review-code_skill

Overview

This skill performs multi-dimensional code reviews and produces structured, actionable reports. It evaluates correctness, readability, performance, security, testing, and architecture to prioritize issues by severity. The goal is clear, fixable findings ranked from critical to informational.

How this skill works

The skill collects the target codebase and metadata, runs a fast scan to surface high-risk areas, then performs deeper per-dimension analyses. Findings are recorded into a structured state and assembled into a prioritized review report with file locations and remediation suggestions. The process is state-driven and iterates until the review reaches completion.

When to use it

• Before pull request merges for medium-to-large changes
• When assessing legacy code or third-party libraries
• During security or performance triage
• To validate test coverage and edge-case handling
• When reviewing architecture or design consistency

Best practices

• Provide spec documents and acceptance criteria up front so the review aligns with project standards
• Limit review scope to logical modules to keep deep reviews focused and actionable
• Prioritize fixes by severity: address critical and high issues before medium/low ones
• Include small, reproducible code samples or failing tests for correctness issues
• Use the structured report to create tracked issues with suggested fixes

Example use cases

• Run a pre-merge review on a feature branch to catch regressions and security risks
• Audit a legacy service to identify performance bottlenecks and missing tests
• Assess a pull request for naming, complexity, and maintainability concerns
• Validate that new code follows architectural patterns and dependency rules
• Generate a prioritized remediation plan after a security scan