Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股VWAP算法/成交量加权策略。当用户说"VWAP"、"成交量加权"、"VWAP策略"、"VWAP突破"、"VWAP算法"、"量价加权"时触发。基于 cn-stock-data 获取数据，计算VWAP、分析偏离度、构建VWAP交易策略。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-vwap-strategy/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | 31% | 0% |
通过 cn-stock-data skill 获取数据：
# [标的] VWAP策略分析报告
## 一、VWAP计算
| 类型 | VWAP | 当前价 | 偏离度 |
|------|------|--------|--------|
## 二、偏离度分析
[历史分布、回归概率]
## 三、交易信号
[突破/支撑信号状态]
## 四、执行建议
[成交量预测、拆单方案]## [标的] VWAP速览
- 日内VWAP 25.32，当前价 25.48 (+0.63%)
- 价格在VWAP上方，偏多
- 偏离度P72，尚未到极值
- 建议：VWAP附近可加仓参考 references/vwap-strategy-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 | 26,276 | 24,815 | -6% | 1 | 1 | 0% | 4,335 | 1,009 | -77% | 0 | 0 | — |
case-02 | pass→pass | 13,244 | 23,035 | +74% | 1 | 1 | 0% | 2,174 | 2,846 | +31% | 0 | 0 | — |
case-03 | fail→pass | 23,958 | 28,017 | +17% | 1 | 1 | 0% | 4,177 | 5,431 | +30% | 0 | 0 | — |
case-04 | pass→pass | 20,674 | 25,940 | +25% | 1 | 1 | 0% | 2,972 | 4,236 | +43% | 0 | 0 | — |
case-22 | pass→pass | 22,422 | 23,841 | +6% | 1 | 1 | 0% | 3,945 | 5,074 | +29% | 0 | 0 | — |
case-05 | pass→pass | 21,650 | 18,988 | -12% | 1 | 1 | 0% | 3,441 | 3,786 | +10% | 0 | 0 | — |
case-06 | fail→fail | 25,669 | 28,435 | +11% | 1 | 1 | 0% | 3,622 | 4,647 | +28% | 0 | 0 | — |
case-07 | pass→pass | 27,587 | 24,884 | -10% | 1 | 1 | 0% | 3,841 | 4,345 | +13% | 0 | 0 | — |
case-08 | pass→pass | 12,559 | 13,273 | +6% | 1 | 1 | 0% | 2,225 | 2,950 | +33% | 0 | 0 | — |
case-09 | fail→pass | 22,931 | 24,476 | +7% | 1 | 1 | 0% | 3,517 | 3,840 | +9% | 0 | 0 | — |
case-10 | fail→pass | 22,155 | 19,171 | -13% | 1 | 1 | 0% | 3,307 | 3,467 | +5% | 0 | 0 | — |
case-11 | pass→pass | 18,179 | 16,622 | -9% | 1 | 1 | 0% | 2,570 | 2,941 | +14% | 0 | 0 | — |
case-12 | pass→pass | 18,559 | 14,851 | -20% | 1 | 1 | 0% | 2,932 | 3,014 | +3% | 0 | 0 | — |
case-13 | pass→pass | 18,665 | 20,898 | +12% | 1 | 1 | 0% | 3,056 | 3,956 | +29% | 0 | 0 | — |
case-14 | pass→pass | 25,147 | 20,998 | -16% | 1 | 1 | 0% | 3,628 | 3,914 | +8% | 0 | 0 | — |
case-15 | pass→pass | 19,268 | 17,218 | -11% | 1 | 1 | 0% | 2,813 | 3,110 | +11% | 0 | 0 | — |
case-16 | pass→pass | 10,889 | 15,717 | +44% | 1 | 1 | 0% | 1,685 | 2,144 | +27% | 0 | 0 | — |
case-17 | fail→pass | 35,729 | 26,131 | -27% | 1 | 1 | 0% | 4,787 | 5,051 | +6% | 0 | 0 | — |
case-23 | pass→pass | 26,974 | 51,867 | +92% | 1 | 1 | 0% | 3,563 | 7,129 | +100% | 0 | 0 | — |
case-18 | pass→pass | 21,920 | 25,036 | +14% | 1 | 1 | 0% | 3,357 | 3,899 | +16% | 0 | 0 | — |
case-19 | pass→pass | 13,651 | 13,322 | -2% | 1 | 1 | 0% | 2,210 | 2,400 | +9% | 0 | 0 | — |
case-20 | fail→fail | 19,881 | 17,546 | -12% | 1 | 1 | 0% | 3,194 | 3,761 | +18% | 0 | 0 | — |
case-21 | pass→pass | 35,607 | 22,308 | -37% | 1 | 1 | 0% | 5,169 | 5,323 | +3% | 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 +17 percentage points is the difference between those two pass rates over the 22 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.