Qa perf
Opinionated Claude Code foundation — Explore → TDD → Audit workflow, auto-detected stack presets (nextjs, fastapi, astro, ...), curl | bash install. MIT.
npx -y skills add christopherlouet/claude-base --skill qa-perfAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 5 stars5 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Application performance optimization. Trigger when the user wants to improve speed, reduce latency, or optimize resources.
SKILL.md
3.1 KB, 783 tokens by cl100k_base, as published. Nobody here has run it
Performance Optimization (pointer)
Canonical thresholds, current Web Vitals (LCP/INP/CLS — note INP replaced FID in 2024), and Chrome perf-team remediation patterns are at:
addyosmani/web-quality-skills— github.com/addyosmani/web-quality-skills (MIT, 1.8k★, maintained by Addy Osmani — Chrome DevTools / Lighthouse engineering lead). Covers Core Web Vitals, perf, a11y, SEO.- web.dev/vitals — web.dev/vitals (Google's canonical Web Vitals reference)
- Vercel React best practices — see
vercel-react-best-practicesskill (foundation-installed) for React-specific patterns
Foundation workflow (when to invoke this skill)
qa-perf is dispatched by qa-loop during the AUDIT phase, in parallel with qa-security / wcag-audit / qa-claudemd. It's a measurement workflow, not an optimisation cookbook:
- Measure first: run Lighthouse / WebPageTest / DevTools Performance against a known scope (the URL, page, or endpoint from
argument-hint). - Compare to canonical thresholds (see table below).
- Identify the bottleneck axis: render-blocking JS? N+1 DB query? Image weight? Bundle size? Each axis has a dedicated vendor remediation guide.
- Recommend with quantified impact (e.g. "lazy-loading hero image saves ~400ms LCP per Lighthouse run #3").
- Re-measure after the fix — a perf change without before/after numbers is theatre.
Canonical Web Vitals thresholds (2024-2026)
| Metric | Good | Needs improvement | Poor | Tool |
|---|---|---|---|---|
| LCP (Largest Contentful Paint) | < 2.5s | 2.5–4s | > 4s | Lighthouse, web-vitals |
| INP (Interaction to Next Paint, replaces FID) | < 200ms | 200–500ms | > 500ms | web-vitals |
| CLS (Cumulative Layout Shift) | < 0.1 | 0.1–0.25 | > 0.25 | Lighthouse, web-vitals |
| TTFB | < 200ms | 200–600ms | > 600ms | DevTools Network |
Foundation discipline (keep across releases)
- No optimisation without measurement: profile before changing code. Guessed bottlenecks are wrong ~70% of the time.
- Before/after numbers mandatory: every perf PR must include the Lighthouse delta or equivalent. Without numbers, the work is unprovable.
- Cache invalidation > caching: adding a cache is easy; correctly invalidating it is the bug surface. Surface cache TTLs in code review.
- N+1 is the #1 backend perf bug: when an endpoint feels slow, instrument query count before optimising anything else.
See also
qa-chromeskill — DevTools manual review (paired layer)dev-react-perfskill — React-specific re-render audit + memoization patternsops-monitoring— production perf instrumentation (OTEL, RUM)vercel-react-best-practicesskill (foundation-installed)- Audit pilot trace:
specs/marketplace-audit/qa-skills-pilot-2026-05-06.md
What ships with it: 1 file
2.4 KB alongside SKILL.md
examples/
- lighthouse-audit.md2.4 KB
Gives 0 of the 12 instructions most performance cost skills give in 783 tokens
Counted across 803 of the 1,058 authors here whose files we hold, read 2026-08-07
- Keep skill files under 500 lines or tokensin 82 of 803, across 16 files
- Use imperative form in instructionsin 80 of 803, across 9 files
- Draft assertions while test runs are in progressin 75 of 803, across 9 files
- Create two to three realistic test promptsin 74 of 803, across 9 files
- Write skill descriptions to be pushyin 72 of 803, across 7 files
- Save test cases to evals JSONin 72 of 803, across 6 files
- Ask questions about edge cases and input formatsin 72 of 803, across 7 files
- Save timing data immediately when runs completein 70 of 803, across 5 files
- Include all trigger conditions in the skill descriptionin 69 of 803, across 3 files
- Launch all test runs in a single turn or simultaneouslyin 69 of 803, across 3 files
- Capture intent before writing a skillin 67 of 803, across 1 file
- Import directly instead of barrel filesin 52 of 803, across 15 files
Said here and by no other author read
- run performance tools against a known scope
- compare results to canonical thresholds
- recommend fixes with quantified impact
- instrument query count for slow endpoints
- surface cache TTLs in code review
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.