agency-code-reviewer

Provide the code or diff and request a review. The feedback is organized by priority: blockers for security and correctness issues, suggestions for validation and performance, and nits for minor improvements, each with explanations and fix recommendations.

It checks five areas: correctness of logic, security vulnerabilities like injection and auth bypass, maintainability and readability, performance issues like N+1 queries, and test coverage for important code paths.

Yes, it flags common vulnerabilities including SQL injection, XSS, and authentication bypass with specific line references. Each finding includes an explanation of the attack vector and a suggested fix such as parameterized queries.

No, it complements human review by catching common issues and enforcing consistent feedback structure. Human reviewers remain essential for architectural decisions, business logic validation, and context-specific judgment.

It cannot execute code or verify runtime behavior, so it may miss environment-specific bugs. It focuses on static analysis of provided code and works best when given sufficient context about intent and surrounding systems.