Systematically finds and fixes performance bottlenecks by measuring first, profiling hot paths, reducing algorithmic and I/O cost, and verifying gains with before/after benchmarks. Use this skill when code, an endpoint, a page, a query, or a job is "slow", "laggy", "timing out", "using too much CPU/memory", or "doesn't scale"; when asked to optimize, speed up, profile, or benchmark something; when chasing high latency / low throughput, p99 spikes, memory leaks, N+1 queries, or excessive allocations; or before/after a perf-sensitive change to prove it actually helped. Covers Python, Node/JS, Go, Java, SQL, and HTTP/web frontends.
