Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股波动率曲面/隐含波动率分析。当用户说"波动率曲面"、"隐含波动率"、"IV曲面"、"volatility surface"、"波动率微笑"、"波动率偏斜"、"IV skew"时触发。基于 cn-stock-data 获取数据，构建与分析波动率曲面。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-volatility-surface/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 103% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-21 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 22% | 0% |
通过 cn-stock-data skill 获取数据：
# [标的] 波动率曲面分析报告
## 一、IV概览
| 指标 | 数值 | 分位数 |
|------|------|--------|
| ATM IV | 22.5% | P55 |
## 二、曲面形态
[偏斜度、凸度、期限结构]
## 三、异常信号
[IV vs RV、偏斜异常]
## 四、交易建议
[基于曲面的期权策略]## [标的] IV速览
- ATM IV 22.5% (P55)，中等水平
- Skew -3.2%，正常偏斜
- VRP +2.1%，卖权有溢价
- 期限结构正常，无倒挂参考 references/volatility-surface-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 | 36,686 | 17,471 | -52% | 1 | 1 | 0% | 5,743 | 1,291 | -78% | 0 | 0 | — |
case-02 | fail→fail | 13,877 | 16,176 | +17% | 1 | 1 | 0% | 2,203 | 2,129 | -3% | 0 | 0 | — |
case-03 | fail→fail | 31,237 | 23,115 | -26% | 1 | 1 | 0% | 4,460 | 4,268 | -4% | 0 | 0 | — |
case-04 | pass→pass | 23,826 | 24,742 | +4% | 1 | 1 | 0% | 3,835 | 4,665 | +22% | 0 | 0 | — |
case-05 | fail→pass | 43,257 | 20,905 | -52% | 1 | 1 | 0% | 1,989 | 4,041 | +103% | 0 | 0 | — |
case-06 | pass→pass | 19,753 | 22,241 | +13% | 1 | 1 | 0% | 3,232 | 4,561 | +41% | 0 | 0 | — |
case-07 | pass→pass | 15,836 | 16,567 | +5% | 1 | 1 | 0% | 2,066 | 3,063 | +48% | 0 | 0 | — |
case-08 | pass→pass | 16,877 | 20,456 | +21% | 1 | 1 | 0% | 2,847 | 3,664 | +29% | 0 | 0 | — |
case-09 | pass→pass | 18,831 | 20,109 | +7% | 1 | 1 | 0% | 3,183 | 3,793 | +19% | 0 | 0 | — |
case-10 | fail→fail | 21,237 | 4,710 | -78% | 1 | 1 | 0% | 2,940 | 1,517 | -48% | 0 | 0 | — |
case-11 | fail→fail | 15,854 | 6,523 | -59% | 1 | 1 | 0% | 2,355 | 1,825 | -23% | 0 | 0 | — |
case-12 | fail→pass | 21,093 | 4,560 | -78% | 1 | 1 | 0% | 2,731 | 1,418 | -48% | 0 | 0 | — |
case-13 | pass→pass | 22,629 | 18,237 | -19% | 1 | 1 | 0% | 3,241 | 3,207 | -1% | 0 | 0 | — |
case-14 | pass→pass | 31,345 | 25,069 | -20% | 1 | 1 | 0% | 5,164 | 5,527 | +7% | 0 | 0 | — |
case-15 | pass→pass | 14,917 | 19,082 | +28% | 1 | 1 | 0% | 2,656 | 3,509 | +32% | 0 | 0 | — |
case-16 | pass→pass | 21,637 | 30,471 | +41% | 1 | 1 | 0% | 3,692 | 5,377 | +46% | 0 | 0 | — |
case-17 | fail→fail | 16,343 | 7,182 | -56% | 1 | 1 | 0% | 2,457 | 1,918 | -22% | 0 | 0 | — |
case-18 | pass→pass | 13,650 | 4,463 | -67% | 1 | 1 | 0% | 1,782 | 1,492 | -16% | 0 | 0 | — |
case-19 | fail→pass | 11,045 | 3,766 | -66% | 1 | 1 | 0% | 1,812 | 1,413 | -22% | 0 | 0 | — |
case-20 | pass→pass | 14,381 | 3,893 | -73% | 1 | 1 | 0% | 2,291 | 1,316 | -43% | 0 | 0 | — |
case-21 | fail→pass | 11,181 | 3,337 | -70% | 1 | 1 | 0% | 1,456 | 1,189 | -18% | 0 | 0 | — |
case-22 | pass→pass | 12,574 | 3,202 | -75% | 1 | 1 | 0% | 1,887 | 1,017 | -46% | 0 | 0 | — |
case-23 | pass→pass | 12,384 | 7,530 | -39% | 1 | 1 | 0% | 1,922 | 1,635 | -15% | 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 21 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 +17 percentage points is the difference between those two pass rates over the 21 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.