Overview & result
github.com/droid-Q/jev-sk…, skill routing built on Jev.
From the original post
github.com/droid-Q/jev-sk… ，基于 jev 的 skills 路由
How Jev fits in the loop
- Jev evaluates the skills that are eligible for the current Codex prompt.
- The selected skill names, paths and relevance probabilities are passed to Codex.
- Selections, no-matches, empty catalogues and routing failures are recorded locally with project, session, model, threshold and latency.
- A web dashboard shows selection rates, daily trends, skill rankings and individual records.
How to reproduce
- Inspect the original showcase and source material at https://github.com/droid-Q/jev-skill-router
- Verify the bounded prompt and input candidate schema configured for Jev
- Benchmark decision latency and classification accuracy against baseline models
Why this build matters
It makes skill routing inspectable: every decision is logged locally and can be filtered by time, project, skill or outcome.
Limitations
- The README's screenshots use synthetic demo data, not measured Jev accuracy.
- Installing the plugin does not trust its hooks; you review and trust them in Codex.
- The dashboard start-up helper runs only in a macOS login session.