Author, evaluate, security-review, and autonomously improve Claude Agent Skills — in any project, with evidence, not vibes.
The core idea: a skill is only as good as the evidence behind it. skill-kit
brings the full evaluation stack from the
skill-testing dev-bench to any
project as an installable plugin — a 21-check static harness, a live behavioral
check, empirical trigger measurement, a blind value-add baseline, and an
autoresearch loop that iteratively tightens a SKILL.md and keeps only the
changes that score better.
| Layer | Tool | What it proves |
|---|---|---|
| 🧹 Static | check-skill |
The skill is well-formed — 21 checks: frontmatter, naming, trigger language, secret & security-smell scans |
| ⚡ Behavioral | behavioral-check |
The documented commands actually run against the real system today |
| 🎯 Empirical trigger | trigger-accuracy |
The skill really fires on its positive prompts and declines its negatives |
| ⚖️ Value-add | value-add-test |
The skill beats the cold model in a blind head-to-head |
| 🔄 Autoresearch | improving-skills |
Iterative modify → score → keep-or-revert loop that tightens an existing SKILL.md |
| ⚓ Goal anchoring | goal-new-skill · goal-improve-skill |
Skill work is driven by a measurable end state, not vibes |
The harness is self-tested (9 pathological fixtures + 34 pytest tests) and CI enforces zero FAILs and zero WARNs on the shipped skills, on every push and PR.
Inside Claude Code, on any machine:
/plugin marketplace add mjenkinsx9/mjenkins-toolbox
/plugin install skill-kit@mjenkins-toolbox
Update later with /plugin update skill-kit. Because a plugin's bin/ joins the
Bash PATH, every helper is callable by bare name in any repo.
skill-kit also installs on Copilot CLI, Codex, Cursor, and Gemini —
see Multi-harness support.
# Lint a skill directory (or a SKILL.md directly) — exit 0 iff zero FAILs:
check-skill .claude/skills/my-skill/skill-kit:improving-skills .claude/skills/my-skill/SKILL.md
Full guides live in docs/. Start with the overview, then dive into the
section you need.
| Doc | Description |
|---|---|
| Overview | What skill-kit is, the core idea, and the evaluation stack |
| Installation | Marketplace install, the bin/-on-PATH convention, requirements |
| Usage | The bin/ commands, slash commands, and the improving-skills scoring loop |
| Multi-harness support | Per-harness status, manifests, and install notes |
| Repository layout | Map of manifests, helpers, commands, skills, and tests |
| Testing the tests | Self-tests + pytest, and the relationship to skill-testing |
Full documentation map: docs/README.md
MIT © 2026 Mike Jenkins · Security policy