Agent Skills
› cosmicstack-labs/mercury-agent-skills
› mobile-performance
mobile-performance
GitHub用于分析和优化移动应用性能，涵盖启动速度、内存管理、电池效率及网络优化。提供iOS与Android的关键指标目标、具体优化工具及最佳实践建议。
Trigger Scenarios
需要诊断App启动慢的问题
分析应用内存泄漏或占用过高
优化网络请求延迟和流量
提升App帧率或降低包体积
Install
npx skills add cosmicstack-labs/mercury-agent-skills --skill mobile-performance -g -y
SKILL.md
Frontmatter
{
"name": "mobile-performance",
"metadata": {
"tags": [
"mobile",
"performance",
"profiling",
"optimization",
"ios",
"android"
],
"author": "cosmicstack-labs",
"version": "1.0.0",
"category": "mobile"
},
"description": "App startup, memory management, battery efficiency, network optimization, and profiling tools"
}
Mobile Performance
Profile, diagnose, and optimize mobile app performance.
Key Metrics
| Metric | Target | Tool |
|---|---|---|
| Cold Start | <1.5s | Xcode Organizer, Play Console |
| Warm Start | <0.5s | Systrace, os_signpost |
| Frames | 60fps (16ms) | Profile GPU, Perfetto |
| APK/IPA Size | <50MB | size-report, analyzer |
| Memory | <150MB | Xcode Memory Debugger, Memory Profiler |
| Network Latency | <200ms p95 | Proxyman, Charles |
Startup Optimization
iOS
- Reduce dynamic framework loading (prefer static libraries)
- Lazy-load non-critical services
- Profile with
Instruments - App Launch - Use
dispatch_oncefor singletons
Android
- Apply
App Startuplibrary for content providers - Profile with
Android Studio - Startup Profiler - Use baseline profiles (Android 12+)
- Defer heavy init to background threads
Memory Management
- Watch for retain cycles (weak/unowned references)
- Image caching: use NSCache/LruCache with size limits
- Monitor for OOMs in production with crash reporting
- Release view controllers when off-screen
Network Optimization
- HTTP/2 multiplexing for concurrent requests
- Response caching with ETags
- Prefetch data predictively
- Compress with Brotli or gzip
- Use Protocol Buffers over JSON for large payloads
Version History
- 38e2523 Current 2026-07-05 19:40
