Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股期现套利/基差交易策略。当用户说"期现套利"、"基差交易"、"index arb"、"期货基差"、"正向套利"、"反向套利"、"基差收敛"时触发。基于 cn-stock-data 获取数据，分析股指期货基差与套利机会。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-index-arb/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 56% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | 12% | 0% |
通过 cn-stock-data skill 获取数据：
# 期现套利分析报告
## 一、基差状态
| 合约 | 期货 | 现货 | 基差 | 年化 |
|------|------|------|------|------|
## 二、套利机会
[策略类型、预期收益]
## 三、执行方案
[现货组合、对冲比例]
## 四、风险控制## 期现套利速览
- IF当月基差 -15点，年化贴水 -3.2%
- IC当月基差 -45点，年化贴水 -8.5%
- IC贴水较深，正向套利空间有限
- 建议：关注IC跨期价差收敛机会参考 references/index-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 | 31,158 | 15,790 | -49% | 1 | 1 | 0% | 4,960 | 1,199 | -76% | 0 | 0 | — |
case-06 | pass→pass | 25,397 | 22,063 | -13% | 1 | 1 | 0% | 3,742 | 4,198 | +12% | 0 | 0 | — |
case-02 | fail→fail | 17,640 | 21,296 | +21% | 1 | 1 | 0% | 2,654 | 1,322 | -50% | 0 | 0 | — |
case-03 | fail→fail | 28,875 | 51,975 | +80% | 1 | 1 | 0% | 4,644 | 8,934 | +92% | 0 | 0 | — |
case-04 | fail→pass | 21,961 | 18,140 | -17% | 1 | 1 | 0% | 3,482 | 3,504 | +1% | 0 | 0 | — |
case-05 | fail→pass | 10,056 | 12,662 | +26% | 1 | 1 | 0% | 1,835 | 2,864 | +56% | 0 | 0 | — |
case-07 | pass→pass | 9,508 | 15,248 | +60% | 1 | 1 | 0% | 1,634 | 3,222 | +97% | 0 | 0 | — |
case-08 | pass→pass | 16,617 | 15,430 | -7% | 1 | 1 | 0% | 2,281 | 3,012 | +32% | 0 | 0 | — |
case-09 | pass→pass | 20,328 | 19,102 | -6% | 1 | 1 | 0% | 3,052 | 3,462 | +13% | 0 | 0 | — |
case-10 | fail→fail | 16,379 | 15,986 | -2% | 1 | 1 | 0% | 2,412 | 3,044 | +26% | 0 | 0 | — |
case-11 | fail→pass | 16,906 | 12,596 | -25% | 1 | 1 | 0% | 2,774 | 2,353 | -15% | 0 | 0 | — |
case-12 | fail→fail | 20,099 | 18,086 | -10% | 1 | 1 | 0% | 2,783 | 3,648 | +31% | 0 | 0 | — |
case-13 | pass→pass | 9,850 | 13,190 | +34% | 1 | 1 | 0% | 1,612 | 2,557 | +59% | 0 | 0 | — |
case-14 | pass→pass | 17,630 | 16,901 | -4% | 1 | 1 | 0% | 2,723 | 3,484 | +28% | 0 | 0 | — |
case-15 | fail→pass | 10,409 | 11,497 | +10% | 1 | 1 | 0% | 1,729 | 2,618 | +51% | 0 | 0 | — |
case-16 | pass→pass | 31,159 | 39,209 | +26% | 1 | 1 | 0% | 4,556 | 6,114 | +34% | 0 | 0 | — |
case-17 | pass→pass | 25,216 | 33,724 | +34% | 1 | 1 | 0% | 4,113 | 5,562 | +35% | 0 | 0 | — |
case-18 | pass→pass | 32,075 | 40,873 | +27% | 1 | 1 | 0% | 6,289 | 8,922 | +42% | 0 | 0 | — |
case-19 | pass→pass | 12,590 | 18,489 | +47% | 1 | 1 | 0% | 1,698 | 3,166 | +86% | 0 | 0 | — |
case-20 | pass→pass | 5,483 | 6,358 | +16% | 1 | 1 | 0% | 1,024 | 1,542 | +51% | 0 | 0 | — |
case-21 | fail→fail | 15,363 | 12,424 | -19% | 1 | 1 | 0% | 2,246 | 2,631 | +17% | 0 | 0 | — |
case-22 | fail→fail | 26,281 | 25,439 | -3% | 1 | 1 | 0% | 4,398 | 4,612 | +5% | 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. 2 cases got worse with the skill loaded, and they are 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/28/2026 | +5% |
Other measured skills in the registry, with their headline benchmark lift.