ai-agents-workspace-starter: Skill for Claude Code

Performance Audit

Overview

Structured checklist for identifying and fixing performance bottlenecks in frontend and backend code.

When to Use

• Investigating slow page loads or API responses
• Reviewing new code for performance anti-patterns
• Auditing database queries for N+1 or missing indexes
• Reducing frontend bundle size or render overhead

When NOT to use:

• Premature optimization without measured bottleneck
• Micro-optimizations with no measurable impact
• Architectural redesigns (that's an architect task)

Checklist

Frontend Performance

• No unnecessary re-renders (missing React.memo, useMemo, useCallback)
• Expensive computations memoized
• Large routes use code splitting (lazy() / dynamic import())
• Images below fold use lazy loading (loading="lazy")
• Long lists use virtualization (react-window, tanstack-virtual)
• Frequent events debounced/throttled (search input, scroll, resize)
• No unused CSS or JS loaded on page
• Critical assets preloaded
• No synchronous operations blocking main thread

Backend Performance

• No N+1 query patterns (use eager loading / selectinload)
• Frequently queried columns have database indexes
• Expensive operations use caching (Redis, in-memory)
• I/O-bound operations are async
• Large datasets use pagination (never unbounded SELECT *)
• No O(n^2) algorithms where O(n) is possible
• Connection pooling configured for DB and external services
• Repeated API calls batched where possible
• Large payloads use compression (gzip/brotli)

General

• No redundant calculations inside loops
• Appropriate data structures for the use case (set vs list for lookups)
• Early returns to skip unnecessary processing
• No unnecessary deep cloning or copying
• Lazy initialization for expensive-to-create objects

Process

1. Measure - identify actual bottleneck (don't guess)
2. Profile - run checklist against the hot path
3. Optimize - implement fix with minimal code change
4. Verify - measure again to confirm improvement
5. Document - add comment explaining why the optimization exists