Code Review Patterns

What it does

This skill assists AI agents in performing code reviews for backend services and APIs, emphasizing quality after functionality is verified. It guides the agent through a two-stage review process: first ensuring the code meets requirements (spec compliance), then assessing its quality based on established heuristics and security checklists. The skill leverages Language Server Protocol (LSP) to enhance semantic understanding during the review process.

When to use it

• When reviewing backend services or APIs for potential bugs and quality issues.
• To ensure code meets specified requirements before evaluating code quality.
• During code reviews where a structured, prioritized approach is needed.
• When security vulnerabilities need to be identified in authentication, data handling, network communication, storage, or externally reachable code.
• For rapid reviews using the provided quick checklist.

Key capabilities

• Two-Stage Review Process: Enforces Spec Compliance review before Code Quality review.
• Reference Material Guidance: Provides access to reference files for review order, security checklists, and heuristics.
• Security Focus: Includes a dedicated security review checklist for critical areas.
• LSP Integration: Utilizes LSP for semantic code understanding (e.g., finding function callers).
• Quick Review Checklist: Offers a rapid assessment of 8 key code quality aspects.

Example prompts

• "Perform a code review on this pull request, focusing first on spec compliance."
• "Review this code for potential security vulnerabilities, using the provided checklist."
• "Can you quickly assess the code quality of this function based on the quick review checklist?"

Tips & gotchas

• Spec Compliance First: The skill emphasizes that code quality reviews should only occur after verifying the code meets requirements.
• False Positives: The agent is instructed to flag issues only when certain, as false positives are detrimental.
• Reference Materials: To ensure a thorough review, consult the provided reference files (review-order-and-checkpoints.md, security-review-checklist.md, code-review-heuristics.md) before forming findings.