Audits and manages Claude skills for quality, technical accuracy, and redundancy using AI-driven evaluation.
Skill Stocktake is an essential maintenance utility for Claude Code environments designed to keep your skill library lean and effective. It automates the audit of custom skills and slash commands across global and project-specific directories, utilizing AI subagents to perform holistic evaluations. By checking for technical freshness, content overlap, and actionable utility, it provides clear verdicts on whether to keep, improve, update, retire, or merge skills, ensuring your AI assistant operates with only the most relevant and high-quality instructions.
Key Features
01Support for both global (~/.claude/skills/) and project-level skill directories.
020 GitHub stars
03Interactive consolidation phase for merging, retiring, or updating skills with user approval.
04Sequential subagent batch processing for manageable and accurate evaluation chunks.
05AI-powered quality checklist evaluating actionability, scope, and technical currency.
06Dual-mode scanning: Quick Scan for recent changes and Full Stocktake for deep audits.
Use Cases
01Cleaning up a cluttered Claude skill library to reduce context noise and token costs.
02Verifying that technical skills remain accurate after library or CLI tool updates.
03Identifying and merging redundant skills created across different development projects.
