator({ [EXTRAS]: { flavor: "mild" } }); + await s.settle(); + const child = await s.spawn(parent, { provider: "codex" }); + assert.deepEqual(child.servedBy.launchOptions, {}); + assert.equal(s.terminals.launchOptionsOf(child.sessionId), undefined); + // Naming it for codex is the orchestrator's own choice, which this plugin does not allow. + await assert.rejects(s.spawn(parent, { provider: "codex", launchOptions: { [EXTRAS]: { flavor: "mild" } } }), + /Extras has not declared its launch options safe for an orchestrator to choose/u); +}); + +test("model-setting arguments are recognised in every form each CLI takes", () => { + const sets = (provider, ...args) => modelSettingArgument(provider, args); + for (const provider of ["opencode", "codex", "qwen", "kimi", "grok", "claude", "cursor"]) { + assert.equal(sets(provider, "--model", "x"), "--model", provider); + assert.equal(sets(provider, "--model=x"), "--model=x", provider); + } + for (const provider of ["opencode", "codex", "qwen", "kimi", "grok"]) { + assert.equal(sets(provider, "-m", "x"), "-m", provider); + assert.equal(sets(provider, "-mx"), "-mx", provider); + } + // Claude has no -m. + assert.equal(sets("claude", "-m", "x"), undefined); + // Codex: a profile, and configuration keys that pick the model or its provider. + assert.equal(sets("codex", "--profile", "work"), "--profile"); + assert.equal(sets("codex", "--profile=work"), "--profile=work"); + assert.equal(sets("codex", "-p", "work"), "-p"); + assert.equal(sets("codex", "-pwork"), "-pwork"); + assert.equal(sets("codex", "-c", "model=\"o3\""), "-c model=\"o3\""); + assert.equal(sets("codex", "-c", " model = o3"), "-c model = o3"); + assert.equal(sets("codex", "--config", "model_provider=acct"), "--config model_provider=acct"); + assert.equal(sets("codex", "--config=profile=work"), "--config=profile=work"); + assert.equal(sets("codex", "-cmodel=o3"), "-cmodel=o3"); + assert.equal(sets("codex", "-c", "\"model\"=o3"), "-c \"model\"=o3"); + // Other configuration keys and flags do not choose the model. + assert.equal(sets("codex", "-c", "model_reasoning_effort=high"), undefined); + assert.equal(sets("codex", "-c", "model"), undefined); + assert.equal(sets("codex", "--search", "-c", "tools.web_search=true"), undefined); + assert.equal(sets("opencode", "--profile", "x", "-p", "y"), undefined); + assert.equal(sets("opencode", "--models"), undefined); +}); + +test("without the delegable declaration a subagent gets exactly the inherited values, never the plugin's defaults", () => { + const pipeline = new LaunchPipeline({ + contributors: () => [{ pluginId: EXTRAS, pluginName: "Extras", serviceId: "extras", secrets: false, + launch: { fields: [{ key: "a", label: "A", kind: "text", default: "da" }, { key: "b", label: "B", kind: "text", default: "db" }, + { key: "c", label: "C", kind: "text" }] } }], + call: async () => null, secret: async () => null, runsRoot: join(tmpdir(), "ctty-unused-runs") + }); + const inherited = { [EXTRAS]: { a: "1", b: "2", c: "" } }; + const delegated = { delegated: true, inherited }; + // A subset is completed from the orchestrator's values, not from the defaults. + assert.deepEqual(pipeline.normalizeOptions("opencode", { [EXTRAS]: { a: "1" } }, delegated), { [EXTRAS]: { a: "1", b: "2", c: "" } }); + assert.deepEqual(pipeline.normalizeOptions("opencode", { [EXTRAS]: { a: "1", b: "2", c: "" } }, delegated), inherited); + const refused = /Extras has not declared its launch options safe for an orchestrator to choose/u; + assert.throws(() => pipeline.normalizeOptions("opencode", { [EXTRAS]: { a: "1", b: "x" } }, delegated), refused); + assert.throws(() => pipeline.normalizeOptions("opencode", { [EXTRAS]: { c: "set" } }, delegated), refused); + assert.throws(() => pipeline.normalizeOptions("opencode", { [EXTRAS]: { a: "1", extra: "1" } }, delegated), refused); + assert.throws(() => pipeline.normalizeOptions("opencode", { [EXTRAS]: {} }, delegated), refused); + // An orchestrator whose own values the plugin no longer accepts as they were is refused, not given defaults. + assert.throws(() => pipeline.normalizeOptions("opencode", { [EXTRAS]: { a: "1" } }, { delegated: true, inherited: { [EXTRAS]: { a: "1", b: "2" } } }), + /Extras launch options changed since this orchestrator started/u); + assert.throws(() => pipeline.normalizeOptions("opencode", { [EXTRAS]: { a: "1" } }, { delegated: true }), refused); +}); + +test("a plugin without fields chosen for the orchestrator is inherited as chosen, not refused", async (t) => { + const s = await setup(t, { delegable: false }); + const parent = s.orchestrator({ [BARE]: {} }); + await s.settle(); + assert.notEqual(s.terminals.getMetadata(parent.id).status, "failed"); + const child = await s.spawn(parent, {}); + await s.settle(); + assert.deepEqual(child.servedBy.launchOptions, { [BARE]: "inherited" }); + assert.deepEqual(s.terminals.launchOptionsOf(child.sessionId), { [BARE]: {} }); + assert.notEqual(s.terminals.getMetadata(child.sessionId).status, "failed"); +}); + +test("a plugin turned off after the orchestrator started: left out unless it set the model, then refused", async (t) => { + const s = await setup(t, { delegable: false }); + const parent = s.orchestrator({ [MODELS]: { account: "glm-flash" }, [EXTRAS]: { flavor: "mild" } }); + await s.settle(); + // Options that do not decide the model: the subagent starts without them. + s.state.gone.add(EXTRAS); + const child = await s.spawn(parent, {}); + await s.settle(); + assert.deepEqual(child.servedBy.launchOptions, { [MODELS]: "inherited" }); + assert.deepEqual(s.terminals.launchOptionsOf(child.sessionId), { [MODELS]: { account: "glm-flash" } }); + // The plugin that set the model: never a silent fall back to the CLI's default model. + s.state.gone.add(MODELS); + const spawned = s.calls.length; + await assert.rejects(s.spawn(parent, {}), + new RegExp(`model is set by its ${MODELS} launch options, and ${MODELS} is now disabled, removed or not trusted; a subagent would run on its CLI's own default model instead`, "u")); + assert.equal(s.calls.length, spawned); +}); + +test("a plugin whose fields changed after the orchestrator started refuses its subagents with that reason", async (t) => { + const s = await setup(t, { delegable: false }); + const parent = s.orchestrator({ [EXTRAS]: { flavor: "mild" } }); + await s.settle(); + s.state.extrasFields.splice(0, 1, { key: "taste", label: "Taste", kind: "text" }); + await assert.rejects(s.spawn(parent, {}), /Extras launch options changed since this orchestrator started, so its subagents cannot repeat the person's choice/u); + s.state.extrasFields.splice(0, 1, { key: "flavor", label: "Flavor", kind: "text" }, { key: "extra", label: "Extra", kind: "text", default: "on" }); + await assert.rejects(s.spawn(parent, {}), /Extras launch options changed since this orchestrator started/u); +}); diff --git a/tests/subagent-signal-exit.test.mjs b/tests/subagent-signal-exit.test.mjs new file mode 100644 index 000000000..c434e05c9 --- /dev/null +++ b/tests/subagent-signal-exit.test.mjs @@ -0,0 +1,118 @@ +import assert from "node:assert/strict"; +import { execFileSync } from "node:child_process"; +import { existsSync } from "node:fs"; +import { mkdtemp, rm } from "node:fs/promises"; +import { tmpdir } from "node:os"; +import { basename, join } from "node:path"; +import test from "node:test"; +import pty from "node-pty"; +import { AgentControlService } from "../src/main/services/AgentControlService.ts"; +import { TerminalManager } from "../src/main/services/TerminalManager.ts"; +import { persistedTerminalSession } from "../src/main/services/TerminalSessionStore.ts"; +import { failureSummary } from "../src/renderer/src/features/home/failureSummary.ts"; +import { availableRegistry } from "./helpers/terminal.mjs"; + +// A subagent whose CLI was killed with SIGKILL showed "done, exit 0", so wait_for_agent reported a crash as success: +// node-pty reports a signal death as { exitCode: 0, signal: 9 }. These tests start real processes (inside sandbox-exec on macOS, as the +// isolation layer wraps an agent there) and kill them. + +const at = { x: 0, y: 0 }; +const SANDBOX = "/usr/bin/sandbox-exec"; +const wrapped = process.platform === "darwin" && existsSync(SANDBOX); +const skip = process.platform === "win32" ? "POSIX signals only" : false; + +/** Every launch becomes the given real command tree, wrapped in sandbox-exec where the isolation layer would. */ +function realSpawner(argv, spawned) { + return (_command, _args, options) => { + const [command, ...args] = wrapped ? [SANDBOX, "-p", "(version 1)(allow default)", ...argv] : argv; + const child = pty.spawn(command, args, { name: "xterm-256color", cols: 80, rows: 24, cwd: options.cwd, env: { PATH: "/usr/bin:/bin", HOME: options.cwd } }); + spawned.push(child); + return child; + }; +} + +async function setup(t, argv) { + const root = await mkdtemp(join(tmpdir(), "ctty-signal-exit-")); + const spawned = []; + const terminals = new TerminalManager(() => undefined, availableRegistry(), undefined, undefined, true, realSpawner(argv, spawned)); + t.after(async () => { + terminals.disposeAll(); + for (const child of spawned) { try { process.kill(child.pid, "SIGKILL"); } catch { /* exited */ } } + await rm(root, { recursive: true, force: true }); + }); + const control = new AgentControlService(terminals); + const orchestrator = terminals.create({ provider: "opencode", profile: "normal", cwd: root, position: at, role: "orchestrator" }); + return { root, spawned, terminals, control, orchestrator }; +} + +async function until(check, timeoutMs = 5_000) { + const deadline = Date.now() + timeoutMs; + while (Date.now() < deadline) { + const value = check(); + if (value) return value; + await new Promise((resolve) => setTimeout(resolve, 25)); + } + throw new Error("timed out"); +} + +const descendants = (pid) => { + try { return execFileSync("/usr/bin/pgrep", ["-P", String(pid)], { encoding: "utf8" }).trim().split("\n").filter(Boolean).map(Number); } catch { return []; } +}; + +test("a subagent whose CLI is killed by SIGKILL is failed (128+9) with the signal named, everywhere its exit is read", { skip }, async (t) => { + const s = await setup(t, ["/bin/sleep", "30"]); + const child = await s.control.spawn({ parentSessionId: s.orchestrator.id, provider: "opencode", cwd: s.root }); + const live = s.spawned.at(-1); + // sandbox-exec execs the CLI: the PTY's own process is the CLI, not a wrapper that could exit 0 after it. + // (node-pty's spawn-helper and then sandbox-exec exec in place first.) + // macOS reports an executable path; Linux ps reports only the command name. + const command = await until(() => { const comm = execFileSync("/bin/ps", ["-o", "comm=", "-p", String(live.pid)], { encoding: "utf8" }).trim(); return basename(comm) === "sleep" ? comm : null; }); + assert.equal(basename(command), "sleep"); + assert.deepEqual(descendants(live.pid), [], "no process below the CLI"); + process.kill(live.pid, "SIGKILL"); + const ended = await until(() => { const m = s.terminals.getMetadata(child.id); return m.exitCode !== null ? m : null; }); + assert.equal(ended.status, "failed"); + assert.equal(ended.exitCode, 137); + assert.match(ended.failureDetails, /killed by signal SIGKILL \(9\)/u); + // wait_for_agent and get_agent_result read the same exit. + const waited = await s.control.waitFor(child.id, { timeoutMs: 1_000 }); + assert.equal(waited.reason, "failed"); + assert.equal(waited.status, "failed"); + assert.equal(waited.exitCode, 137); + const result = await s.control.result(child.id); + assert.equal(result.state, "failed"); + assert.equal(result.exitCode, 137); + // The saved session restores as failed, and HOME's failure summary shows the exit code with the signal line. + const saved = persistedTerminalSession(ended); + assert.equal(saved.lastState, "failed"); + assert.equal(saved.exitCode, 137); + const summary = failureSummary(ended, "en"); + assert.match(summary.headline, /137/u); + assert.match(summary.diagnostics, /^The process was killed by signal SIGKILL \(9\)\./u); +}); + +test("a wrapper shell whose CLI child is killed reports the signal as a non-zero exit: the subagent is failed", { skip }, async (t) => { + const s = await setup(t, ["/bin/sh", "-c", "/bin/sleep 30; exit $?"]); + const child = await s.control.spawn({ parentSessionId: s.orchestrator.id, provider: "opencode", cwd: s.root }); + const shell = s.spawned.at(-1); + const cli = await until(() => descendants(shell.pid)[0]); + process.kill(cli, "SIGKILL"); + const ended = await until(() => { const m = s.terminals.getMetadata(child.id); return m.exitCode !== null ? m : null; }); + assert.equal(ended.status, "failed"); + assert.equal(ended.exitCode, 137); + // The shell exited normally with 137: no signal reached node-pty, so nothing is named. + assert.doesNotMatch(ended.failureDetails ?? "", /killed by signal/u); + assert.equal((await s.control.waitFor(child.id, { timeoutMs: 1_000 })).reason, "failed"); +}); + +test("a clean exit stays done", { skip }, async (t) => { + const s = await setup(t, ["/usr/bin/true"]); + const child = await s.control.spawn({ parentSessionId: s.orchestrator.id, provider: "opencode", cwd: s.root }); + const ended = await until(() => { const m = s.terminals.getMetadata(child.id); return m.exitCode !== null ? m : null; }); + assert.equal(ended.status, "done"); + assert.equal(ended.exitCode, 0); + assert.equal(ended.failureDetails, null); + assert.equal((await s.control.waitFor(child.id, { timeoutMs: 1_000 })).reason, "done"); + assert.equal((await s.control.result(child.id)).state, "done"); + assert.equal(persistedTerminalSession(ended).lastState, "exited"); +}); diff --git a/tests/terminal-lifecycle.test.mjs b/tests/terminal-lifecycle.test.mjs index 8ca7f9c16..fc1b3e545 100644 --- a/tests/terminal-lifecycle.test.mjs +++ b/tests/terminal-lifecycle.test.mjs @@ -290,15 +290,31 @@ test("session metadata revisions advance before lifecycle events cross IPC", asy assert.match(source, /metadata\.revision \+= 1;[\s\S]{0,120}?this\.emit\(IPC\.terminalSession/); }); -test("revoking lifecycle hooks makes live agent status unavailable until a restarted session gets a new parser", async () => { +test("revoking lifecycle hooks makes live agent status unavailable until a restarted session gets a new parser", async (t) => { const source = await readFile(terminalManagerPath, "utf8"); - - assert.match(source, /setLifecycleHooksEnabled\(enabled: boolean\): void/); - assert.match(source, /this\.lifecycleHooksEnabled = next;\s*if \(next\) return;/); - assert.match(source, /session\.lifecycle = null;/); - assert.match(source, /session\.metadata\.provider === "terminal"/); - assert.match(source, /session\.metadata\.status = "unavailable";\s*this\.emitSession\(session\.metadata\)/); - assert.match(source, /if \(!this\.lifecycleHooksEnabled \|\| !session/); + const { TerminalManager } = await import("../src/main/services/TerminalManager.ts"); + const { availableRegistry, fakeSpawner } = await import("./helpers/terminal.mjs"); + const manager = new TerminalManager(() => undefined, availableRegistry(), undefined, undefined, true, fakeSpawner([])); + t.after(() => manager.disposeAll()); + const session = manager.create({ provider: "codex", profile: "normal", cwd: process.cwd(), position: { x: 0, y: 0 } }); + const shell = manager.create({ provider: "terminal", profile: "normal", cwd: process.cwd(), position: { x: 0, y: 0 } }); + const shellStatus = manager.getMetadata(shell.id).status; + manager.applyProviderSignal(session.id, { kind: "lifecycle", state: "working" }); + assert.equal(manager.getMetadata(session.id).status, "working"); + manager.setLifecycleHooksEnabled(false); + assert.equal(manager.getMetadata(session.id).status, "unavailable"); + assert.equal(manager.sessions.get(session.id).lifecycle, null); + assert.equal(manager.getMetadata(shell.id).status, shellStatus, "a plain terminal has no agent status to revoke"); + for (const state of ["needs_approval", "working", "idle"]) { + manager.applyProviderSignal(session.id, { kind: "lifecycle", state }); + assert.equal(manager.getMetadata(session.id).status, "unavailable", "a hook signal never restores a disabled status"); + } + manager.setLifecycleHooksEnabled(true); + manager.applyProviderSignal(session.id, { kind: "lifecycle", state: "working" }); + assert.equal(manager.getMetadata(session.id).status, "unavailable", "re-enabling applies only to new or restarted sessions"); + const fresh = manager.create({ provider: "codex", profile: "normal", cwd: process.cwd(), position: { x: 0, y: 0 } }); + manager.applyProviderSignal(fresh.id, { kind: "lifecycle", state: "working" }); + assert.equal(manager.getMetadata(fresh.id).status, "working"); assert.match(source, /session\.lifecycle = this\.lifecycleHooksEnabled\s*\? createProviderLifecycleParser/); }); @@ -306,8 +322,9 @@ test("only explicit turn-stop hooks set completed art and a new turn clears it", const [card, manager, main] = await Promise.all([ readFile(terminalCardPath, "utf8"), readFile(terminalManagerPath, "utf8"), - readFile(new URL("../src/main/index.ts", import.meta.url), "utf8") + readFile(new URL("../src/main/services/agent-runtime/applyRuntimeSignal.ts", import.meta.url), "utf8") ]); + // main/index.ts hands every runtime signal to applyRuntimeSignal, which passes the hook's event on. assert.match(main, /state: signal\.state,\s*event: signal\.event/); assert.match(manager, /\["Stop", "StopFailure", "StopCancelled"\]\.includes\(signal\.event \?\? ""\)/); assert.match(manager, /nextStatus === "working" \? false : completed \|\| Boolean\(session\.metadata\.turnCompleted\)/); @@ -377,7 +394,6 @@ test("failed PTYs preserve their final sanitized output as failure details", asy // Masked whole before the last lines are chosen (a cut inside a secret would leave its tail readable). assert.match(source, /terminalFailureDetails\(this\.redactSecrets\(current\.bufferChunks\.slice\(current\.bufferStart\)\.join\(""\)\)\)/); - assert.match(source, /current\.metadata\.failureDetails = exitCode === 0/); }); function effectDependenciesContaining(source, marker) { diff --git a/tests/usage-collector.test.mjs b/tests/usage-collector.test.mjs index ba0172ed4..8e8e385f0 100644 --- a/tests/usage-collector.test.mjs +++ b/tests/usage-collector.test.mjs @@ -1,6 +1,6 @@ import test from 'node:test'; import assert from 'node:assert/strict'; -import { execFileSync } from 'node:child_process'; +import { execFileSync, spawn } from 'node:child_process'; import { createHash, randomUUID } from 'node:crypto'; import { existsSync } from 'node:fs'; import { appendFile, chmod, mkdir, mkdtemp, readFile, readdir, rename, rm, symlink, utimes, writeFile } from 'node:fs/promises'; @@ -193,7 +193,7 @@ test('Hermes queries select accounting columns only, read-only, and never touch }).collect(T0); const statements = argumentsSeen.filter((argument) => /^\s*SELECT/i.test(argument)); assert.equal(statements.length, 2); - assert.ok(argumentsSeen.includes('-readonly') && argumentsSeen.includes('-safe') && argumentsSeen.some((a) => /^file:.*\?mode=ro&immutable=1$/.test(a))); + assert.ok(argumentsSeen.includes('-readonly') && argumentsSeen.includes('-safe') && argumentsSeen.some((a) => /^file:.*\?mode=ro$/.test(a))); for (const statement of statements) { assert.doesNotMatch(statement, /\*|system_prompt|title|messages|content|INSERT|UPDATE|DELETE|CREATE|DROP|ATTACH|PRAGMA\s/i); } @@ -228,6 +228,98 @@ test('Hermes WAL stores: live writer, stopped writer without -wal/-shm, and orph assert.deepEqual((await readdir(dir)).sort(), ['state.db', 'state.db-wal']); }); +/** A collector whose sqlite3 calls are recorded (database URI per call) and can be made to fail first. */ +function recordingCollector(home, uris, failFirst = 0) { + let failures = failFirst; + return new LocalUsageCollector(home, {}, {}, { + sqlite: SQLITE, + sqliteRunner: async (file, args, options) => { + uris.push(args.find((argument) => argument.startsWith('file:'))); + if (failures > 0) { + failures--; + throw Object.assign(new Error('Command failed'), { stderr: 'Error: in prepare, database is locked (5)' }); + } + return { stdout: execFileSync(file, args, { ...options, encoding: 'utf8' }) }; + } + }); +} + +/** + * A sqlite3 process holding `statements` open in an exclusive transaction. The writer runs in its own process: POSIX + * locks are per process, and this process opening and closing the database (snapshots, header reads) would drop them. + */ +async function exclusiveWriter(t, db, statements) { + const child = spawn(SQLITE, ['-init', devNull, '-batch', db], { env: {}, stdio: ['pipe', 'pipe', 'pipe'] }); + t.after(() => { if (child.exitCode === null && child.signalCode === null) child.kill(); }); + let output = ''; + const ready = new Promise((resolve, reject) => { + child.stdout.on('data', (chunk) => { output += chunk; if (output.includes('writer-ready')) resolve(); }); + child.once('exit', () => reject(new Error(`writer exited early: ${output}`))); + }); + child.stdin.write(`BEGIN EXCLUSIVE; ${statements}; SELECT 'writer-ready';\n`); + await ready; + return async () => { + const exited = new Promise((resolve) => child.once('exit', resolve)); + child.stdin.end('COMMIT;\n.quit\n'); + await exited; + }; +} + +test('Hermes rollback-journal stores are read with locks, never as immutable; a busy writer is retried, then skipped', { skip: needsSqlite }, async (t) => { + const home = await fixture(t); + const dir = join(home, '.hermes'); + const path = await hermesDb(dir, session('r', 'openai-codex', [0, 0, 0, 0], T0 - DAY) + modelRow('r', 'gpt-5.5', 'openai-codex', '', [100, 10, 0, 0], T0 - DAY, T0 - DAY)); + const header = await readFile(path); + assert.deepEqual([header[18], header[19]], [1, 1], 'the fixture is a rollback-journal database'); + const uris = []; + const collector = recordingCollector(home, uris); + await collectUntouched(collector, T0, home); + assert.ok(uris.length > 0 && uris.every((uri) => /\?mode=ro$/.test(uri)), `rollback-journal reads take locks: ${uris.join(' ')}`); + + // A writer in the middle of a transaction holds the exclusive lock: an immutable read would ignore it and could + // see half-written pages. The locked read waits, retries a bounded number of times, then skips this poll. + const commit = await exclusiveWriter(t, path, 'UPDATE session_model_usage SET input_tokens = 150'); + uris.length = 0; + // Not collectUntouched: the writer's rollback journal exists while its transaction is open. + const busy = await collector.collect(T0 + MINUTE); + assert.deepEqual(busy.events, []); + assert.match(busy.coverage.join('\n'), /Hermes default: database is busy; not collected this time\./); + assert.equal(uris.length, 3, 'three attempts, then the poll moves on'); + assert.deepEqual((await readdir(dir)).sort(), ['state.db', 'state.db-journal'], 'only the writer\'s own journal exists'); + await commit(); + assert.deepEqual((await readdir(dir)).sort(), ['state.db'], 'no side files remain'); + + const after = await collectUntouched(collector, T0 + 2 * MINUTE, home); + assert.deepEqual(after.events.map((e) => e.input), [50], 'the committed change is counted on the next poll'); + assert.deepEqual((await readdir(dir)).sort(), ['state.db'], 'reading created no side files'); + + // A transient lock is retried within the same poll. + const retried = []; + sql(path, `UPDATE session_model_usage SET input_tokens = 170`); + const second = recordingCollector(home, retried, 1); + const first = await collectUntouched(second, T0 + 3 * MINUTE, home); + assert.doesNotMatch(first.coverage.join('\n'), /busy/); + assert.equal(retried.length, 3, 'one failed attempt, then the schema and the accounting query'); +}); + +test('Hermes WAL stores without side files keep the immutable read', { skip: needsSqlite }, async (t) => { + const { DatabaseSync } = await import('node:sqlite').catch(() => ({})); + if (!DatabaseSync) return t.skip('node:sqlite unavailable'); + const home = await fixture(t); + const dir = join(home, '.hermes'), path = join(dir, 'state.db'); + await mkdir(dir); + const writer = new DatabaseSync(path); + writer.exec('PRAGMA journal_mode=WAL;' + SCHEMA + session('w', 'openai-codex', [0, 0, 0, 0], T0 - DAY)); + writer.close(); + assert.deepEqual((await readdir(dir)).sort(), ['state.db']); + const header = await readFile(path); + assert.deepEqual([header[18], header[19]], [2, 2], 'the fixture is a WAL database'); + const uris = []; + await collectUntouched(recordingCollector(home, uris), T0, home); + assert.ok(uris.length > 0 && uris.every((uri) => /\?mode=ro&immutable=1$/.test(uri)), uris.join(' ')); + assert.deepEqual((await readdir(dir)).sort(), ['state.db']); +}); + // ---- JSONL fixtures const NOW = Math.floor(Date.now() / 1000) * 1000; const tokens = ([input, cached, output]) => ({ input_tokens: input, cached_input_tokens: cached, output_tokens: output, reasoning_output_tokens: Math.floor(output / 2), total_tokens: input + output }); diff --git a/tests/usage-history-service.test.mjs b/tests/usage-history-service.test.mjs index 82cfce4a8..3bb6960d6 100644 --- a/tests/usage-history-service.test.mjs +++ b/tests/usage-history-service.test.mjs @@ -1,7 +1,8 @@ import assert from "node:assert/strict"; import { execFileSync } from "node:child_process"; -import { appendFile, chmod, lstat, mkdir, mkdtemp, readdir, readFile, rm, stat, symlink, writeFile } from "node:fs/promises"; +import { appendFile, chmod, lstat, mkdir, mkdtemp, readdir, readFile, rename, rm, stat, symlink, writeFile } from "node:fs/promises"; import { tmpdir, devNull } from "node:os"; +import { createRequire, syncBuiltinESMExports } from "node:module"; import { join } from "node:path"; import test from "node:test"; import { LocalUsageCollector, discoverSqlite } from "../src/main/services/LocalUsageCollector.ts"; @@ -353,6 +354,58 @@ test("structurally invalid history is quarantined the same way", async () => { }); }); +const fsPromises = createRequire(import.meta.url)("node:fs/promises"); + +/** Runs `run` while every open of `path` first calls `before` and its handle calls `onClose` before closing. */ +async function withOpenHook(t, path, { before = async () => undefined, onClose = async () => undefined }, run) { + const realOpen = fsPromises.open; + t.mock.method(fsPromises, "open", async (target, ...rest) => { + if (target !== path) return realOpen(target, ...rest); + await before(); + const handle = await realOpen(target, ...rest); + const close = handle.close.bind(handle); + handle.close = async () => { await onClose(); return close(); }; + return handle; + }); + syncBuiltinESMExports(); + try { + await run(); + } finally { + t.mock.restoreAll(); + syncBuiltinESMExports(); + } +} + +test("a history file swapped between the symlink check and the open is refused and left alone", async (t) => { + await withDirectory(async (directory) => { + const path = join(directory, FILE); + await writeFile(path, "{}"); + await writeFile(join(directory, "planted.json"), "planted"); + const { history } = service(directory); + await withOpenHook(t, path, { before: () => rename(join(directory, "planted.json"), path) }, () => history.load()); + assert.match(history.get().error, /could not be read \(ELOOP\)/); + assert.equal(await readFile(path, "utf8"), "planted", "the swapped-in file is not touched"); + }); +}); + +test("the preserved original is the content that was read, even if the path is swapped afterwards", async (t) => { + await withDirectory(async (directory) => { + const original = JSON.stringify(envelope( + { events: [event("good", { from: T0 - DAY, to: T0 - DAY }), event("bad", { input: -5 })] }, + { version: 2, files: { [KEY]: cursor(1) }, hermes: {} } + )); + const path = join(directory, FILE); + await writeFile(path, original); + await writeFile(join(directory, "planted.json"), "planted"); + const { history } = service(directory); + await withOpenHook(t, path, { onClose: () => rename(join(directory, "planted.json"), path) }, () => history.load()); + const preserved = (await readdir(directory)).filter((name) => name.startsWith("usage-history.corrupt-")); + assert.equal(preserved.length, 1); + assert.equal(await readFile(join(directory, preserved[0]), "utf8"), original); + if (process.platform !== "win32") assert.equal((await stat(join(directory, preserved[0]))).mode & 0o777, 0o600); + }); +}); + test("partially invalid records are dropped openly and the original file is preserved", async () => { await withDirectory(async (directory) => { const original = JSON.stringify(envelope(