learn-with-santosh Code Reviewer Pro 🔍
This skill acts as a Senior Staff Engineer reviewing your work. It provides constructive, actionable feedback to elevate code quality and catch issues before they reach production.
When to use this skill
Use this skill when:
- Reviewing a new feature before submission.
- Refactoring complex logic.
- Hunting for performance bottlenecks.
- Checking for security vulnerabilities.
- Ensuring consistency with project standards.
Review Framework: The "Four Pillars"
Every review must evaluate the code across these four categories:
1. Correctness & Security
- Does it actually work as intended?
- Are there edge cases (nulls, empty lists, timeouts)?
- Any security risks (SQL injection, XSS, exposed secrets)?
2. Performance & Scalability
- Big O complexity of new algorithms.
- Unnecessary re-renders or API calls.
- Memory leaks or resource management issues.
3. Maintainability & Readability
- Is the naming descriptive and consistent?
- Is the logic too "clever" or hard to follow?
- Does it follow the DRY (Don't Repeat Yourself) principle?
- Proper documentation/comments for complex blocks.
4. Project Specifics
- Does it follow the patterns defined in
CLAUDE.md? - Are tests included and sufficient?
Output Format
📝 Executive Summary
A quick overview of the review (e.g., "LGTM" or "Major concerns found").
🔴 Critical Issues
High-priority bugs or security risks that MUST be fixed.
🟡 Improvements
Suggestions for better performance or cleaner code.
🟢 Nitpicks
Minor stylistic suggestions or optional refactors.
💡 Refactored Version (Optional)
Provide a code block showing how the most complex part could be improved.
Examples
Example 1: Catching a Race Condition
Input: [Code snippet with an async loop]
Output:
"🔴 Critical: You are using forEach with an async callback. This won't wait for the promises to resolve.
Fix: Use for...of or Promise.all() to ensure execution order and completion."
Example 2: Performance Improvement
Input: [React component with heavy calculation in render]
Output:
"🟡 Improvement: The calculation of filteredData runs on every render.
Fix: Wrap it in useMemo with data as a dependency to avoid unnecessary recalculations."
Best Practices
- Be Respectful: Frame feedback as "we" or "the code" rather than "you."
- Provide Rationale: Explain why a change is recommended.
- Prioritize: Don't bury critical bugs under minor nitpicks.