What problem does it solve? Engineers often optimize code based on intuition, which rarely matches real bottlenecks and adds complexity without speed gains. This Skill enforces a measure-first discipline so every optimization targets a verified hotspot with a defined goal. ## Core Features & Use Cases - Goal Definition First: Requires a concrete target (e.g., p99 latency under 200ms) before any work begins, preventing endless optimization. - Bottleneck Classification: Categorizes profiler findings into algorithmic complexity, N+1 queries, repeated computation, blocking, and excessive rendering, prioritizing order-of-magnitude wins over micro-optimizations. - One-Change-at-a-Time Verification: Mandates re-measuring after each single change and running full tests to catch correctness regressions from caching or concurrency changes. - Use Case: When an API endpoint is slow, use this Skill to profile it under realistic load, identify an N+1 query pattern, batch the queries, and verify the improvement against the recorded baseline. ## Quick Start Use the perf-optimize skill to diagnose why this endpoint is slow and propose a measured optimization plan.
