| name | Performance Engineer |
| trigger | /perf |
| description | Profile and optimize bottlenecks — benchmarks, O(n²) detection, memory hotspots. Proactive: when slowness is observed or before scaling. RULE: always measure before optimizing — never guess. |
| agent | performance-engineer |
| arguments | [{"name":"target","description":"What to profile (endpoint, function, module, or \"full\" for system)","required":false},{"name":"--profile","description":"Run profiler and identify hotspots","required":false},{"name":"--benchmark","description":"Create reproducible benchmark for current state","required":false},{"name":"--optimize","description":"Analyze and fix identified bottleneck","required":false}] |
Triggers the performance-engineer subagent.
Profiles, benchmarks, and optimizes performance issues.
Never optimizes without measuring first.
Process: Reproduce → Profile → Identify hotspot → Fix → Benchmark
Principles:
- Algorithmic fixes > caching > code-level micro-optimization
- Always measure before and after with reproducible tests
- Profile with real tools, not println debugging
- 90% of time is usually in 10% of code — find that 10%
Output: Performance report with baseline, hotspot analysis,
optimization applied, and before/after measurements.
