메뉴
Structured performance audit covering Core Web Vitals, bundle size, memory profiling, and runtime efficiency.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Rules and strategies for managing agent context window size, avoiding bloat, and preserving signal-to-noise ratio.
Standard patterns for error handling, retry logic, circuit breakers, and graceful degradation.
Detect and remove contradictions across agent policies before execution.
Multi-step tool workflows via code orchestration to reduce latency, context pollution, and token overhead.
Bind project-specific prompts to local schema and workflow artifacts while keeping the harness core generic and globally reusable.
Operational session protocol for task-scoped leases, reconciliation, checkpoints, inspection, queue promotion, and handoff across long-running work.
| name | performance-audit |
| description | Structured performance audit covering Core Web Vitals, bundle size, memory profiling, and runtime efficiency. |
Provide a structured performance audit procedure for web applications, covering load performance, runtime efficiency, and resource usage.
Run Lighthouse in both mobile and desktop modes:
chrome-devtools → lighthouse_audit (device: "mobile", mode: "navigation")
chrome-devtools → lighthouse_audit (device: "desktop", mode: "navigation")
| Metric | Target | Tools |
|---|---|---|
| LCP (Largest Contentful Paint) | < 2.5s | Lighthouse, Performance trace |
| INP (Interaction to Next Paint) | < 200ms | Performance trace |
| CLS (Cumulative Layout Shift) | < 0.1 | Lighthouse, Performance trace |
| FCP (First Contentful Paint) | < 1.8s | Lighthouse |
| TTFB (Time to First Byte) | < 800ms | Network panel |
| Check | Target | How |
|---|---|---|
| Total JS bundle (gzipped) | < 200 KB (initial load) | npm run build + analyzer |
| Largest dependency | Flag > 50 KB gzipped | Bundle analyzer |
| Tree-shaking | No unused exports in bundle | Build output analysis |
| Code splitting | Route-level chunks, not monolith | Build output analysis |
| Duplicate dependencies | Zero duplicates | npm ls --all or analyzer |
Use Chrome DevTools Performance trace:
chrome-devtools → performance_start_trace (reload: true, autoStop: true)
| Check | Target |
|---|---|
| Long tasks (> 50ms) | Identify and split or defer |
| Layout thrashing | No forced synchronous layouts |
| Memory leaks | Heap snapshot comparison (before/after interaction cycle) |
| Animation jank | 60fps for all animations |
| Event listener count | No unbounded growth |
chrome-devtools → list_network_requests
| Check | Target |
|---|---|
| Total requests on load | < 50 |
| Uncompressed assets | All text assets gzipped/brotli |
| Missing cache headers | Static assets have Cache-Control |
| Render-blocking resources | Minimize or defer |
| Image optimization | WebP/AVIF, responsive srcset, lazy loading |
chrome-devtools → take_memory_snapshot (filePath: "baseline.heapsnapshot")
# Perform user interaction cycle
chrome-devtools → take_memory_snapshot (filePath: "after-interaction.heapsnapshot")
Compare snapshots for:
systematic-debugging — for diagnosing specific performance issuestesting-policy — include performance benchmarks in CIcompletion-gate — performance regressions block the gate