From 5d8f53e4fc9969d0f2c2a98c854ff90204dbe486 Mon Sep 17 00:00:00 2001 From: Asaf Mazuz Date: Wed, 23 Sep 2026 16:31:17 +0300 Subject: [PATCH] Add jev-guard (klauswg) listing Catalog curator listing for https://github.com/klauswg/jev-guard. AI-assisted source review; no live TypeSafe spend. --- README.md | 1 + community/projects/tools/README.md | 1 + community/projects/tools/klauswg-jev-guard.md | 47 +++++++++++++++++++ 3 files changed, 49 insertions(+) create mode 100644 community/projects/tools/klauswg-jev-guard.md diff --git a/README.md b/README.md index b756802e..5b5eb59f 100644 --- a/README.md +++ b/README.md @@ -328,6 +328,7 @@ The [independent model research](community/projects/tools/README.md#independent- - [jev-harness (TypeSafeAI)](https://github.com/TypeSafeAI/jev-harness) - Research proposal-review harness: LLM proposes, Jev answers narrow questions, code leaves host evidence (TypeSafeAI community; distinct from AntonioCoppe/jev-harness). [Project guide](community/projects/tools/typesafeai-jev-harness.md). - [jev-guard](https://github.com/leepokai/jev-guard) - Coding-agent security hooks: TypeSafe Jev risk-scores tool calls with session context (deny/ask/allow), flags prompt injection, and scans skills across Claude Code, Codex, Cursor, and more. [Project guide](community/projects/tools/jev-guard.md). - [jev-guard (CMaintz)](https://github.com/CMaintz/jev-guard) - Portable TypeSafe Jev allow/block/hold guard for agent tool calls with LangChain and Vercel AI SDK adapters (distinct from leepokai/jev-guard). [Project guide](community/projects/tools/cmaintz-jev-guard.md). +- [jev-guard (klauswg)](https://github.com/klauswg/jev-guard) - Exchange deposit/withdrawal risk triage: TypeSafe Jev judges; Java code adjudicates (distinct from leepokai/CMaintz jev-guard). [Project guide](community/projects/tools/klauswg-jev-guard.md). - [JevGuard](https://github.com/Jhonnyr97/JevGuard) - Claude Code/Codex plugin: generate project rules from CLAUDE.md/AGENTS.md and enforce them with TypeSafe Jev on PreToolUse/Stop (distinct from jev-guard risk scoring). [Project guide](community/projects/tools/jevguard.md). - [jev-drone](https://github.com/RomanSlack/jev-drone) - Explores Jev tactical judgments in a MuJoCo quadrotor simulation, with local flight control and a separate experimental tunnel policy. [Project guide](community/projects/tools/jev-drone.md). - [jev-effort-router](https://github.com/AlphaPerseii3000/jev-effort-router) - Hermes plugin for Ollama:Cloud: TypeSafe Jev routes both model and reasoning effort per turn via `llm_request` rewrite. [Project guide](community/projects/tools/jev-effort-router.md). diff --git a/community/projects/tools/README.md b/community/projects/tools/README.md index 10f3e59d..ca3a599b 100644 --- a/community/projects/tools/README.md +++ b/community/projects/tools/README.md @@ -164,6 +164,7 @@ See the [computer-use guide](../../../docs/computer-use.md) for a comparison, fo | [jev-harness (TypeSafeAI)](typesafeai-jev-harness.md) | Research proposal-review contract: LLM proposes, Jev answers four narrow questions, code emits host evidence (distinct from AntonioCoppe/jev-harness). | TypeScript · source-only (`jev-harness` 0.0.0, MIT) | | [jev-guard](jev-guard.md) | Risk-score coding-agent tool calls with TypeSafe Jev (deny/ask/allow), flag injection in results, and scan skills across many agents. | Node.js · npm CLI/hooks (`jev-guard` 0.3.1) | | [jev-guard (CMaintz)](cmaintz-jev-guard.md) | Framework-agnostic allow/block/hold tool-call guard with TypeSafe Jev + LangChain/Vercel adapters (distinct from leepokai jev-guard). | TypeScript · library (`jev-guard` 0.1.0, MIT) | +| [jev-guard (klauswg)](klauswg-jev-guard.md) | Exchange deposit/withdrawal risk triage: TypeSafe Jev answers; Java hard rules and gates decide (distinct from coding-agent jev-guard). | Java · Spring Boot demo (MIT) | | [JevGuard](jevguard.md) | Enforce CLAUDE.md/AGENTS.md-derived rules on Claude Code/Codex via TypeSafe Jev PreToolUse/Stop hooks (distinct from jev-guard risk firewall). | TypeScript · Claude/Codex plugin (`jevguard` 0.1.0) | | [jev-in-codex](jev-in-codex.md) | Rank Codex capabilities, search hits, and output excerpts with TypeSafe Jev via local MCP. | TypeScript · MCP server + Codex plugin (0.1.0) | | [jev-issue-radar](jev-issue-radar.md) | Find duplicate/related GitHub issues with TypeSafe Jev evidence choices in a local read-only dashboard. | Node.js · loopback server + static UI (0.1.1) | diff --git a/community/projects/tools/klauswg-jev-guard.md b/community/projects/tools/klauswg-jev-guard.md new file mode 100644 index 00000000..d845e6ca --- /dev/null +++ b/community/projects/tools/klauswg-jev-guard.md @@ -0,0 +1,47 @@ +# jev-guard (klauswg) + +[All projects](../README.md) · [Developer tools](README.md#developer-tools) + +Real-time risk triage gateway for exchange deposits/withdrawals: TypeSafe Jev answers risk/pattern/freeze questions; Java code owns hard rules, thresholds, and final actions. Distinct from [leepokai/jev-guard](jev-guard.md) and [CMaintz/jev-guard](cmaintz-jev-guard.md). + +| At a glance | Details | +| --- | --- | +| Source | [Source](https://github.com/klauswg/jev-guard) | +| Maintainer | [klauswg](https://github.com/klauswg). Independently curated; this entry is not an upstream submission or endorsement. | +| Format | Java 17+ Spring Boot demo service (`com.jevguard`) with mock and live TypeSafe clients. | +| Requirements | JDK 17+, Maven; optional `TYPESAFE_API_KEY` for live mode. Mock mode runs the full pipeline offline. | +| License | [MIT](https://github.com/klauswg/jev-guard/blob/067a17073f0a7fd5f5fc18508bf9aec49ce46ac2/LICENSE). | +| Disclosure | AI-assisted catalog review; no affiliation. Listing is not an endorsement. Source inspected (README, LICENSE, Java layout). Live TypeSafe calibration and Spring Boot run were **not** executed on the review host. Upstream n=100 calibration is synthetic/author-labeled—not a real AML detection rate. | + +## When to use + +Use it as a **reference architecture** for Jev-as-triage + deterministic adjudication on irreversible money flows. Prefer coding-agent [jev-guard](jev-guard.md) / [cmaintz-jev-guard](cmaintz-jev-guard.md) for tool-call firewalls—not this finance demo. + +## How it works + +FeatureBuilder digests a 24h window → HardRuleEngine may veto without a model call → StateRenderer builds sanitized English state → Jev answers risk_level / pattern / should_freeze / observation-only route → GateLogic composes AUTO_PASS / MANUAL_REVIEW / FREEZE. Failures degrade to manual review, never auto-pass. JSONL audit masks addresses. + +## Get started + +```sh +git clone https://github.com/klauswg/jev-guard.git +cd jev-guard +git checkout 067a17073f0a7fd5f5fc18508bf9aec49ce46ac2 +mvn spring-boot:run +# second terminal: ./demo.sh # mock fixtures, no API key +``` + +## Examples and demos + +- `./demo.sh` / `demo.ps1` simulated events; `GET /v1/metrics`, `GET /v1/decisions`. +- `docs/calibration-report.md`, `eval/labeled-samples.jsonl` (author-reported). + +## Limits and data handling + +Demo binds localhost; blacklist is sample data, not a live sanctions feed; in-memory profiles reset on restart. Live mode sends transfer feature text to TypeSafe. Not production-ready without your auth and data sources. This listing did not call live Jev. + +## Review and maintenance + +Reviewed on **2026-09-23** at [commit 067a170](https://github.com/klauswg/jev-guard/tree/067a17073f0a7fd5f5fc18508bf9aec49ce46ac2) (MIT). AI-assisted source review of README, LICENSE, package layout. No live TypeSafe spend. + +Related: [jev-guard](jev-guard.md), [cmaintz-jev-guard](cmaintz-jev-guard.md), [jev-java](jev-java.md).