AI Code Reviewer

AI Code Reviewer is an evidence-first, read-only code review Skill designed to help developers discover meaningful problems in code, pull requests, and AI-generated implementations. It analyzes correctness, security, performance, architecture, testing, maintainability, specification compliance, dependency risks, and regression risk. The Skill uses a structured TRACE → VERIFY → RANK workflow to investigate potential issues, evaluate available evidence, suppress weak or speculative findings, deduplicate related problems, and prioritize actionable risks. Each substantive finding includes severity, confidence, evidence, location, impact, root cause when established, recommendation, and suggested testing where useful. The Skill is designed with a least-privilege approach: it does not inherently require network access, secrets, environment variables, external services, or source-code modification. Repository content is treated as untrusted data to resist prompt injection, and the Skill never claims that tests, scanners, compilers, or other tools were executed unless they actually were and were explicitly authorized.