What it is
jev-pref is a semantic linter for coding agents. It turns project preferences from AGENTS.md into jev-pref.json rules, checks them against diffs, staged files, or pull requests, and returns findings to the agent.
How it uses Jev
Jev classifies evidence in each diff hunk, staged file set, or pull request against project-defined semantic checks. jev-pref maps Jev's probabilities to approve, advisory, or fix_now outcomes, feeds findings back to the coding agent, and exits nonzero on blocking findings.
Primitives:choicescore
Technique worth stealing
Shape broad preferences into concrete, evidence-grounded checks with one judgment each, and leave deterministic rules to existing tooling.
Try it
Run npx jev-pref setup, then npx jev-pref review --hunks.
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-centricyes0.72
- Shows a System One patternunclear0.44
- Handles uncertaintyno0.05
- Measuredno0.04
- Runnableyes0.98
- Worth recommendingunclear0.52
- Model replicano0.04
- Problem scopescore on a 0–2 scale1.37
- About Jevyes0.97
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.