performance-benchmark

Measure performance baselines, detect regressions, and compare stack alternatives before/after changes. Modes: page performance (Core Web Vitals), API latenc...

This skill is a coherent, instruction-only benchmarking guide — it does not install code or ask for credentials. Before using it: (1) ensure you have the benchmark tools it references (k6, autocannon, hyperfine, or a 'benchmark' CLI) or provide your own implementations; (2) never run load tests against production without explicit authorization — the SKILL.md includes the same warning; (3) review what you store in .benchmarks/ before committing (it may contain timing data or URLs you don't want public); (4) if you plan to run benchmarks from CI, ensure your runners and secrets are properly scoped; and (5) if you expect the skill to provide a runnable CLI, note it doesn't — you must map the documented commands to real tools or scripts. Overall this is internally consistent and does what it says.