jevbooks

← All projects

jev-engineering

eugeniughelbur/jev-engineering · Homepage

Tool-call safety gate for AI coding agents, backed by Jev.

What it is

jev-gate is a safety gate for AI coding agents like Claude Code, Codex and Cursor. It answers allow, ask or deny for every tool call, and ships a 300-call injection test showing what the gate catches and what walks past it.

How it uses Jev

After deterministic hard rules and a read-only allowlist, Jev makes one request with two typed questions in parallel to judge a shell command. Its calibrated probabilities feed thresholds that produce allow, ask or deny. Any error or timeout falls back to the harness's normal permission prompt.

Primitives:choicescore

Technique worth stealing

Order the layers deliberately: denials before the allowlist, hard rules before the model, because the model is least reliable where a hard rule is easiest.

Try it

claude plugin marketplace add eugeniughelbur/jev-engineering; claude plugin install jev-engineering@jev-engineering

View on GitHub

judged by Jevjev-1.13.0

Evidence

Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.

  • Jev-centricyes0.77
  • Shows a System One patternyes0.86
  • Handles uncertaintyno0.13
  • Measuredyes0.70
  • Runnableyes0.91
  • Worth recommendingunclear0.49
  • Model replicano0.05
  • Problem scopescore on a 0–2 scale1.95
  • About Jevyes0.98

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 21 Sept 2026.