agentleFS
Sign inStart free
ArangoGutierrezArangoGutierrez/claude-toolkit/.claude/skills/skill-eval/SKILL.md

skill-eval

Skill4 starsChanged 3 months agoMITView on GitHub

Measure skill discoverability — run each skill's evals.json (positive + decoy prompts) through a real headless-claude activation probe and emit a pass/fail scorecard. Use to check routing/decoy behavior or catch mis-trigger regressions. Triggered by /skill-eval [skill ...]. Not for measuring output correctness or running skills for their side effects.

We split this skill into the rules it gives an agent

2 rules. 0 are common, 0 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 16 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Run scripts/skill-eval.sh with zero args (all skills that have an evals.json) or a list of skill names.

    line 24
  2. Very specific:

    All ERROR verdicts → claude invocation failing.

    line 42

“Common” is measured: kept by at least 6 in 10 of the 16 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

skill-eval

Purpose

Quantify whether a skill activates on prompts that should trigger it and stays silent on near-neighbor decoys, by probing real claude activation.

Instructions

Run scripts/skill-eval.sh with zero args (all skills that have an evals.json) or a list of skill names. Read the printed scorecard; a non-zero exit means at least one FAIL or ERROR.

Available Scripts

  • scripts/skill-eval.sh [skill ...] — orchestrator.
  • scripts/probe.sh / score.jq / report.sh / extract.jq — the units.

Prerequisites

claude CLI and jq on PATH. Each evaluated skill needs an evals.json (schema: {skill, cases:[{id, prompt, expect:"activate"|"silent"}]}).

Limitations

Discoverability only (no output-correctness). Activation is measured headlessly, a proxy for interactive routing. Rates use N attempts (default 5) — tune with SKILL_EVAL_N / SKILL_EVAL_PASS / SKILL_EVAL_DECOY.

Troubleshooting

  • All ERROR verdicts → claude invocation failing; run one probe manually.
  • Decoys always fire → tighten the skill's description negative-scope.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free