What problem does it solve? Over time, project documentation drifts from the code it describes: rules go stale, resolved issues stay listed as open, and docs contradict each other or the repository. This Skill performs a read-only, out-of-cycle integrity audit that checks AI-context documentation against actual repository evidence and reports discrepancies with severity ratings. ## Core Features & Use Cases - Evidence-Based Verification: Classifies claims as descriptive or normative and resolves them against code, tests, scripts, schemas, and generated artifacts using a defined evidence hierarchy. - Structured Discrepancy Report: Produces a standardized audit report with severity levels (blocker, material, minor), code-backed findings, mechanical fixes, and decisions requiring user input. - Scoped Audit Breadth: Defaults to the narrowest audit that answers the request, expanding into source evidence only where a discrepancy requires verification. - Use Case: After several months of development, ask for a documentation integrity audit to find stale lifecycle rules, broken file references, missing frontmatter, and docs that contradict current code—then apply only the mechanical fixes you authorize. ## Quick Start Audit the AI-context documentation for integrity and report any discrepancies against current repository evidence.
