Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Opinionated coding style and tooling preferences by Anthony Fu.
.claude/skills/dokhacgiakhoa-antfu-coding-style/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-02 | ✓→✓ | = Same ✓ | -6% | 0% |
| case-04 | ✓→✓ | = Same ✓ | -43% | 0% |
> Goal: Maximal Developer Experience (DX) through minimalism and smart defaults. "Works out of the box."
constants.ts.const count: number = 0const count = 0const assertion object or String Unions.type Mode = 'dark' | 'light' > enum Mode { ... }pnpm (Fast, disk efficient).ni (NPM/Yarn/PNPM/Bun Agnostic runner).ni -> installnr dev -> run devnu -> upgradeVitest (Fast, Jest compatible, Native ESM).*.test.ts.toMatchSnapshot() sparingly, prefer toMatchInlineSnapshot().<div text="red 4xl" hover="text-blue"> (Controversial but preferred by Antfu for clean template).<script setup lang="ts">.ref with .value or explicit .value macros if configured.defineProps and defineEmits at the top.watch / computed grouped together.The "Antfu" Vibe:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 24,103 | 22,566 | -6% | 1 | 1 | 0% | 4,864 | 3,102 | -36% | 0 | 0 | — |
case-02 | pass→pass | 16,956 | 9,915 | -42% | 1 | 1 | 0% | 2,845 | 2,686 | -6% | 0 | 0 | — |
case-03 | fail→pass | 15,609 | 10,864 | -30% | 1 | 1 | 0% | 3,102 | 2,382 | -23% | 0 | 0 | — |
case-04 | pass→pass | 13,947 | 6,582 | -53% | 1 | 1 | 0% | 2,601 | 1,492 | -43% | 0 | 0 | — |
case-05 | fail→pass | 13,480 | 5,191 | -61% | 1 | 1 | 0% | 2,096 | 1,509 | -28% | 0 | 0 | — |
case-06 | pass→pass | 10,462 | 9,955 | -5% | 1 | 1 | 0% | 2,012 | 1,998 | -1% | 0 | 0 | — |
case-07 | pass→pass | 16,156 | 6,831 | -58% | 1 | 1 | 0% | 2,260 | 1,586 | -30% | 0 | 0 | — |
case-08 | pass→pass | 11,742 | 5,954 | -49% | 1 | 1 | 0% | 2,159 | 1,655 | -23% | 0 | 0 | — |
case-09 | pass→pass | 15,098 | 6,241 | -59% | 1 | 1 | 0% | 2,553 | 1,636 | -36% | 0 | 0 | — |
case-10 | pass→pass | 12,033 | 4,127 | -66% | 1 | 1 | 0% | 2,132 | 1,273 | -40% | 0 | 0 | — |
case-11 | pass→pass | 13,055 | 8,252 | -37% | 1 | 1 | 0% | 2,278 | 1,957 | -14% | 0 | 0 | — |
case-12 | pass→pass | 4,902 | 2,396 | -51% | 1 | 1 | 0% | 833 | 933 | +12% | 0 | 0 | — |
case-13 | pass→pass | 4,937 | 3,258 | -34% | 1 | 1 | 0% | 905 | 1,056 | +17% | 0 | 0 | — |
case-14 | pass→pass | 10,251 | 5,760 | -44% | 1 | 1 | 0% | 1,587 | 1,623 | +2% | 0 | 0 | — |
case-15 | pass→pass | 12,097 | 5,943 | -51% | 1 | 1 | 0% | 2,085 | 1,612 | -23% | 0 | 0 | — |
case-16 | pass→pass | 10,395 | 4,685 | -55% | 1 | 1 | 0% | 1,905 | 1,130 | -41% | 0 | 0 | — |
case-17 | pass→pass | 11,431 | 8,498 | -26% | 1 | 1 | 0% | 1,935 | 1,747 | -10% | 0 | 0 | — |
case-18 | pass→pass | 10,292 | 11,061 | +7% | 1 | 1 | 0% | 1,733 | 2,104 | +21% | 0 | 0 | — |
case-19 | pass→pass | 9,638 | 3,845 | -60% | 1 | 1 | 0% | 1,861 | 1,220 | -34% | 0 | 0 | — |
case-20 | fail→fail | 13,639 | 9,814 | -28% | 1 | 1 | 0% | 2,972 | 2,526 | -15% | 0 | 0 | — |
case-21 | pass→pass | 11,220 | 5,188 | -54% | 1 | 1 | 0% | 1,805 | 1,619 | -10% | 0 | 0 | — |
case-22 | pass→pass | 7,586 | 4,452 | -41% | 1 | 1 | 0% | 1,228 | 1,245 | +1% | 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 +14 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.