Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股隔夜效应/跳空分析。当用户说"隔夜效应"、"overnight"、"跳空"、"高开低开"、"隔夜收益"、"缺口"、"overnight return"时触发。基于 cn-stock-data 获取数据，分析隔夜收益特征、跳空模式、隔夜风险。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-overnight-effect/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-16 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 18% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 17% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | 41% | 0% |
通过 cn-stock-data skill 获取数据：
# [标的] 隔夜效应分析报告
## 一、隔夜收益统计
| 指标 | 数值 | 历史分位 |
|------|------|----------|
## 二、跳空模式
[缺口类型判断、回补概率]
## 三、隔夜风险因素
[外盘联动、公告影响]
## 四、策略建议
[隔夜持仓建议、风控措施]## [标的] 隔夜速览
- 隔夜收益 +0.8%，高开
- 缺口类型：持续缺口，回补概率低
- 美股隔夜涨 +0.5%，正向联动
- 建议：缺口支撑有效，可持有参考 references/overnight-effect-guide.md 获取详细方法论与 A股实证研究。
python# 调用 skill result = run_skill({ "param1": "value1", "param2": "value2" })
bashpython scripts/run_skill.py --input data.json
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 32,579 | 19,232 | -41% | 1 | 1 | 0% | 4,597 | 2,573 | -44% | 0 | 0 | — |
case-02 | fail→fail | 16,131 | 21,908 | +36% | 1 | 1 | 0% | 2,049 | 2,493 | +22% | 0 | 0 | — |
case-03 | fail→fail | 30,512 | 27,112 | -11% | 1 | 1 | 0% | 4,352 | 5,151 | +18% | 0 | 0 | — |
case-04 | pass→pass | 15,502 | 21,479 | +39% | 1 | 1 | 0% | 2,470 | 2,912 | +18% | 0 | 0 | — |
case-05 | pass→pass | 16,576 | 14,813 | -11% | 1 | 1 | 0% | 2,400 | 2,814 | +17% | 0 | 0 | — |
case-06 | pass→pass | 13,051 | 16,556 | +27% | 1 | 1 | 0% | 2,076 | 2,937 | +41% | 0 | 0 | — |
case-07 | pass→pass | 8,114 | 13,079 | +61% | 1 | 1 | 0% | 1,265 | 2,020 | +60% | 0 | 0 | — |
case-08 | pass→pass | 12,233 | 15,504 | +27% | 1 | 1 | 0% | 1,940 | 2,807 | +45% | 0 | 0 | — |
case-09 | fail→fail | 41,734 | 38,309 | -8% | 1 | 1 | 0% | 5,781 | 6,928 | +20% | 0 | 0 | — |
case-10 | fail→fail | 12,195 | 17,396 | +43% | 1 | 1 | 0% | 1,707 | 2,226 | +30% | 0 | 0 | — |
case-11 | pass→pass | 28,482 | 24,898 | -13% | 1 | 1 | 0% | 3,635 | 3,900 | +7% | 0 | 0 | — |
case-12 | pass→pass | 21,752 | 45,222 | +108% | 1 | 1 | 0% | 3,205 | 3,896 | +22% | 0 | 0 | — |
case-13 | pass→pass | 31,898 | 21,919 | -31% | 1 | 1 | 0% | 3,087 | 3,796 | +23% | 0 | 0 | — |
case-14 | fail→fail | 25,796 | 16,642 | -35% | 1 | 1 | 0% | 3,860 | 4,336 | +12% | 0 | 0 | — |
case-15 | pass→pass | 12,944 | 8,852 | -32% | 1 | 1 | 0% | 1,725 | 2,027 | +18% | 0 | 0 | — |
case-16 | fail→pass | 31,349 | 21,556 | -31% | 1 | 1 | 0% | 4,305 | 4,253 | -1% | 0 | 0 | — |
case-17 | fail→pass | 23,170 | 18,900 | -18% | 1 | 1 | 0% | 3,448 | 3,707 | +8% | 0 | 0 | — |
case-18 | pass→pass | 27,384 | 20,423 | -25% | 1 | 1 | 0% | 3,337 | 3,893 | +17% | 0 | 0 | — |
case-19 | pass→pass | 26,641 | 31,161 | +17% | 1 | 1 | 0% | 3,462 | 4,746 | +37% | 0 | 0 | — |
case-20 | fail→fail | 50,381 | 48,788 | -3% | 1 | 1 | 0% | 5,557 | 6,491 | +17% | 0 | 0 | — |
case-21 | fail→fail | 30,822 | 33,788 | +10% | 1 | 1 | 0% | 6,183 | 7,685 | +24% | 0 | 0 | — |
case-22 | fail→fail | 21,491 | 38,848 | +81% | 1 | 1 | 0% | 4,421 | 6,733 | +52% | 0 | 0 | — |
case-23 | fail→fail | 24,829 | 22,618 | -9% | 1 | 1 | 0% | 3,301 | 4,040 | +22% | 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, and 22 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +9 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.