Skill Stocktake is a comprehensive auditing framework designed to maintain a high-quality library of custom Claude Code skills and commands. It automates the process of identifying redundant, outdated, or low-quality instructions through sequential batch evaluation by an AI subagent. By checking for content overlap, technical freshness, and actionability, it ensures your agent's capabilities remain sharp, efficient, and cost-effective, offering clear verdicts on whether to keep, improve, update, retire, or merge specific skills.
Key Features
01Comprehensive results caching and resume support for large skill inventories
020 GitHub stars
03Full Stocktake mode for deep-dive library-wide quality assessments (20-30 min)
04Quick Scan mode for rapid auditing of recently modified skills (5-10 min)
05Sequential subagent evaluation using Claude 3 Opus for high-reasoning quality checks
06Automated identification of content overlap across skills, MEMORY.md, and CLAUDE.md
