perf-recommendations

name	perf-recommendations
description	Generate optimization recommendations and validate performance improvements against budgets and targets
model	sonnet-4-6
token_budget	{"input":8000,"output":4000}

Performance Recommendations Skill

Identifies performance bottlenecks, recommends optimizations, and validates improvements.

Optimization Recommendations

CPU Optimization

• Profile >50ms tasks, break into chunks
• Use requestIdleCallback for non-critical work
• Tree-shake unused code, code-split large modules
• Lazy load routes, remove unused polyfills
• Optimize loop performance

Memory Optimization

• Remove event listeners on cleanup
• Clear timers/intervals, nullify references
• Profile with Heap Snapshots regularly
• Lazy load images/videos
• Implement cache eviction policies
• Use object pooling

Render Optimization

• Optimize React renders (useMemo, useCallback)
• Reduce re-renders, use React Profiler
• Implement virtualization for long lists
• Debounce scroll/resize handlers
• Use transform/opacity instead of position changes

Network Optimization

• Implement caching strategies (HTTP headers)
• Request batching (combine multiple requests)
• Response compression (gzip, brotli)
• Service workers for offline support
• Image optimization (WebP, responsive)

Startup Optimization

• Lazy load non-critical features
• Prefetch critical resources
• Minimize initial bundle
• Async load analytics/ads
• Preload fonts and critical images

Battery Optimization

• Reduce polling frequency
• Implement exponential backoff
• Cancel requests when backgrounded
• Use efficient networking (HTTP/2)
• Avoid continuous CPU usage

Validation Against Budgets

Checks:

• Cold Startup: < 2000ms
• Memory Active: < 400MB
• FPS: >= 60
• Bundle Size: <= 50MB

Reports improvement percentages and pass/fail for each metric.

Recommendation Report

Documents:

• Critical Issues (MUST FIX)
• High Priority (FIX SOON)
• Medium Priority (PLAN REMEDIATION)
• Low Priority (DOCUMENT FOR FUTURE)
• Total Effort & ROI

Optimization Validation

1. Baseline Profile (before changes)
2. Apply Optimization
3. New Profile (after changes)
4. Compare Results
5. Validate Budgets
6. Report Impact

Guardrails

• Fail: FPS < 60 mobile, < 30 browser
• Fail: Memory > 400MB, any leaks
• Fail: Bundle > 50MB
• Fail: Startup cold > 3s
• Warn: Startup approaching limits
• Warn: Cache approaching budgets
• Warn: API p95 > 500ms

Triggers

Use when:

• Performance profiling complete
• Bottlenecks identified
• Optimization strategy needed
• Changes validated for performance
• Release performance validation needed

---