What problem does it solve? Written specifications often contain contradictions, arithmetic errors, and gaps that only surface during a build. This Skill audits a document set with independent reviewer agents that never saw the author's reasoning, then verifies every claimed fix in the actual text before a build depends on the documents. ## Core Features & Use Cases - Layered adversarial audits: A cross-set consistency pass, a mandatory re-verify round after fixes, a targeted confirmation round, and per-document deep dives with document-specific lenses. - Provable findings and verdicts: Every finding cites a file and section with a prescribed fix, classified as BLOCKER, MAJOR, or MINOR, and rolled up into CLEAN, PASS-WITH-FIXES, or FIX-FIRST verdicts. - Audit trail recording: Writes dated audit logs to audits/ with reconciled counts, per-round verdicts, and honest caveats about what was not verified. - Use Case: Before building from a requirements pipeline (BRD, PRD, FSD, TDD), run the audit to catch cross-document contradictions, recompute formulas, and confirm the fix pass did not introduce new defects. ## Quick Start Ask the AI to audit the docs in your specification folder before starting the build.
