Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when drafting or revising the literature-review section for a 《中国农村经济》 manuscript, when the Chinese / English reference balance is off, or when canonical 三农 theory references are missing. 本技能服务于《中国农村经济》(China Rural Economy, CRE)。
.claude/skills/brycewang-stanford-cre-literature-review/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 53% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 54% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 56% | 0% |
《中国农村经济》偏好三层结构的综述：
《中国农村经济》尤其看重三农领域文献是否到位：既要有国际理论与方法文献，也要充分对话国内农经学界的权威研究。一篇只引国际顶刊、不引国内农经经典的稿件，会被审稿人认为"未扎根本土文献"。
经验法则（具体不必拘泥，导向是"中外并重 + 理论扎实"）：
[作者 (年份)] 利用[农户 / 村级数据 + 方法]研究了[三农问题]，发现……。
然而，该研究[局限]，未考虑[本文要补的空白，如农户自选择 / 区域异质 / 机制]。
与之相比，本文……。每段综述至少出现一次"与之相比 / 区别于 / 本文延伸了 / 本文补充了"。
《中国农村经济》偏好对话式综述（研究之间相互回应），而非罗列式（按作者顺序点名）。
> 张三（2020）研究了土地流转，李四（2021）研究了农户增收，王五（2022）研究了合作社。
> 土地流转能否提升农户收入，已有研究存在分歧（张三，2020；李四，2021）：一类研究认为流转通过规模经营提高生产率，另一类则强调流转中的契约不完全限制了增收效果。两类判断的差异部分源于对农户自选择的处理不同。本文沿这一思路，使用数据]并构造识别策略]，进一步考察了……。
判断标准：把作者名都隐去后，综述还能读出"研究 A 怎么说、研究 B 怎么反驳、本文怎么定位"，就是对话式。
综述末尾必须导出本文的差异化定位，模板：
> 本文与现有文献的重要区别体现在以下几个方面： > 第一，识别策略上的差异——如何处理农户自选择 / 政策内生]； > 第二，数据或样本上的差异——更新 / 更细的农户或村级数据]； > 第三，机制或异质性上的差异]。 > 与本刊（或《中国农村观察》）近期发表的 XXX（作者，202X）一文相比，本文 ……
与本刊及姊妹刊已发表文章对比是审稿人的关注点——综述若不提近 3 年两刊同主题文章，会被视为"未充分对话"。
【结构】三层 / 两层 / 一层
【中外配比】中文 X% / 英文 Y%
【理论文献占比】X%
【三农本土权威文献】到位 / 缺位
【五年内文献比例】X%
【缺漏清单】[...]
【下一步】cre-identification| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 25,986 | 28,856 | +11% | 1 | 1 | 0% | 3,168 | 4,858 | +53% | 0 | 0 | — |
case-15 | pass→pass | 15,559 | 18,091 | +16% | 1 | 1 | 0% | 2,410 | 3,837 | +59% | 0 | 0 | — |
case-02 | fail→pass | 26,072 | 27,584 | +6% | 1 | 1 | 0% | 3,113 | 4,444 | +43% | 0 | 0 | — |
case-03 | pass→fail | 32,493 | 34,404 | +6% | 1 | 1 | 0% | 4,382 | 6,029 | +38% | 0 | 0 | — |
case-04 | pass→fail | 24,295 | 29,470 | +21% | 1 | 1 | 0% | 4,517 | 5,462 | +21% | 0 | 0 | — |
case-05 | pass→pass | 22,582 | 21,480 | -5% | 1 | 1 | 0% | 2,480 | 3,946 | +59% | 0 | 0 | — |
case-12 | pass→pass | 23,241 | 22,060 | -5% | 1 | 1 | 0% | 2,758 | 4,087 | +48% | 0 | 0 | — |
case-06 | pass→fail | 31,811 | 34,153 | +7% | 1 | 1 | 0% | 4,239 | 6,006 | +42% | 0 | 0 | — |
case-07 | fail→pass | 24,027 | 15,637 | -35% | 1 | 1 | 0% | 2,785 | 3,874 | +39% | 0 | 0 | — |
case-08 | pass→pass | 30,948 | 24,100 | -22% | 1 | 1 | 0% | 3,290 | 4,263 | +30% | 0 | 0 | — |
case-09 | fail→pass | 28,870 | 27,562 | -5% | 1 | 1 | 0% | 3,156 | 4,849 | +54% | 0 | 0 | — |
case-10 | pass→pass | 27,471 | 27,683 | +1% | 1 | 1 | 0% | 3,235 | 4,512 | +39% | 0 | 0 | — |
case-11 | fail→pass | 23,863 | 24,206 | +1% | 1 | 1 | 0% | 2,659 | 4,152 | +56% | 0 | 0 | — |
case-13 | pass→pass | 23,174 | 21,737 | -6% | 1 | 1 | 0% | 2,481 | 3,991 | +61% | 0 | 0 | — |
case-14 | pass→pass | 25,525 | 19,904 | -22% | 1 | 1 | 0% | 2,604 | 3,525 | +35% | 0 | 0 | — |
case-16 | fail→pass | 25,736 | 27,056 | +5% | 1 | 1 | 0% | 3,526 | 4,489 | +27% | 0 | 0 | — |
case-17 | fail→pass | 21,211 | 25,718 | +21% | 1 | 1 | 0% | 2,680 | 4,267 | +59% | 0 | 0 | — |
case-18 | pass→pass | 18,752 | 18,493 | -1% | 1 | 1 | 0% | 2,538 | 3,807 | +50% | 0 | 0 | — |
case-19 | pass→pass | 23,554 | 21,409 | -9% | 1 | 1 | 0% | 2,426 | 4,066 | +68% | 0 | 0 | — |
case-20 | pass→pass | 19,819 | 16,364 | -17% | 1 | 1 | 0% | 2,578 | 3,780 | +47% | 0 | 0 | — |
case-21 | pass→pass | 19,106 | 19,626 | +3% | 1 | 1 | 0% | 2,437 | 4,206 | +73% | 0 | 0 | — |
case-22 | pass→pass | 26,950 | 27,583 | +2% | 1 | 1 | 0% | 4,002 | 5,089 | +27% | 0 | 0 | — |
case-23 | fail→pass | 19,495 | 5,188 | -73% | 1 | 1 | 0% | 2,776 | 2,272 | -18% | 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 +22 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.