What problem does it solve? Code changes often ship without consistent review standards, leaving correctness, security, and complexity issues undetected until production. This Skill applies a structured TRUST 5 review (Tested, Readable, Unified, Secured, Trackable) to every diff with deterministic verdicts. ## Core Features & Use Cases - TRUST 5 Review: Evaluates changed code against coverage, naming, consistency, OWASP security, and commit conventions, producing APPROVE or REQUEST_CHANGES verdicts. - Risk-Tiered Multi-Provider Review: Routes low/medium risk changes to a single provider and high/critical changes to multi-provider dissent review, with degraded fallback tracking. - Findings Split and Repair Loop: Separates authoritative Correctness/Security findings from advisory Complexity findings, and feeds REQUEST_CHANGES checklists back into the autonomous fix loop with a retry budget of 2. - Use Case: After implementing an OAuth2 feature, run the review on the diff to receive severity-ordered security findings, UI design-context drift checks, and a decision receipt before merging. ## Quick Start Ask the agent to run /auto review on the current diff or a specific path such as /auto review HEAD~3..HEAD --risk-tier high.
