performance

What problem does it solve? Teams often optimize code based on intuition rather than evidence, shipping changes that add complexity without measurable gains or that regress tail latency and memory. This Skill enforces a measure-first discipline for performance work and cache design. ## Core Features & Use Cases - Measurement Discipline: Requires naming a target metric (p99 latency, throughput, CPU, allocation) and capturing a baseline on production-shaped data before changing code. - Controlled Optimization: Enforces one performance change per commit, re-measured under the same workload, with checks that memory, error rate, and tail latency did not regress. - Cache Design Governance: Requires naming source of truth, invalidation trigger, stale tolerance, key contract, stampede protection, TTL/jitter, and metrics before keeping a cache. - Use Case: When a service endpoint is slow, use this Skill to profile the actual bottleneck, benchmark the fix against a baseline, and validate that p99 latency improved without regressing adjacent metrics. ## Quick Start Use the performance skill to profile this slow endpoint, establish a latency baseline, and design a caching strategy with explicit invalidation rules.