What it is
Hunch lets a person or coding agent ask plain-English questions across a repository and get scored source locations to investigate. It also runs recurring review rules on diffs, locally or on pull requests.
How it uses Jev
Hunch sends each in-scope chunk to Jev and returns scored source locations. For review, Jev checks code against plain-English or typed questions and picks one label per finding. Results rank candidates and flag configured concerns; they do not prove a bug.
Primitives:choicescore
Technique worth stealing
Phrase behavior questions in plain English so Jev can match implementations regardless of the names the code uses.
Try it
npx skills add Kelbie/hunch, then npx -y --min-release-age=0 @kelbie/hunch find "..."
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-centricyes0.83
- Shows a System One patternno0.34
- Handles uncertaintyno0.05
- Measuredno0.04
- Runnableyes0.97
- Worth recommendingunclear0.56
- Model replicano0.04
- Problem scopescore on a 0–2 scale1.93
- About Jevyes0.95
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 21 Sept 2026.