cwjechw98-lang

Jev Gates — Development skill for Claude Code

Development community

Three gates for any coding agent: a deterministic approval gate before irreversible actions, a completion gate that demands evidence instead of «done», and a rubric regression gate.

How to install Jev Gates

This entry records only its repository, not the path inside it, so there is no exact command to give. Open cwjechw98-lang/jev-gates and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Jev Gates does

Three gates for any coding agent: a deterministic approval gate before irreversible actions, a completion gate that demands evidence instead of «done», and a rubric regression gate. Powered by TypeSafe Jev (System One). Works with Claude Code, Codex, DeepSeek Harness, Cursor, opencode, Hermes, plain shell and CI.

Alternatives in Development

  • Lazycodex — The one and only agent harness for complex codebases 3.3k ★
  • Fast Jev Compaction — Claude Code plugin that replaces the compaction summary with Jev decisions: every tool call and result is scor 425 ★
  • Run Daily — Trigger the daily digest now (background via OS scheduler) and monitor until completion 355 ★

README

Jev Gates — three glowing gates in cyan, violet and amber on a dark grid

Stop saying "done" without proof.

Three small gates between an AI coding agent and its consequences. **No SDK. No dependencies. No server.** Plain Node 18+ scripts that any harness can run.

[![CI](https://github.com/cwjechw98-lang/jev-gates/actions/workflows/ci.yml/badge.svg)](https://github.com/cwjechw98-lang/jev-gates/actions/workflows/ci.yml) [![License: MIT](https://img.shields.io/badge/license-MIT-3fb950?style=flat-square)](LICENSE) [![Node 18+](https://img.shields.io/badge/node-%E2%89%A5%2018-339933?style=flat-square&logo=node.js&logoColor=white)](package.json) [![Dependencies: 0](https://img.shields.io/badge/dependencies-0-22d3ee?style=flat-square)](package.json) [![Cost per check](https://img.shields.io/badge/cost%20per%20check-%240.0001-fbbf24?style=flat-square)](#-cost) [![Model calls](https://img.shields.io/badge/model%20calls-2%20of%203%20gates-a78bfa?style=flat-square)](#the-three-gates)

[Quick start](#-quick-start) · [The three gates](#the-three-gates) · [Any harness](#-works-with-any-harness) · [Design rules](docs/DESIGN.md) · [Русский](README.ru.md)


A coding agent reports success. It has three ways to be wrong, and each one needs a different kind of check.

🔒 It acts when it should have asked

Approval prompts get switched off — for speed, for unattended runs, or because the harness never had them. Then nothing stands between the agent and a force-push.

✅ It says "done" without proof

Not usually lying: it read its own summary of what it did, and the summary was optimistic. A stricter prompt does not fix this.

🧪 It trusts a broken rubric

A new yes/no question looks reasonable, gets wired into a pipeline, and quietly answers "yes"