skill-lint

To check skill health and detect drift, evaluate freshness, usage, integrity, format, and dependencies to produce a CURRENT, STALE, BROKEN, or DELETE verdict for each skill across root and workspace repositories.

Deterministic skill linting works by scoring file metadata, content references, and presence checks using file stats, grep counts, and path existence checks, avoiding LLM judgment to generate reliable health verdicts.

Automate broken link detection in SKILL.md files by scanning root and workspace skill collections, resolving cross-skill references, and checking path existence to identify missing frontmatter and drift.

Yes, you can use skill health checks to identify stale or unused skills by evaluating usage and freshness dimensions with deterministic scoring thresholds, mapping the aggregated total to specific verdicts.

You need to run a skill health assessment when managing multiple skill repositories to detect drift, broken references, missing frontmatter, and format degradation before deploying updates.

Deterministic scoring for skill verification is limited to file metadata, grep counts, and path existence checks, meaning it cannot evaluate semantic correctness or behavioral accuracy of skill logic.