omp-jevify

What omp-jevify does

omp-jevify packages a jevify skill for Oh My Pi that generalises rubric-first bulk judging beyond test pruning. You write questions first, sample a few items, freeze the rubric, then let Jev’s judge_batch score every item in one pass — optionally calibrating against a slower model and acting only on flagged rows. Concrete jobs include checking whether commit hunks match the message, finding call sites left behind by a refactor, triaging compiler diagnostics, and turning a frozen rubric into an in-editor check while code is written. Install is an omp plugin pin; the skill can auto-trigger from goal language, the jevify keyword, or /skill:jevify. Helper Python loads into the eval kernel for extraction (diff, rg, ast-grep) with dry-run-safe span deletion by default. Eval harnesses measure auto-trigger rates on held-out prompts. Use omp-jevify when you have roughly twenty or more similar items and one repeated question — not for one-off single-file reviews where a slower generative pass is enough.

Where it fits in the Jev stack

• PR hunk relevance checks against the commit message
• Bulk call-site audits after renames
• Log / diagnostic triage with a frozen rubric

Browse more in Agent tooling, or return to the Jev AI tools directory.