Jev Skill Router — Security skill for Claude Code
Jev-powered skill router & security auditor for any AI agent (Codex, Claude Code, OpenCode, Hermes): ONE cheap decision per request tells the model WHICH skill to load; scans skill libraries for promp.
How to install Jev Skill Router
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open aleksvega/jev-skill-router and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Jev Skill Router does
Jev-powered skill router & security auditor for any AI agent (Codex, Claude Code, OpenCode, Hermes): ONE cheap decision per request tells the model WHICH skill to load; scans skill libraries for prompt injection & dangerous commands.
Alternatives in Security
- Deepsec — Deepsec is a security harness for finding vulnerabilities in your codebase powered by coding agents 7.8k ★
- Claude Code Security Review — An AI-powered security review GitHub Action using Claude to analyze code changes for security vulnerabilities 3.9k ★
- OpenTag — Open-source, channel-native agent gateway for Slack 499 ★
README
jev-skill-router
[](https://www.npmjs.com/package/jev-skill-router) [](LICENSE) [](https://nodejs.org) [-orange)](https://github.com/tamaratran/fast-jev-compaction)
**A Jev-powered skill router & security auditor for any AI agent** — Codex, Claude Code, OpenCode, Hermes Agent, anything with `SKILL.md` skills.
The problem: agents have dozens of skills, but the model *forgets to use them* — or uses the wrong one. jev-skill-router sends ONE cheap Jev (System One) decision request per user request and returns a short instruction: *which skill to load, how complex the request is, how confident the choice is.*
Second problem it solves: **skill supply-chain safety.** Install skills from any GitHub repo, then run a Jev security audit that flags **prompt injection** and **dangerous shell commands** per skill.
Built on the public Jev endpoint (`POST https://openrouter.ai/api/alpha/decisions`, model `typesafe/jev-1.13`) with a normal OpenRouter key. One decision costs ~$0.00002 and takes ~0.5 s.
Install (one command)
npm i -g jev-skill-router
export OPENROUTER_API_KEY=sk-or-v1-... # key from env only, never hardcodedWhy not just "list skills in AGENTS.md"?
| Approach | Cost / request | Latency | Stays in sync with skill library | Safety scan |
|---|---|---|---|---|
| Full skill catalog pasted in AGENTS.md | tokens EVERY request, grows with library | — | manual re-paste | ❌ |
| "The model remembers" (system prompt) | tokens, degrades with 100+ skills | — | ❌ model forgets | ❌ |
| LLM call to route | $0.01–0.10 | 3–10 s | manual | ❌ |
| jev-skill-router (Jev) | ~$0.00002 | ~0.5–1.5 s | auto (reads SKILL.md dirs) | ✅ built-in |
Use
# Tell the agRelated Skills
Hermes Claude Code Bridge
Run Claude Code or OpenCode as an OpenAI-compatible backend — a security-first, read-only Lead DevOps/SRE agen
Danshari Skill
🧹 断舍离.skill / danshari.skill — 第一个劝你删 skill 的 skill：审计并移除你的 agent 不再需要的 skill（模型已会的 / harness 已自带的 / MCP 已接管的
Adversarial Testing
Use when authoring evals/predicates (Scout, TDD-Engineer), framing an audit (Auditor), or justifying which pha
Reviewer Gap Audit
You are the compact global omission auditor for a decomposed pull-request review. Primary specialists already
Route Status
Show recent auto-router decisions and a rolling distribution of model routings. Reads from ~/.claude/cache/rou
Skill Auditor Seguranca
Executable security and LGPD gate for AI-built projects. Scans for exposed secrets, .env leaks, Supabase servi
Related Agents
Payment Tracker
Scans your Gmail for payment confirmations and tells you what's been paid on your behalf.
Ensemble Curator
High-volume triage agent for external-AI-agent PRs (Codex, Jules, Hermes, Droid, Aider, etc.). Analyzes, conso
Agent Of Empires
Claude Code, OpenCode, Mistral Vibe, Codex CLI, Gemini CLI Coding Agent Terminal Session manager via tmux and
