What problem does it solve? Self-reviewing code shares the author's mental model and blind spots, producing rubber-stamp approvals on changes a fresh reviewer would flag. This Skill forces genuine perspective shifts so critical bugs, maintainability issues, and security vulnerabilities surface before merge. ## Core Features & Use Cases - Three adversarial personas: The Saboteur hunts production-breaking edge cases, the New Hire flags maintainability and clarity problems, and the Security Auditor applies an OWASP-informed checklist. - Mandatory findings with severity promotion: Each persona must report at least one issue, and findings caught by 2+ personas are promoted one severity level (NOTE to WARNING to CRITICAL). - Structured verdicts: Every review ends with a BLOCK, CONCERNS, or CLEAN verdict plus deduplicated, severity-ranked findings. - Use Case: Before merging a self-authored PR touching authentication code, run an adversarial review on the diff to catch a missing null check and an unvalidated input that a same-model review would have approved. ## Quick Start Ask the AI to run an adversarial review on your staged changes or a specific diff such as the last three commits before merging your pull request.
