Skill Stocktake
shimo4228/skill-stocktake
Audita e cuida da saúde do seu catálogo de skills com inteligência
Ver no GitHubSkill para Claude que audita todas as suas skills e comandos instalados, avaliando qualidade, sobreposição e relevância. Produz verditos de Manter, Melhorar, Atualizar, Aposentar ou Mesclar baseado em checklist e análise holística com IA.
Instalação
npx convez-ai shimo4228/skill-stocktake1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Identificar skills obsoletas ou com baixo uso
- Detectar redundância entre skills existentes
- Priorizar melhorias e atualizações tecnológicas
- Manter inventário organizado de competências
Veredito

Ferramenta especializada e bem arquitetada para governança de skills em ambientes Claude avançados. Essencial para quem mantém catálogos grandes ou quer implementar ciclos de curação contínua de agentes.
README
A tradução para o português deste README ainda está na fila.
Language: English | 日本語
skill-stocktake
An Agent Skill that audits all your Claude skills and commands for quality. Uses a checklist + AI holistic judgment to produce Keep / Improve / Update / Retire / Merge verdicts.
Install
Claude Code
# Copy skill + scripts into your global skills directory
cp -r skills/skill-stocktake ~/.claude/skills/skill-stocktake
SkillsMP
/skills add shimo4228/skill-stocktake
Modes
| Mode | Trigger | Duration |
|---|---|---|
| Quick Scan | results.json exists (default) | 5-10 min |
| Full Stocktake | results.json absent, or /skill-stocktake full | 20-30 min |
How It Works
Quick Scan
Re-evaluates only skills that changed since the last run. Uses scripts/quick-diff.sh to detect mtime changes, then carries forward unchanged results.
Full Stocktake
- Phase 1 — Inventory:
scripts/scan.shenumerates all skill files, extracts frontmatter, and collects usage stats - Phase 2 — Quality Evaluation: AI subagent reads each skill and applies the checklist (overlap, freshness, usage frequency)
- Phase 3 — Summary Table: Verdicts with actionable reasons
- Phase 4 — Consolidation: Retire/Merge/Improve actions with user confirmation
Verdict Criteria
| Verdict | Meaning |
|---|---|
| Keep | Useful and current |
| Improve | Worth keeping, but specific improvements needed |
| Update | Referenced technology is outdated |
| Retire | Low quality, stale, or cost-asymmetric |
| Merge into [X] | Substantial overlap with another skill |
Scripts
| Script | Purpose |
|---|---|
scripts/scan.sh | Enumerate skill files with frontmatter and mtime |
scripts/quick-diff.sh | Detect changed/new skills since last evaluation |
scripts/save-results.sh | Merge evaluation results into results.json |
Requirements
jq(JSON processing)bash4+- Claude Code with Agent tool support (for AI evaluation)
About this skill
This skill implements the Curate phase of the Agent Knowledge Cycle (AKC) — a Zenodo-citable six-phase bidirectional growth loop (DOI 10.5281/zenodo.19200726) for sustaining intent alignment between an AI agent and its operator over time. AKC is one of three research lines by @shimo4228, alongside Contemplative Agent (DOI 10.5281/zenodo.19212118) — autonomous agents grounded in four contemplative axioms — and Agent Attribution Practice (AAP) (DOI 10.5281/zenodo.19652013) — harness-neutral ADRs on accountability distribution.
License
MIT