What problem does it solve? Code reviews often rubber-stamp changes without verifying that the implementation matches the spec or probing for real defects. This Skill enforces a two-stage review — spec compliance and code quality — and arms the reviewer with a checklist of failure modes this repository has actually been bitten by, such as checks that cannot fail, tenancy holes, and mock-only coverage across service boundaries. ## Core Features & Use Cases - Two-stage review gate: Verifies the code implements what was designed (no gaps, no scope creep), then audits patterns, security, accessibility, performance, and maintainability. - Repo-specific defect checklist: Targets historically real bugs — unfalsifiable checks, shared rows any authenticated user can mutate, direct provider SDK imports, per-user choices hidden behind global env flags, stored-but-never-read settings, and non-English text in persisted artifacts. - Structured findings and deferral discipline: Issues must be fixed and re-verified; deferrals require a tracked row plus a mechanism (asserted trigger, KNOWN_RED row, or named test), and results are emitted via the shared aif-gate-result contract. - Use Case: After implementing a multi-tenant feature, run this review to catch whether a new lint rule can actually fail, whether any shared row leaks across users, and whether cross-service behavior is covered by a live smoke test rather than mocks. ## Quick Start Review the current change set for spec compliance and code quality, then emit the findings as an aif-gate-result report.
