Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when targeting 《经济学(季刊)》(China Economic Quarterly, CEQ), the Peking University / CCER journal closest to international top-field economics standards in China. Apply to enforce a clean structural model or credibly identified causal design, rigorous econometric exposition, frontier-aligned framing, and CEQ formatting. Strictest on methodology among Chinese econ journals.
.claude/skills/brycewang-stanford-china-economic-quarterly/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 12% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 53% | 0% |
北京大学中国经济研究中心（CCER）主办，是国内最贴近国际主流经济学范式的中文刊。编委与审稿人多有海外训练，看重"问题—模型/识别—证据"的逻辑闭环。可发表中文或英文。与《经济研究》相比，CEQ 更强调方法规范性与技术严谨，对政策口号的容忍度更低。
business-management-journal（《经济管理》） / china-accounting-review（《会计评论》） / china-economic-studies（《中国经济问题》） / china-industrial-economics（《中国工业经济》）。若这些刊物的读者对象更贴近，不要因为名称、地区或变量相似而强行投本刊。../../resources/source-basis.md 和 ../../resources/official-source-map.md，从其中的官方来源锚点或同一主办/出版体系入口开始核验，并在回答中说明核验日期。【匹配度】高 / 中 / 低
【识别策略】结构 / 准实验(类型) / 仅相关（需补）
【现代DID合规】是 / 否（缺哪一步）
【推断风险】<聚类/弱工具/多重检验 待补项>
【贡献定位】对标到 <具体文献> | 仍是套话（需改）
【建议去向】CEQ / 经济研究 / 行业刊| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | fail→fail | 15,253 | 19,570 | +28% | 1 | 1 | 0% | 1,213 | 3,426 | +182% | 0 | 0 | — |
case-07 | fail→pass | 28,526 | 20,810 | -27% | 1 | 1 | 0% | 2,828 | 3,323 | +18% | 0 | 0 | — |
case-08 | pass→pass | 29,701 | 33,500 | +13% | 1 | 1 | 0% | 3,151 | 4,925 | +56% | 0 | 0 | — |
case-01 | fail→fail | 31,961 | 25,162 | -21% | 1 | 1 | 0% | 3,509 | 4,044 | +15% | 0 | 0 | — |
case-02 | fail→pass | 35,508 | 27,756 | -22% | 1 | 1 | 0% | 4,012 | 4,444 | +11% | 0 | 0 | — |
case-03 | fail→pass | 32,643 | 31,576 | -3% | 1 | 1 | 0% | 4,199 | 4,723 | +12% | 0 | 0 | — |
case-04 | pass→fail | 28,136 | 31,553 | +12% | 1 | 1 | 0% | 3,147 | 4,779 | +52% | 0 | 0 | — |
case-05 | pass→pass | 32,467 | 24,945 | -23% | 1 | 1 | 0% | 3,581 | 4,758 | +33% | 0 | 0 | — |
case-09 | fail→pass | 26,492 | 23,145 | -13% | 1 | 1 | 0% | 2,611 | 3,935 | +51% | 0 | 0 | — |
case-10 | pass→pass | 28,052 | 23,676 | -16% | 1 | 1 | 0% | 2,802 | 4,084 | +46% | 0 | 0 | — |
case-11 | pass→pass | 29,126 | 25,490 | -12% | 1 | 1 | 0% | 2,628 | 4,032 | +53% | 0 | 0 | — |
case-12 | fail→pass | 28,631 | 30,394 | +6% | 1 | 1 | 0% | 2,982 | 4,561 | +53% | 0 | 0 | — |
case-13 | fail→pass | 32,604 | 25,046 | -23% | 1 | 1 | 0% | 3,400 | 4,164 | +22% | 0 | 0 | — |
case-14 | pass→pass | 20,661 | 26,659 | +29% | 1 | 1 | 0% | 2,565 | 3,961 | +54% | 0 | 0 | — |
case-15 | fail→pass | 26,432 | 31,810 | +20% | 1 | 1 | 0% | 2,973 | 4,881 | +64% | 0 | 0 | — |
case-16 | pass→pass | 21,675 | 41,901 | +93% | 1 | 1 | 0% | 3,101 | 4,298 | +39% | 0 | 0 | — |
case-17 | pass→pass | 24,234 | 26,935 | +11% | 1 | 1 | 0% | 2,526 | 4,020 | +59% | 0 | 0 | — |
case-18 | pass→pass | 25,822 | 20,320 | -21% | 1 | 1 | 0% | 2,596 | 3,825 | +47% | 0 | 0 | — |
case-19 | fail→pass | 31,331 | 28,510 | -9% | 1 | 1 | 0% | 3,468 | 4,555 | +31% | 0 | 0 | — |
case-20 | fail→pass | 27,362 | 28,059 | +3% | 1 | 1 | 0% | 3,097 | 4,273 | +38% | 0 | 0 | — |
case-21 | pass→pass | 34,575 | 20,161 | -42% | 1 | 1 | 0% | 3,538 | 4,157 | +17% | 0 | 0 | — |
case-22 | pass→pass | 30,538 | 29,380 | -4% | 1 | 1 | 0% | 3,386 | 4,394 | +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. 22 cases were attempted. The headline lift of +36 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.
Other measured skills in the registry, with their headline benchmark lift.