Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股离散度交易/相关性策略。当用户说"离散度"、"dispersion"、"相关性交易"、"correlation trading"、"指数vs成分股波动率"、"离散度套利"时触发。基于 cn-stock-data 获取数据，分析指数与成分股波动率离散度。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-dispersion-trade/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-18 | ✓→✗ | ▼ Worse | 5% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 16% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 31% | 0% |
通过 cn-stock-data skill 获取数据：
# 离散度交易分析报告
## 一、离散度指标
| 指标 | 数值 | 分位数 |
|------|------|--------|
## 二、相关性分析
[隐含vs已实现相关性]
## 三、交易方案
[具体期权组合]
## 四、风险控制## 离散度速览
- 隐含相关性 0.65 vs 已实现 0.55
- 相关性溢价 +10%，偏高
- 建议：做多离散度(卖指数期权+买成分股)
- 风险：市场恐慌时相关性飙升参考 references/dispersion-trade-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 | 41,592 | 37,321 | -10% | 1 | 1 | 0% | 5,381 | 6,437 | +20% | 0 | 0 | — |
case-02 | fail→fail | 18,802 | 13,479 | -28% | 1 | 1 | 0% | 2,579 | 2,582 | +0% | 0 | 0 | — |
case-03 | fail→fail | 36,009 | 37,529 | +4% | 1 | 1 | 0% | 5,065 | 6,686 | +32% | 0 | 0 | — |
case-04 | pass→pass | 18,768 | 15,331 | -18% | 1 | 1 | 0% | 2,582 | 2,987 | +16% | 0 | 0 | — |
case-05 | pass→pass | 20,625 | 18,402 | -11% | 1 | 1 | 0% | 2,744 | 3,604 | +31% | 0 | 0 | — |
case-06 | pass→pass | 27,518 | 25,428 | -8% | 1 | 1 | 0% | 3,755 | 4,170 | +11% | 0 | 0 | — |
case-07 | pass→pass | 25,289 | 21,886 | -13% | 1 | 1 | 0% | 4,055 | 4,276 | +5% | 0 | 0 | — |
case-08 | fail→pass | 23,383 | 20,276 | -13% | 1 | 1 | 0% | 3,304 | 4,210 | +27% | 0 | 0 | — |
case-09 | fail→fail | 43,390 | 49,166 | +13% | 1 | 1 | 0% | 6,634 | 7,521 | +13% | 0 | 0 | — |
case-10 | pass→pass | 24,746 | 27,236 | +10% | 1 | 1 | 0% | 3,556 | 4,676 | +31% | 0 | 0 | — |
case-11 | pass→pass | 22,306 | 27,762 | +24% | 1 | 1 | 0% | 3,055 | 5,027 | +65% | 0 | 0 | — |
case-12 | pass→pass | 22,892 | 22,719 | -1% | 1 | 1 | 0% | 3,327 | 3,592 | +8% | 0 | 0 | — |
case-13 | pass→pass | 15,335 | 22,823 | +49% | 1 | 1 | 0% | 2,068 | 3,252 | +57% | 0 | 0 | — |
case-14 | fail→fail | 18,387 | 16,486 | -10% | 1 | 1 | 0% | 2,486 | 2,783 | +12% | 0 | 0 | — |
case-15 | fail→fail | 17,188 | 14,461 | -16% | 1 | 1 | 0% | 2,432 | 2,813 | +16% | 0 | 0 | — |
case-16 | pass→pass | 22,221 | 15,597 | -30% | 1 | 1 | 0% | 2,801 | 3,042 | +9% | 0 | 0 | — |
case-17 | fail→pass | 14,887 | 17,113 | +15% | 1 | 1 | 0% | 2,995 | 3,196 | +7% | 0 | 0 | — |
case-18 | pass→fail | 30,112 | 21,197 | -30% | 1 | 1 | 0% | 4,787 | 5,021 | +5% | 0 | 0 | — |
case-19 | pass→pass | 25,710 | 23,462 | -9% | 1 | 1 | 0% | 3,568 | 4,207 | +18% | 0 | 0 | — |
case-20 | pass→pass | 21,026 | 25,461 | +21% | 1 | 1 | 0% | 3,799 | 3,999 | +5% | 0 | 0 | — |
case-21 | fail→fail | 15,909 | 11,729 | -26% | 1 | 1 | 0% | 2,051 | 2,477 | +21% | 0 | 0 | — |
case-22 | fail→fail | 23,899 | 12,988 | -46% | 1 | 1 | 0% | 2,385 | 2,559 | +7% | 0 | 0 | — |
case-23 | pass→pass | 15,534 | 27,878 | +79% | 1 | 1 | 0% | 2,975 | 3,375 | +13% | 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. 1 case got worse with the skill loaded, and it is 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.