Performance Optimization
Measure before optimizing — performance work without measurement is guessing that adds complexity without improving what matters. Its workflow is measure → identify → fix → verify → guard: establish a baseline with both synthetic (Lighthouse, DevTools) and real-user (RUM, CrUX) data, find the actual bottleneck, fix that one thing, then measure again against Core Web Vitals targets (LCP ≤ 2.5s, INP ≤ 200ms, CLS ≤ 0.1).
A symptom-to-measurement decision tree routes each complaint to the right profiler, and the skill catalogs concrete anti-patterns with fixes: N+1 queries, unbounded fetching, unoptimized images, unnecessary React re-renders, oversized bundles, missing caching. Enforceable performance budgets close the loop in CI. It applies pattern-measure-first, emits an artifact-perf-audit, and can delegate to the addy-web-performance-auditor persona.
See Also
- addy-web-performance-auditor — the persona this delegates auditing to.
- pattern-measure-first — the discipline it embodies.
- artifact-perf-audit — the audit it produces.
- stage-review — the canonical stage this implements.