Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →知乎回答/文章写作 - 回答结构、专业语气、引用规范、话题关联、盐值优化
.claude/skills/dongsheng123132-zhihu-writer/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-18 | ✗→✓ | ▲ Improved | 26% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-02 | ✓→✗ | ▼ Worse | 28% | 0% |
| case-07 | ✓→✗ | ▼ Worse | 37% | 0% |
| case-11 | ✓→✗ | ▼ Worse | 21% | 0% |
帮助创作者写出高赞、高收藏的知乎内容，包括问题回答和专栏文章。
【开头】直接回答问题 + 亮明立场（1-2句）
"作为一个做了5年XX的人，我的回答是：..."
【分割线】
---
【正文】分点论述，有理有据
一、第一个论点
论据 + 数据/案例支撑
（可配图表增强说服力）
二、第二个论点
论据 + 个人经验
三、第三个论点
论据 + 行业分析
【总结】回扣问题，给出实操建议
【引流】（可选）
更多XX内容可以关注我的专栏：XXX帮我回答知乎问题"普通人如何开始学习AI？"，要专业但通俗帮我写一篇知乎专栏文章，主题是"为什么我不建议用ChatGPT学编程"我的这个回答数据不好，帮我分析并改写（附原文）最近DeepSeek很火，帮我写一个相关的知乎回答思路| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 23,317 | 38,933 | +67% | 1 | 1 | 0% | 3,234 | 3,839 | +19% | 0 | 0 | — |
case-02 | pass→fail | 20,191 | 25,012 | +24% | 1 | 1 | 0% | 3,219 | 4,115 | +28% | 0 | 0 | — |
case-03 | pass→pass | 17,853 | 16,942 | -5% | 1 | 1 | 0% | 2,118 | 3,083 | +46% | 0 | 0 | — |
case-04 | fail→fail | 15,027 | 19,341 | +29% | 1 | 1 | 0% | 1,989 | 3,259 | +64% | 0 | 0 | — |
case-05 | pass→pass | 8,318 | 9,195 | +11% | 1 | 1 | 0% | 1,339 | 2,063 | +54% | 0 | 0 | — |
case-06 | pass→pass | 9,345 | 13,489 | +44% | 1 | 1 | 0% | 1,165 | 2,392 | +105% | 0 | 0 | — |
case-07 | pass→fail | 28,408 | 29,984 | +6% | 1 | 1 | 0% | 3,603 | 4,927 | +37% | 0 | 0 | — |
case-08 | fail→fail | 27,767 | 26,363 | -5% | 1 | 1 | 0% | 3,475 | 3,987 | +15% | 0 | 0 | — |
case-09 | fail→fail | 16,560 | 16,266 | -2% | 1 | 1 | 0% | 2,434 | 3,009 | +24% | 0 | 0 | — |
case-10 | pass→pass | 22,501 | 26,928 | +20% | 1 | 1 | 0% | 3,320 | 4,225 | +27% | 0 | 0 | — |
case-11 | pass→fail | 26,268 | 26,864 | +2% | 1 | 1 | 0% | 3,540 | 4,278 | +21% | 0 | 0 | — |
case-12 | fail→fail | 18,048 | 17,732 | -2% | 1 | 1 | 0% | 2,669 | 3,363 | +26% | 0 | 0 | — |
case-13 | pass→pass | 18,154 | 16,660 | -8% | 1 | 1 | 0% | 2,408 | 3,313 | +38% | 0 | 0 | — |
case-14 | fail→fail | 27,877 | 23,133 | -17% | 1 | 1 | 0% | 3,560 | 3,635 | +2% | 0 | 0 | — |
case-15 | pass→pass | 20,708 | 22,467 | +8% | 1 | 1 | 0% | 2,622 | 3,453 | +32% | 0 | 0 | — |
case-16 | pass→pass | 22,720 | 25,707 | +13% | 1 | 1 | 0% | 3,027 | 4,027 | +33% | 0 | 0 | — |
case-17 | pass→pass | 15,968 | 14,571 | -9% | 1 | 1 | 0% | 2,331 | 2,943 | +26% | 0 | 0 | — |
case-18 | fail→pass | 19,096 | 22,372 | +17% | 1 | 1 | 0% | 2,815 | 3,549 | +26% | 0 | 0 | — |
case-19 | fail→fail | 26,279 | 23,529 | -10% | 1 | 1 | 0% | 3,259 | 3,736 | +15% | 0 | 0 | — |
case-20 | fail→fail | 20,544 | 18,877 | -8% | 1 | 1 | 0% | 2,712 | 3,522 | +30% | 0 | 0 | — |
case-21 | fail→pass | 23,550 | 21,116 | -10% | 1 | 1 | 0% | 3,056 | 3,655 | +20% | 0 | 0 | — |
case-22 | pass→pass | 30,665 | 27,753 | -9% | 1 | 1 | 0% | 3,881 | 4,331 | +12% | 0 | 0 | — |
case-23 | fail→fail | 17,112 | 19,241 | +12% | 1 | 1 | 0% | 2,529 | 3,289 | +30% | 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. 23 cases were attempted. The headline lift of -4 percentage points is the difference between those two pass rates over the 23 comparable cases. 3 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.
Other measured skills in the registry, with their headline benchmark lift.