What problem does it solve? Performance work often becomes guesswork: code gets optimized without evidence, regressions ship silently, and no one can prove users are better off. This Skill enforces a measure-first loop so every optimization has a baseline, a named bottleneck, a verified fix, and a guard against regression. ## Core Features & Use Cases - Structured Investigation Loop: Follows a five-step cycle of measure, identify, fix, verify, and guard, with a symptom-to-measurement table covering slow page loads, laggy interactions, slow APIs, memory growth, and intermittent latency. - Bottleneck Classification and Fixes: Categorizes issues into database, network/API, frontend loading, frontend rendering, caching, CPU/memory, concurrency, and tooling, with concrete anti-pattern fixes such as resolving N+1 queries, bounding list endpoints, and reducing bundle size. - Core Web Vitals Guidance: Provides LCP, INP, and CLS thresholds and routes web-facing tasks to a web performance auditor when browser rendering review is needed. - Use Case: A dashboard endpoint feels slow in production. Use this Skill to baseline response time, identify an N+1 query pattern, apply a batched query fix, verify with before/after query counts, and add a query-count budget as a regression guard. ## Quick Start Ask the AI to investigate why a specific page or API endpoint is slow and produce a measured before-and-after performance report.
