Where does the tool look for skills to audit?

It scans the global directory at ~/.claude/skills/ and, if run from a project root, the project-specific directory at .claude/skills/.

What is the difference between Quick Scan and Full Stocktake?

Quick Scan re-evaluates only the skills that have changed since the last run, typically taking 5–10 minutes. Full Stocktake performs a complete review of all skills and usually takes 20–30 minutes.

Does Skill Stocktake automatically delete my files?

No. All operations involving retiring, merging, or deleting skills require explicit user confirmation before any files are modified or removed.

Can I resume an audit if it is interrupted?

Yes, Skill Stocktake saves intermediate results to a results.json file. If an audit is interrupted, it can resume from the last evaluated skill.

How does the AI evaluate skill quality?

A general-purpose subagent checks each skill against criteria like actionability, scope fit, uniqueness compared to other skills, and the currency of technical references.

Skill Stocktake

Name: Skill Stocktake
Author: hieuck

byhieuck

0•

개발자 도구

Audits and optimizes your Claude skills library through automated quality checks and AI-powered evaluations.

Skill Stocktake is a comprehensive maintenance utility for Claude Code users designed to ensure their custom skills remain high-quality, relevant, and efficient. By executing the /skill-stocktake command, the tool performs a holistic audit of global and project-specific skill directories, identifying redundant content, outdated technical references, and overlapping instructions. It utilizes an intelligent subagent to evaluate each skill against a standardized quality checklist, providing actionable verdicts such as Keep, Improve, Update, or Retire. This allows developers to maintain a lean, high-performance workspace by systematically pruning or merging skills that no longer add unique value.

주요 기능

01Support for both global (~/.claude/skills/) and project-level skill paths

02Detailed verdict reporting with actionable improvement suggestions

030 GitHub stars

04Intermediate result caching to support resuming long-running audits

05Dual-mode auditing: Quick Scan for diffs and Full Stocktake for deep reviews

06AI-powered subagent evaluation using holistic quality checklists

사용 사례

01Identifying and merging overlapping or redundant skill definitions

02Verifying the freshness of technical references and API flags in skills

03Maintaining a clean and efficient library of Claude Code skills

주요 기능

01Support for both global (~/.claude/skills/) and project-level skill paths

02Detailed verdict reporting with actionable improvement suggestions

030 GitHub stars

04Intermediate result caching to support resuming long-running audits

05Dual-mode auditing: Quick Scan for diffs and Full Stocktake for deep reviews

06AI-powered subagent evaluation using holistic quality checklists

사용 사례

01Identifying and merging overlapping or redundant skill definitions

02Verifying the freshness of technical references and API flags in skills

03Maintaining a clean and efficient library of Claude Code skills