Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股午间效应/午盘策略分析。当用户说"午间效应"、"午盘"、"午休"、"lunch effect"、"午盘开盘"、"11:30"、"13:00"、"中午休市"时触发。基于 cn-stock-data 获取数据，分析午间休市效应、午盘开盘特征、午盘交易策略。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-lunch-effect/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 56% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 23% | 0% |
| case-22 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 33% | 0% |
通过 cn-stock-data skill 获取数据：
# [标的/市场] 午间效应分析报告
## 一、午间跳空
| 指标 | 数值 |
|------|------|
| 上午收盘 | 25.30 |
| 午盘开盘 | 25.35 |
## 二、午盘特征
[方向延续/反转概率]
## 三、午间信息
[午间公告/消息汇总]
## 四、午盘策略建议## [标的] 午盘速览
- 上午涨 +1.5%，午间微幅高开
- 午盘延续概率 55%
- 午间无重大公告
- 建议：观察13:15确认方向参考 references/lunch-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 | 27,119 | 32,552 | +20% | 1 | 1 | 0% | 3,747 | 4,548 | +21% | 0 | 0 | — |
case-02 | fail→fail | 22,349 | 10,154 | -55% | 1 | 1 | 0% | 2,171 | 1,186 | -45% | 0 | 0 | — |
case-03 | fail→fail | 23,344 | 36,243 | +55% | 1 | 1 | 0% | 3,374 | 5,989 | +78% | 0 | 0 | — |
case-09 | fail→fail | 15,492 | 16,559 | +7% | 1 | 1 | 0% | 1,977 | 2,437 | +23% | 0 | 0 | — |
case-04 | pass→pass | 31,335 | 44,963 | +43% | 1 | 1 | 0% | 5,725 | 7,586 | +33% | 0 | 0 | — |
case-05 | fail→pass | 31,233 | 40,017 | +28% | 1 | 1 | 0% | 2,288 | 3,575 | +56% | 0 | 0 | — |
case-06 | pass→pass | 22,677 | 44,228 | +95% | 1 | 1 | 0% | 4,238 | 6,863 | +62% | 0 | 0 | — |
case-07 | fail→pass | 19,780 | 12,853 | -35% | 1 | 1 | 0% | 2,575 | 2,617 | +2% | 0 | 0 | — |
case-08 | fail→fail | 29,042 | 11,526 | -60% | 1 | 1 | 0% | 3,527 | 1,279 | -64% | 0 | 0 | — |
case-10 | pass→pass | 29,974 | 24,217 | -19% | 1 | 1 | 0% | 2,957 | 4,316 | +46% | 0 | 0 | — |
case-11 | pass→pass | 17,734 | 14,858 | -16% | 1 | 1 | 0% | 2,472 | 2,883 | +17% | 0 | 0 | — |
case-12 | fail→pass | 21,829 | 17,680 | -19% | 1 | 1 | 0% | 2,684 | 3,299 | +23% | 0 | 0 | — |
case-13 | pass→pass | 16,835 | 19,110 | +14% | 1 | 1 | 0% | 2,247 | 3,000 | +34% | 0 | 0 | — |
case-14 | pass→pass | 24,605 | 20,339 | -17% | 1 | 1 | 0% | 3,010 | 3,735 | +24% | 0 | 0 | — |
case-15 | pass→pass | 33,748 | 18,084 | -46% | 1 | 1 | 0% | 2,704 | 3,421 | +27% | 0 | 0 | — |
case-16 | pass→pass | 16,992 | 18,176 | +7% | 1 | 1 | 0% | 2,786 | 2,728 | -2% | 0 | 0 | — |
case-17 | pass→pass | 27,042 | 23,081 | -15% | 1 | 1 | 0% | 3,290 | 4,172 | +27% | 0 | 0 | — |
case-18 | pass→pass | 19,125 | 17,032 | -11% | 1 | 1 | 0% | 2,295 | 2,954 | +29% | 0 | 0 | — |
case-19 | pass→pass | 19,114 | 19,877 | +4% | 1 | 1 | 0% | 2,692 | 3,682 | +37% | 0 | 0 | — |
case-20 | pass→pass | 14,687 | 13,608 | -7% | 1 | 1 | 0% | 2,111 | 2,664 | +26% | 0 | 0 | — |
case-21 | pass→pass | 13,427 | 10,087 | -25% | 1 | 1 | 0% | 1,646 | 2,188 | +33% | 0 | 0 | — |
case-22 | fail→pass | 20,850 | 19,097 | -8% | 1 | 1 | 0% | 3,172 | 3,703 | +17% | 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, and 20 counted toward the lift figure. The other 2 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 +18 percentage points is the difference between those two pass rates over the 20 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.