jevbooks

← All projects

jev-skill-suggester

win4r/jev-skill-suggester

Bounded installed-skill recommendations with TypeSafe Jev.

Agent Decisions44%Runner-up: Classification & RoutingTiered thresholds

What it is

A Codex Skill and standalone Python CLI that recommends which installed skill to read first for a task, for agents with many overlapping skills. It does not execute or install skills.

How it uses Jev

Jev uses Choice to rank skill descriptions and a separate Noul to judge if the group is useful, keeping up to three candidates per group. After reviewing body snippets, Choice and per-candidate Noul decide the recommendation. Thresholds: fit ≥ 0.80, Choice confidence ≥ 0.65.

Primitives:choicenoul

Technique worth stealing

Two-stage bounded recommendation: coarse Choice ranking plus Noul usefulness gate, then snippet review before final Choice/Noul.

Try it

python3 -I -B scripts/suggest.py suggest --task '...' --mode jev --prompt-key --out results/report.json

View on GitHub

judged by Jevjev-1.13.0

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-centricunclear0.59
  • Shows a System One patternno0.30
  • Handles uncertaintyunclear0.45
  • Measuredno0.11
  • Runnableyes0.97
  • Worth recommendingunclear0.58
  • Model replicano0.04
  • Problem scopescore on a 0–2 scale1.05
  • About Jevyes0.97

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.