Install any skill in seconds. Free to start, no credit card required.
Get Started Free →React & Next.js engineering standards.
.claude/skills/dokhacgiakhoa-react-best-practices/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-01 | ✓→✓ | = Same ✓ | 15% | 0% |
| case-03 | ✓→✓ | = Same ✓ | 12% | 0% |
| case-04 | ✓→✓ | = Same ✓ | -2% | 0% |
| case-05 | ✓→✓ | = Same ✓ | 27% | 0% |
> Goal: Build high-performance, maintainable, and accessible React applications using modern patterns (React 19+, Next.js App Router).
app/) for all new projects.'use client' when interactivity (state, hooks, event listeners) is strictly needed.<Suspense> boundaries to stream UI parts that depend on slow data.useEffect for data fetching.children prop to tackle prop drilling.useHookName.any.next/image with proper width, height, and sizes.next/font to eliminate layout shift (CLS).next/dynamic or React.lazy for heavy client components below the fold.@next/bundle-analyzer.clsx + tailwind-merge (typically cn() helper) for conditional class merging.md:, lg: prefixes.textapp/ ├── (marketing)/ # Route group ├── dashboard/ │ ├── page.tsx │ ├── layout.tsx │ ├── _components/ # Private feature components │ └── loading.tsx components/ # Shared components (UI Kit) lib/ # Utilities and helpers hooks/ # Shared hooks types/ # Global types
dangerouslySetInnerHTML.NextAuth.js (Auth.js) or Clerk. Protect routes via Middleware.Checklist before PR:
'use client' used only where necessary?key props?| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 19,161 | 17,057 | -11% | 1 | 1 | 0% | 3,588 | 4,119 | +15% | 0 | 0 | — |
case-02 | fail→pass | 22,981 | 19,182 | -17% | 1 | 1 | 0% | 4,727 | 4,744 | +0% | 0 | 0 | — |
case-03 | pass→pass | 17,284 | 14,418 | -17% | 1 | 1 | 0% | 3,272 | 3,679 | +12% | 0 | 0 | — |
case-04 | pass→pass | 20,021 | 11,697 | -42% | 1 | 1 | 0% | 3,133 | 3,060 | -2% | 0 | 0 | — |
case-05 | pass→pass | 13,833 | 11,993 | -13% | 1 | 1 | 0% | 2,413 | 3,073 | +27% | 0 | 0 | — |
case-06 | pass→pass | 14,819 | 11,452 | -23% | 1 | 1 | 0% | 2,719 | 3,048 | +12% | 0 | 0 | — |
case-07 | pass→pass | 10,298 | 8,615 | -16% | 1 | 1 | 0% | 1,841 | 1,974 | +7% | 0 | 0 | — |
case-08 | pass→pass | 9,839 | 8,107 | -18% | 1 | 1 | 0% | 1,729 | 2,224 | +29% | 0 | 0 | — |
case-09 | pass→pass | 14,913 | 11,144 | -25% | 1 | 1 | 0% | 2,198 | 2,793 | +27% | 0 | 0 | — |
case-10 | pass→pass | 13,239 | 14,356 | +8% | 1 | 1 | 0% | 2,496 | 3,411 | +37% | 0 | 0 | — |
case-11 | fail→fail | 11,760 | 8,720 | -26% | 1 | 1 | 0% | 2,150 | 2,270 | +6% | 0 | 0 | — |
case-12 | pass→pass | 11,257 | 11,715 | +4% | 1 | 1 | 0% | 1,979 | 2,459 | +24% | 0 | 0 | — |
case-13 | pass→pass | 13,764 | 17,372 | +26% | 1 | 1 | 0% | 2,339 | 3,201 | +37% | 0 | 0 | — |
case-14 | pass→pass | 9,719 | 6,874 | -29% | 1 | 1 | 0% | 1,875 | 1,992 | +6% | 0 | 0 | — |
case-15 | pass→pass | 12,558 | 10,842 | -14% | 1 | 1 | 0% | 2,348 | 2,920 | +24% | 0 | 0 | — |
case-16 | pass→pass | 21,183 | 16,835 | -21% | 1 | 1 | 0% | 2,938 | 3,719 | +27% | 0 | 0 | — |
case-17 | pass→pass | 8,567 | 6,481 | -24% | 1 | 1 | 0% | 1,703 | 1,650 | -3% | 0 | 0 | — |
case-18 | pass→pass | 13,155 | 7,811 | -41% | 1 | 1 | 0% | 2,300 | 2,118 | -8% | 0 | 0 | — |
case-19 | pass→pass | 13,659 | 12,022 | -12% | 1 | 1 | 0% | 2,466 | 2,680 | +9% | 0 | 0 | — |
case-20 | pass→pass | 16,345 | 16,199 | -1% | 1 | 1 | 0% | 3,071 | 3,296 | +7% | 0 | 0 | — |
case-21 | pass→pass | 19,017 | 15,317 | -19% | 1 | 1 | 0% | 3,898 | 3,870 | -1% | 0 | 0 | — |
case-22 | pass→pass | 16,709 | 14,135 | -15% | 1 | 1 | 0% | 3,142 | 3,306 | +5% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +5 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.