What problem does it solve? Single-pass code and design-system audits can miss findings or produce unverified claims, and there is no trustworthy way to confirm that a review actually covered everything. This Skill runs a complete no-sampling primary audit plus a genuinely independent second audit from a different provider, then reconciles the two with evidence so no finding is accepted on trust alone. ## Core Features & Use Cases - Dual isolated audits: Executes a complete primary audit and an independent peer audit in a separate read-only context against the identical frozen inventory and rubric. - Evidence-based reconciliation: Compares findings dimension by dimension, reproduces every claim against source, and preserves disagreements instead of resolving them by vote. - Fail-closed governance: Returns REVIEW-BLOCKED whenever coverage is incomplete, identity is missing, isolation is absent, or the worktree drifts, never fabricating a PASS. - Use Case: Before a design-system release, run this Skill to audit every component, rule, and test with two independent providers, then receive a report of verified findings, rejected claims, and residual risk. ## Quick Start Run a deep cross-provider audit of this repository with full coverage and reconcile the independent findings.
