Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →按「AI浪」调性写作——人工智能与智能体方向，快节奏口语化的技术讲解。当用户要写 AI 底层原理（大模型/注意力/幻觉）、工程实践（RAG/Agent 设计）、bug 排查、前沿技术（新模型/新框架）、工具对比（LangChain/LlamaIndex）等主题时使用。不适合纯后端/架构（→wei-danxin）、生活省钱（→wei-mingdeng）。
.claude/skills/helloworldtang-wei-ailang/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 21% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 12% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 13% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 27% | 0% |
专注 AI 与智能体领域的知识分享，已认证科技自媒体。读者是对 AI 底层原理、工程实践、前沿技术感兴趣的人。
wc -m 落在区间../references/writing-standards.md 禁用词表）> 引用块）../references/writing-standards.md../references/dedup-rules.mdwei-danxin；生活省钱/职场非技术 → wei-mingdeng| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 26,017 | 24,259 | -7% | 1 | 1 | 0% | 4,031 | 4,865 | +21% | 0 | 0 | — |
case-02 | fail→fail | 24,559 | 24,610 | +0% | 1 | 1 | 0% | 4,419 | 4,743 | +7% | 0 | 0 | — |
case-03 | fail→pass | 27,821 | 25,080 | -10% | 1 | 1 | 0% | 4,023 | 4,505 | +12% | 0 | 0 | — |
case-04 | pass→fail | 29,673 | 35,273 | +19% | 1 | 1 | 0% | 4,837 | 6,369 | +32% | 0 | 0 | — |
case-05 | fail→fail | 24,223 | 26,539 | +10% | 1 | 1 | 0% | 3,844 | 5,614 | +46% | 0 | 0 | — |
case-06 | pass→fail | 28,674 | 26,342 | -8% | 1 | 1 | 0% | 3,887 | 4,651 | +20% | 0 | 0 | — |
case-07 | pass→pass | 24,072 | 23,065 | -4% | 1 | 1 | 0% | 4,102 | 4,375 | +7% | 0 | 0 | — |
case-08 | fail→pass | 28,353 | 26,462 | -7% | 1 | 1 | 0% | 4,707 | 5,080 | +8% | 0 | 0 | — |
case-09 | pass→pass | 24,614 | 25,099 | +2% | 1 | 1 | 0% | 3,758 | 4,673 | +24% | 0 | 0 | — |
case-10 | fail→fail | 23,925 | 25,618 | +7% | 1 | 1 | 0% | 3,866 | 4,329 | +12% | 0 | 0 | — |
case-11 | fail→pass | 23,905 | 25,186 | +5% | 1 | 1 | 0% | 4,198 | 4,763 | +13% | 0 | 0 | — |
case-12 | fail→pass | 27,111 | 31,224 | +15% | 1 | 1 | 0% | 5,016 | 6,384 | +27% | 0 | 0 | — |
case-13 | fail→pass | 21,230 | 21,569 | +2% | 1 | 1 | 0% | 3,581 | 4,323 | +21% | 0 | 0 | — |
case-14 | pass→pass | 26,751 | 34,276 | +28% | 1 | 1 | 0% | 4,465 | 6,335 | +42% | 0 | 0 | — |
case-15 | fail→fail | 23,560 | 24,527 | +4% | 1 | 1 | 0% | 3,941 | 5,037 | +28% | 0 | 0 | — |
case-16 | fail→pass | 25,193 | 26,126 | +4% | 1 | 1 | 0% | 4,376 | 5,463 | +25% | 0 | 0 | — |
case-17 | fail→pass | 24,669 | 26,394 | +7% | 1 | 1 | 0% | 4,311 | 5,410 | +25% | 0 | 0 | — |
case-18 | fail→pass | 20,361 | 25,569 | +26% | 1 | 1 | 0% | 3,494 | 4,861 | +39% | 0 | 0 | — |
case-19 | fail→fail | 26,274 | 19,897 | -24% | 1 | 1 | 0% | 4,887 | 3,994 | -18% | 0 | 0 | — |
case-20 | pass→pass | 25,886 | 28,135 | +9% | 1 | 1 | 0% | 4,384 | 5,668 | +29% | 0 | 0 | — |
case-21 | fail→pass | 31,541 | 23,634 | -25% | 1 | 1 | 0% | 5,128 | 5,043 | -2% | 0 | 0 | — |
case-22 | fail→pass | 23,444 | 21,925 | -6% | 1 | 1 | 0% | 4,129 | 4,134 | +0% | 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 +41 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/3/2026 | +5% |
Other measured skills in the registry, with their headline benchmark lift.