What it is
A GitHub Action for pull requests written by coding agents. It reads the diff and posts a single comment saying what the PR really does, which hunks need a human, why, and what to verify.
How it uses Jev
Jev judges every hunk, answering yes/no questions with calibrated probabilities. A generative model is called only for hunks Jev flags. Gates stand on Jev alone; warnings without a second look are counted, not shown.
Primitives:choice
Technique worth stealing
Use a small calibrated-probability model to triage every hunk, and call a generative model only on the hunks it flags.
Try it
Add the workflow file and two secrets, then open a pull request. Or run npx github:davidrydberg/git-judge-jev --base main.
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-centricunclear0.51
- Shows a System One patternno0.23
- Handles uncertaintyno0.05
- Measuredno0.15
- Runnableyes0.92
- Worth recommendingno0.26
- Model replicano0.12
- Problem scopescore on a 0–2 scale1.55
- About Jevyes0.95
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.