Jevc — AI skill for Claude Code
Compile agent policy prose into deterministic verdict programs: narrow evidence questions for the model, the verdict computed in code.
How to install Jevc
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open doronp/jevc and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Jevc does
Compile agent policy prose into deterministic verdict programs: narrow evidence questions for the model, the verdict computed in code. Install: npm i -g jev-compiler
Alternatives in AI
- Graphjin — GraphJin - An Automagical Compiler to Connect AI to Your Databases 3034 44 1 3k ★
- Upstream Watch — Review pydantic-ai and pydantic-ai-harness releases published since the last watch entry, and record a verdict 361 ★
- Engage.Scorecard — Calibrate model verdict trust (Wilson-bounded miss-rate) to short-circuit re-validation 348 ★
README

jevc
**Turn the rules your agent keeps ignoring into gates you can test.**
Your `CLAUDE.md` says *NEVER commit unless the user explicitly asks*, and the agent commits anyway — because a markdown rule is a suggestion the model re-reads fresh every turn. `jevc` compiles a rule like that into a **Jev program**: a few narrow typed questions answered by TypeSafe's System One model (`jev-1.13.0` — non-generative, returns probabilities, never text), plus a reducer that computes the verdict in ordinary code. The rule stops being prose the model weighs and becomes an invariant you can read, diff and unit-test.
It is for anyone shipping a gate an LLM currently judges — a Claude Code `PreToolUse` hook, a tool-call guard, a model router, an output verifier — who needs that decision to be reproducible and reviewable.
Start here
npm install -g jev-compiler # the package is jev-compiler; the command it installs is jevc
jevc check # replays the recorded corpus, offline
jevc scan . # what this project could enforce
From source instead: `git clone https://github.com/doronp/jevc && cd jevc && npm install && npm run build`, then `npx jevc …` from the checkout. Node >= 22 to run the test suite.
$ npx jevc check
58 fixtures, 58 passing, 0 failing
No API key, no network — not for the tests, the examples, or anything in this README except `jevc check --live`.
As a Claude Code plugin: `/plugin marketplace add doronp/jevc`, then `/plugin install jevc@jevc`. It adds one skill, `/jevc:jevc`, which runs the scan, lift and compile below for you and asks before it installs a gate ([docs/wiring.md](docs/wiring.md#claude-code-skill-and-slash-command)).
Your rules are already in the repo
Point `jevc scan` at a project. It finds the instruction files agent harnesses actually read — `CLAUDE.md`, `AGENTS
Related Skills
Nethergate
An audited gate that lets AI agents (Claude Code / Codex) defend a Linux server through narrow MCP tools — no
System Ir
System Context Compiler — compile code, config, infra, runtime, and intent into an evidence-backed machine mod
Scc
System Context Compiler — compile code, config, infra, runtime, and intent into an evidence-backed machine mod
Grokkable Output
Agent skill that makes AI replies parseable in one pass — verdict first, prose over fake structure, claims tha
Word Compiler
Context compiler for long-form fiction that turns character, voice, and scene plans into guided LLM prose gene
Implement Structural
Implements the model extension behind a STRUCTURAL /analyze-policy verdict — baseline change for enacted law,
Related Agents
Prose Asserter
Judges a prose-test walk against the case's expected path and the code-computed world delta, returning a four-
Digest
Generic digest compiler. Input: path to project config file and number of days. Reads pre-computed per-commit
Light Compiler
Use this agent when you need to compile a Delphi project and verify that code changes build successfully. This