Contract
- Input: problem or task defined by the skill body.
- Output: Markdown artifact or structured result with completion criteria met.
- Side effects: none (design/review/documentation only unless explicitly stated).
- Dependencies: none (self-contained unless linked to other skills).
- Stop condition: all process steps completed; artifact saved; criteria checked.
- Risk: low.
- Boundary: produces reasoning or documentation artifacts; does not modify external systems unless explicitly instructed.
Skill Audit
Use this skill to verify the repository-local Skills platform.
Contract
- Input: the local Skills bundle, lockfile, symlink tree, and platform schemas.
- Output: a concise audit report with warnings and errors ranked by impact.
- Scope: verify bundle health; do not modify the skills during the audit.
- Rule: report parity, lock coverage, and contract drift separately.
- Rule: treat validator output as evidence, not as the audit itself.
- Rule: distinguish inventory drift, symlink drift, and contract drift so the maintenance action is obvious.
- Rule: if the bundle is healthy, say so explicitly instead of summarizing only the validator.
Steps
- Run the platform validator in
.agents/skills/platform/validate-skills.mjs. - Run the semantic auditor in
.agents/skills/platform/audit-semantics.mjswhen the audit includes templates, retired names, Markdown links, or contract drift. - Inspect the output for missing canonical skills, broken
.claude/links, and lockfile drift. - Inspect the inventory for stale aliases or mismatched metadata that the validator may not classify crisply.
- Summarize findings as a prioritized maintenance report.
Completion criteria
.agents/and.claude/parity is checked- the lockfile coverage gap is reported
- any missing manifest or contract issue is surfaced
- the report is prioritized and actionable
- the validator result is captured in the audit evidence
- semantic audit output is captured when run
- the report states whether the bundle is healthy, warning-only, or failing