perf

name	perf
description	Profile, benchmark, and optimize application performance.

/perf

Profile, benchmark, and optimize application performance.

Usage

/perf [target] [--profile] [--benchmark] [--lighthouse]

Arguments

• target: File, endpoint, component, or area to analyze
• --profile: Run profiler and identify bottlenecks
• --benchmark: Run benchmarks and compare
• --lighthouse: Run Lighthouse audit (web only)

Instructions

When this skill is invoked:

Agent Behavior

Autonomy:

• Identify bottlenecks and implement optimizations
• Measure before and after to prove improvement
• Follow platform-specific best practices

Safety:

• Never optimize without measuring first
• Keep optimizations readable — don't sacrifice clarity
• Run tests after every optimization to prevent regressions

Performance Process

Phase 1: Measure Baseline

Establish current performance metrics before changing anything.

Web (Next.js / React):

# Lighthouse CI
npx lighthouse http://localhost:3000 --output json --output-path ./perf-baseline.json

# Bundle analysis
npx @next/bundle-analyzer
# or
npx webpack-bundle-analyzer stats.json

Python (FastAPI):

# Endpoint profiling
uv run python -m cProfile -o profile.out src/{project}/main.py
uv run py-spy record -o profile.svg -- python src/{project}/main.py

# Load testing
uv run locust -f tests/load/locustfile.py

iOS:

Instruments → Time Profiler, Allocations, Network
Xcode → Debug Navigator → CPU / Memory / Network gauges

Android:

Android Studio → Profiler → CPU / Memory / Network
./gradlew benchmark

Phase 2: Identify Bottlenecks

Analyze profiling results and categorize:

Category	Symptoms	Common Causes
Slow queries	High DB time, N+1	Missing indexes, unoptimized joins
Memory leaks	Growing memory, OOM	Unclosed connections, retained references
Bundle size	Slow page load	Large dependencies, no tree-shaking
Render perf	Janky UI, low FPS	Unnecessary re-renders, large lists
Network	Slow API calls	No caching, large payloads, no compression

Phase 3: Optimize

Apply targeted fixes based on findings:

Database:

• Add missing indexes
• Fix N+1 queries (use eager loading / joins)
• Add query result caching
• Optimize query structure

API:

• Add response caching headers
• Implement pagination
• Compress responses (gzip/brotli)
• Use connection pooling

Frontend:

• Lazy load components and routes
• Optimize images (next/image, WebP, sizing)
• Reduce bundle size (dynamic imports, tree-shaking)
• Memoize expensive computations

Mobile:

• Use FlatList / LazyColumn / LazyVStack for lists
• Optimize image loading and caching
• Reduce main thread work
• Minimize bridge calls (React Native)

Phase 4: Verify

Measure again and compare:

## Performance Comparison

| Metric | Before | After | Change |
|--------|--------|-------|--------|
| API Response (p95) | 450ms | 120ms | -73% |
| Bundle Size | 1.2MB | 680KB | -43% |
| Lighthouse Score | 62 | 94 | +52% |
| Memory Usage | 256MB | 180MB | -30% |
| DB Query Count | 47 | 12 | -74% |

Run full test suite to confirm no regressions:

{test_all_command}

Performance Report Format

## Performance Report

**Target:** {what was analyzed}
**Date:** {date}

### Baseline Metrics
{metrics before optimization}

### Bottlenecks Found
1. {bottleneck}: {impact}
2. {bottleneck}: {impact}

### Optimizations Applied
1. {change}: {expected improvement}
2. {change}: {expected improvement}

### Results
{metrics after optimization with comparison}

### Recommendations
- {further optimizations not yet applied}

Example Output

$ /perf /api/users --profile

Profiling /api/users endpoint...

Baseline (10 requests, p95):
  Response time: 450ms
  DB queries: 47
  Memory: 45MB

Bottlenecks found:
  1. N+1 query on user.posts (38 extra queries)
  2. No index on users.email (full table scan)
  3. Serializing unused fields (posts.body)

Applying fixes...
  Added eager loading for user.posts
  Created index on users.email
  Added field selection to query

Results:
  Response time: 120ms (-73%)
  DB queries: 3 (-94%)
  Memory: 38MB (-16%)

All tests passing (142/142).