Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股波动率套利/Vega交易策略。当用户说"波动率套利"、"vol arb"、"Vega交易"、"做多波动率"、"做空波动率"、"波动率交易"、"straddle"、"strangle"时触发。基于 cn-stock-data 获取数据，设计波动率套利策略。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-volatility-arb/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -8% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -52% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 15% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 41% | 0% |
通过 cn-stock-data skill 获取数据：
# 波动率套利策略报告
## 一、波动率评估
| 指标 | 数值 | 信号 |
|------|------|------|
## 二、策略方案
[具体期权组合、Greeks]
## 三、对冲计划
[Delta对冲方案、频率]
## 四、风险控制
[止损、最大亏损]## 波动率套利速览
- IV 25% vs HV 20%，IV偏高
- 建议：卖出Straddle，收取权利金
- Delta对冲：日度调整
- 最大亏损控制在权利金的150%参考 references/volatility-arb-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 | 33,474 | 14,160 | -58% | 1 | 1 | 0% | 5,368 | 1,638 | -69% | 0 | 0 | — |
case-02 | fail→fail | 13,073 | 9,671 | -26% | 1 | 1 | 0% | 1,936 | 2,199 | +14% | 0 | 0 | — |
case-03 | fail→fail | 35,293 | 31,111 | -12% | 1 | 1 | 0% | 4,542 | 5,817 | +28% | 0 | 0 | — |
case-04 | pass→pass | 20,369 | 21,920 | +8% | 1 | 1 | 0% | 4,671 | 5,385 | +15% | 0 | 0 | — |
case-05 | pass→pass | 31,560 | 39,027 | +24% | 1 | 1 | 0% | 6,272 | 8,870 | +41% | 0 | 0 | — |
case-06 | pass→pass | 17,922 | 19,422 | +8% | 1 | 1 | 0% | 3,503 | 4,594 | +31% | 0 | 0 | — |
case-07 | fail→pass | 12,695 | 7,182 | -43% | 1 | 1 | 0% | 1,914 | 1,766 | -8% | 0 | 0 | — |
case-08 | fail→pass | 16,036 | 8,959 | -44% | 1 | 1 | 0% | 2,072 | 2,018 | -3% | 0 | 0 | — |
case-09 | pass→pass | 14,427 | 13,950 | -3% | 1 | 1 | 0% | 2,317 | 3,034 | +31% | 0 | 0 | — |
case-10 | pass→pass | 25,327 | 23,765 | -6% | 1 | 1 | 0% | 3,694 | 4,274 | +16% | 0 | 0 | — |
case-11 | pass→pass | 15,100 | 13,335 | -12% | 1 | 1 | 0% | 2,220 | 2,775 | +25% | 0 | 0 | — |
case-12 | pass→pass | 18,823 | 21,457 | +14% | 1 | 1 | 0% | 3,186 | 4,325 | +36% | 0 | 0 | — |
case-13 | pass→pass | 18,964 | 15,352 | -19% | 1 | 1 | 0% | 2,625 | 2,627 | +0% | 0 | 0 | — |
case-14 | pass→pass | 17,699 | 24,402 | +38% | 1 | 1 | 0% | 2,869 | 4,284 | +49% | 0 | 0 | — |
case-15 | pass→pass | 16,639 | 15,366 | -8% | 1 | 1 | 0% | 2,550 | 3,013 | +18% | 0 | 0 | — |
case-16 | fail→fail | 23,503 | 24,749 | +5% | 1 | 1 | 0% | 3,302 | 4,535 | +37% | 0 | 0 | — |
case-17 | fail→pass | 13,598 | 2,249 | -83% | 1 | 1 | 0% | 2,088 | 993 | -52% | 0 | 0 | — |
case-18 | pass→pass | 14,664 | 10,689 | -27% | 1 | 1 | 0% | 1,937 | 2,251 | +16% | 0 | 0 | — |
case-19 | pass→pass | 16,910 | 5,671 | -66% | 1 | 1 | 0% | 2,440 | 1,502 | -38% | 0 | 0 | — |
case-20 | pass→pass | 22,080 | 19,648 | -11% | 1 | 1 | 0% | 2,835 | 3,659 | +29% | 0 | 0 | — |
case-21 | pass→pass | 25,241 | 22,096 | -12% | 1 | 1 | 0% | 3,162 | 3,788 | +20% | 0 | 0 | — |
case-22 | fail→fail | 15,541 | 9,230 | -41% | 1 | 1 | 0% | 2,309 | 2,218 | -4% | 0 | 0 | — |
case-23 | pass→pass | 20,551 | 12,000 | -42% | 1 | 1 | 0% | 2,574 | 2,425 | -6% | 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 +13 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.