Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →PROACTIVE SKILL — automatically apply whenever writing, generating, or editing any text that is mainly in Chinese. No user trigger needed. Also use when the user explicitly asks to "fix Chinese", "修改中文", "去翻译腔", "去AI味", "fix Chinese formatting", or "review Chinese text". Eliminates AI-sounding expressions and translation artifacts. Enforces Chinese-English mixed formatting rules (spacing, punctuation, bold formatting).
.claude/skills/brycewang-stanford-fix-chinese/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 59% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | -11% | 0% |
Review the provided text or file and apply all fixes below. Make edits in-place.
Claude Code 是, 第 1 章**这里** 和**那里**| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,812 | 2,171 | -68% | 1 | 1 | 0% | 1,201 | 591 | -51% | 0 | 0 | — |
case-02 | fail→pass | 6,962 | 3,488 | -50% | 1 | 1 | 0% | 1,283 | 968 | -25% | 0 | 0 | — |
case-03 | pass→pass | 6,445 | 3,906 | -39% | 1 | 1 | 0% | 1,156 | 1,024 | -11% | 0 | 0 | — |
case-04 | pass→pass | 9,191 | 5,351 | -42% | 1 | 1 | 0% | 1,547 | 1,173 | -24% | 0 | 0 | — |
case-05 | fail→fail | 7,273 | 2,612 | -64% | 1 | 1 | 0% | 1,314 | 695 | -47% | 0 | 0 | — |
case-06 | pass→pass | 7,439 | 5,908 | -21% | 1 | 1 | 0% | 1,368 | 1,308 | -4% | 0 | 0 | — |
case-07 | fail→pass | 5,587 | 6,422 | +15% | 1 | 1 | 0% | 1,002 | 1,593 | +59% | 0 | 0 | — |
case-08 | pass→pass | 6,401 | 3,819 | -40% | 1 | 1 | 0% | 1,212 | 896 | -26% | 0 | 0 | — |
case-09 | pass→pass | 6,891 | 3,898 | -43% | 1 | 1 | 0% | 1,154 | 779 | -32% | 0 | 0 | — |
case-10 | pass→pass | 8,033 | 5,517 | -31% | 1 | 1 | 0% | 1,595 | 1,383 | -13% | 0 | 0 | — |
case-11 | pass→pass | 6,749 | 4,901 | -27% | 1 | 1 | 0% | 1,268 | 1,320 | +4% | 0 | 0 | — |
case-12 | pass→pass | 7,343 | 3,022 | -59% | 1 | 1 | 0% | 1,382 | 880 | -36% | 0 | 0 | — |
case-13 | pass→pass | 6,235 | 3,959 | -37% | 1 | 1 | 0% | 1,243 | 1,042 | -16% | 0 | 0 | — |
case-14 | fail→fail | 6,398 | 4,365 | -32% | 1 | 1 | 0% | 1,265 | 1,161 | -8% | 0 | 0 | — |
case-15 | fail→pass | 6,943 | 2,868 | -59% | 1 | 1 | 0% | 1,301 | 771 | -41% | 0 | 0 | — |
case-16 | pass→pass | 7,459 | 7,183 | -4% | 1 | 1 | 0% | 1,344 | 1,312 | -2% | 0 | 0 | — |
case-17 | fail→pass | 4,242 | 2,969 | -30% | 1 | 1 | 0% | 886 | 876 | -1% | 0 | 0 | — |
case-18 | pass→pass | 5,473 | 3,672 | -33% | 1 | 1 | 0% | 1,094 | 1,050 | -4% | 0 | 0 | — |
case-19 | pass→pass | 7,745 | 5,958 | -23% | 1 | 1 | 0% | 1,405 | 1,595 | +14% | 0 | 0 | — |
case-20 | pass→pass | 4,416 | 3,046 | -31% | 1 | 1 | 0% | 920 | 860 | -7% | 0 | 0 | — |
case-21 | pass→pass | 23,834 | 21,401 | -10% | 1 | 1 | 0% | 4,120 | 4,373 | +6% | 0 | 0 | — |
case-22 | pass→pass | 16,287 | 15,451 | -5% | 1 | 1 | 0% | 3,262 | 3,926 | +20% | 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 +18 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.