Finish AgentCTL, Gateway, desktop, and Polylogue runtime integration

.agentctl/project.toml

@@ -8,6 +8,7 @@ root_markers = ["flake.nix", "modules", "dots"]
 [environment]
 kind = "nix-develop"
 command = ["nix", "develop", "--accept-flake-config", "--command"]
+preflight = ["true"]
 inherit = ["HOME", "USER", "LANG", "TERM", "SSH_AUTH_SOCK", "XDG_RUNTIME_DIR", "DBUS_SESSION_BUS_ADDRESS", "WAYLAND_DISPLAY", "DISPLAY"]
 unset = ["PYTHONPATH", "PYTHONHOME", "VIRTUAL_ENV", "_PYTHON_SYSCONFIGDATA_NAME", "_PYTHON_HOST_PLATFORM", "PYTHONPYCACHEPREFIX"]
 

README.md

@@ -17,7 +17,7 @@ local data systems, and frequent operational changes.
 | Area                 | Current scope                                                                                                                                                        |
 | -------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
 | Workstation          | Hyprland and Noctalia desktop, Home Manager configuration, GPU modes, audio, terminal capture, local AI services, development environments, and desktop applications |
-| Services             | typed service inventory, systemd resource classes, monitoring, and common policy for user and system units                                                         |
+| Services             | typed service inventory, systemd resource classes, monitoring, and common policy for user and system units                                                           |
 | Local data           | Sinex, Polylogue, Lynchpin, ActivityWatch, machine telemetry, shell history, and terminal recordings                                                                 |
 | Storage and recovery | impermanence, explicit persistence, Btrfs snapshots, Borg archives, restore drills, and separate treatment for durable, rebuildable, and bulk data                   |
 | Agent tooling        | shared instructions and skills, generated MCP profiles, browser and desktop control, local model backends, and a trusted repository gateway                          |
@@ -131,15 +131,15 @@ nix develop
 
 Common commands:
 
-| Command       | Purpose                                                                       |
-| ------------- | ----------------------------------------------------------------------------- |
-| `check`       | run the curated default verification tier sequentially                        |
-| `lint`        | run static Nix and shell checks without modifying files                       |
-| `format`      | format supported source with treefmt                                          |
+| Command       | Purpose                                                                        |
+| ------------- | ------------------------------------------------------------------------------ |
+| `check`       | run the curated default verification tier sequentially                         |
+| `lint`        | run static Nix and shell checks without modifying files                        |
+| `format`      | format supported source with treefmt                                           |
 | `switch`      | build and activate the workstation through the shared lock and resource policy |
-| `boot`        | build and register the next boot generation without activating it             |
-| `test-system` | test activation without changing the boot default                             |
-| `test-vm`     | build the NixOS VM smoke test                                                 |
+| `boot`        | build and register the next boot generation without activating it              |
+| `test-system` | test activation without changing the boot default                              |
+| `test-vm`     | build the NixOS VM smoke test                                                  |
 
 Direct commands stay direct. Submit scheduled or heavy work as a named
 AgentCTL project operation; `switch` remains the supported activation command.

docs/agent-hook-parity.md

@@ -1,15 +1,15 @@
 # Agent hook parity
 
-This matrix records the current boundary between Claude Code and Codex hooks. It is reviewed against the generated Codex file at each configuration change.
+This matrix records the current boundary between Claude Code and Codex hooks. It is reviewed against the generated Codex file at each configuration change, including the configured primary Polylogue hooks spool shared by every writer.
 
-| Capability | Claude Code | Codex | Evidence and action |
-| --- | --- | --- | --- |
-| Session start and resume | Enforced | Enforced | Both clients run configured recall and Polylogue session capture. |
-| User prompt capture | Enforced | Enforced | Both clients send `UserPromptSubmit` to Polylogue. |
-| Pre-compaction handoff | Enforced | Enforced | Both clients run `sinnix-context-handoff`. |
-| Tool-use evidence | Enforced | Enforced | Both clients send tool events to Polylogue. |
-| Stop capture | Enforced | Enforced | Both clients capture the stop event through Polylogue. |
-| Agent model dispatch guard | Enforced | Manual | Claude has the structured `Agent` matcher and `pretooluse-agent-model.sh`; Codex applies the explicit-model contract at AgentCTL launch. |
-| Destructive Bash guard | Enforced | Manual | Claude has the structured `Bash` matcher and `pretooluse-bash.sh`; Codex follows the shared Git protocol. |
+| Capability                 | Claude Code | Codex    | Evidence and action                                                                                                                      |
+| -------------------------- | ----------- | -------- | ---------------------------------------------------------------------------------------------------------------------------------------- |
+| Session start and resume   | Enforced    | Enforced | Both clients run configured recall and Polylogue session capture.                                                                        |
+| User prompt capture        | Enforced    | Enforced | Both clients send `UserPromptSubmit` to Polylogue.                                                                                       |
+| Pre-compaction handoff     | Enforced    | Enforced | Both clients run `sinnix-context-handoff`.                                                                                               |
+| Tool-use evidence          | Enforced    | Enforced | Both clients send tool events to Polylogue.                                                                                              |
+| Stop capture               | Enforced    | Enforced | Both clients capture the stop event through Polylogue.                                                                                   |
+| Agent model dispatch guard | Enforced    | Manual   | Claude has the structured `Agent` matcher and `pretooluse-agent-model.sh`; Codex applies the explicit-model contract at AgentCTL launch. |
+| Destructive Bash guard     | Enforced    | Manual   | Claude has the structured `Bash` matcher and `pretooluse-bash.sh`; Codex follows the shared Git protocol.                                |
 
 The generated Codex configuration is the authority for Codex-supported rows. This document is a parity record, not a second hook registry.

docs/agent-skill-executables.md

@@ -2,19 +2,19 @@
 
 This audit records every executable beneath `dots/_ai/skills`. AgentCTL is the only retained lifecycle owner. The allowed native runner is a private argv translator called by Sinnixd; no skill executable owns job state, cancellation, retries, manifests, or process cleanup.
 
-| Executable | Positive caller | Retained role |
-| --- | --- | --- |
-| `orchestrate/scripts/build_plan_batch_prompts.py` | `orchestrate` skill | Writes operator-reviewed prompt files only. |
-| `orchestrate/scripts/probe_agent_runtime.sh` | `orchestrate` skill and runtime-modes reference | Direct vendor and quota availability probe. |
-| `agent-runtime/scripts/run_agent_prompt.sh` | Sinnixd service contract | Private backend argv translator for an attested job. |
-| `chatgpt-conversations/scripts/sinnix-chatgpt-conversations` | `chatgpt-conversations` skill | Read-only conversation extraction from the shared browser. |
-| `desktop-control-plane/scripts/chrome-control.sh` | Agent module and desktop-control-plane skill | Visible browser UI helper. |
-| `desktop-control-plane/scripts/hypr-control.sh` | Agent module and desktop-control-plane skill | Visible window-manager UI helper. |
-| `desktop-control-plane/scripts/keyboard-control.sh` | Agent module | Explicit keyboard UI helper. |
-| `desktop-control-plane/scripts/kitty-remote-control.sh` | Agent module and desktop-control-plane skill | Visible terminal UI helper. |
-| `desktop-control-plane/scripts/screenshot-color-lab.sh` | Agent module and capture registry | Screenshot and display diagnostic helper. |
-| `grok/scripts/partition_by_size.sh` | `grok` skill | Measured source partitioning for a code audit. |
-| `html-report/generators/embed-path-popups.py` | `html-report` skill | HTML report navigation generator. |
-| `investigate/scripts/freeze.sh` | `investigate` skill | Evidence preservation before recovery mutation. |
-| `investigate/scripts/recover-probe.sh` | `investigate` skill | Read-only recovery authority probe. |
-| `skill-authoring/scripts/validate_skill.py` | `skill-authoring` skill and agent-tools check | Skill metadata and structure validation. |
+| Executable                                                   | Positive caller                                 | Retained role                                              |
+| ------------------------------------------------------------ | ----------------------------------------------- | ---------------------------------------------------------- |
+| `orchestrate/scripts/build_plan_batch_prompts.py`            | `orchestrate` skill                             | Writes operator-reviewed prompt files only.                |
+| `orchestrate/scripts/probe_agent_runtime.sh`                 | `orchestrate` skill and runtime-modes reference | Direct vendor and quota availability probe.                |
+| `agent-runtime/scripts/run_agent_prompt.sh`                  | Sinnixd service contract                        | Private backend argv translator for an attested job.       |
+| `chatgpt-conversations/scripts/sinnix-chatgpt-conversations` | `chatgpt-conversations` skill                   | Read-only conversation extraction from the shared browser. |
+| `desktop-control-plane/scripts/chrome-control.sh`            | Agent module and desktop-control-plane skill    | Visible browser UI helper.                                 |
+| `desktop-control-plane/scripts/hypr-control.sh`              | Agent module and desktop-control-plane skill    | Visible window-manager UI helper.                          |
+| `desktop-control-plane/scripts/keyboard-control.sh`          | Agent module                                    | Explicit keyboard UI helper.                               |
+| `desktop-control-plane/scripts/kitty-remote-control.sh`      | Agent module and desktop-control-plane skill    | Visible terminal UI helper.                                |
+| `desktop-control-plane/scripts/screenshot-color-lab.sh`      | Agent module and capture registry               | Screenshot and display diagnostic helper.                  |
+| `grok/scripts/partition_by_size.sh`                          | `grok` skill                                    | Measured source partitioning for a code audit.             |
+| `html-report/generators/embed-path-popups.py`                | `html-report` skill                             | HTML report navigation generator.                          |
+| `investigate/scripts/freeze.sh`                              | `investigate` skill                             | Evidence preservation before recovery mutation.            |
+| `investigate/scripts/recover-probe.sh`                       | `investigate` skill                             | Read-only recovery authority probe.                        |
+| `skill-authoring/scripts/validate_skill.py`                  | `skill-authoring` skill and agent-tools check   | Skill metadata and structure validation.                   |