What problem does it solve? As a collection of AI agent skills grows past a dozen interlinked entries, descriptions drift, cross-references break, and overlapping skills compete for the same triggers. This Skill runs a structured periodic audit across an entire skills directory so maintainers catch quality decay before it causes mis-triggering or dead-end instructions. ## Core Features & Use Cases - Four-category audit: Checks staleness (skills left behind as ecosystem conventions evolve), redundancy (unintentional trigger overlap without boundary statements), broken cross-references (pointers to missing files or skills), and description quality (too vague or too broad to trigger reliably). - Severity-ranked report: Produces a summary table plus per-skill findings with concrete evidence (file, line, or date), a clean list, and a prioritized recommendation list. - Strict scope control: Audits only the repository's own skills directory, explicitly excluding global agent harness installs, project memory, and unrelated folders. - Use Case: After adding a batch of new skills to your Claude Code ecosystem, run the audit to confirm no two skills claim the same trigger phrases and that every referenced file still exists. ## Quick Start Audit all skills in this repository's skills directory and report any staleness, redundancy, broken references, or description problems.
