Performance Optimization guides an agent through frontend, backend and database performance work with a five-step loop: measure, identify, fix, verify and guard. It sets Core Web Vitals targets and gives a symptom tree for deciding what to measure first.
The skill covers N+1 queries, pagination, query plans and indexes, connection pools, caching design, image optimization, re-renders and bundle size. A change that stays within noise is reverted, and every attempt is logged in a ledger.
What You Get
- A measure, identify, fix, verify and guard workflow
- Core Web Vitals thresholds and example performance budgets
- Anti-pattern entries with worked fixes in a references file
- A red flag list and a verification checklist
Common Use Cases
- Fixing slow API responses
- Improving Core Web Vitals
- Reviewing a suspected performance regression
- Setting performance budgets in CI
What Makes It Unique
The skill treats a neutral result as a revert, requires the same measurement method for baseline and result and asks for a ledger of tried ideas so dead ends are not repeated. Its backend entries cover pool exhaustion and cache key design, which general advice often skips.
Under The Hood
The skill loads a SKILL.md and, when needed, a references file of worked patterns. The agent runs profiling and measurement tools such as Lighthouse, the web-vitals library, query EXPLAIN and bundle size checks, and may add CI steps with npx. It ships no scripts.