Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股TWAP时间加权/匀速执行策略。当用户说"TWAP"、"时间加权"、"匀速下单"、"TWAP算法"、"均匀执行"、"时间切片"时触发。基于 cn-stock-data 获取数据，设计TWAP执行方案、分析执行偏差、优化时间切片。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-twap-strategy/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-04 | ✓→✗ | ▼ Worse | -69% | 0% |
| case-05 | ✓→✗ | ▼ Worse | -62% | 0% |
| case-06 | ✓→✗ | ▼ Worse | -69% | 0% |
通过 cn-stock-data skill 获取数据：
# [标的] TWAP执行方案
## 一、方案设计
| 参数 | 设置 |
|------|------|
| 总量 | 10万股 |
| 时间片 | 5分钟 |
## 二、流动性评估
[各时段流动性、参与率]
## 三、执行偏差预估
[预期偏差、成本估算]
## 四、增强建议
[自适应调整方案]## [标的] TWAP方案速览
- 10万股拆分48个时间片(5分钟)
- 每片约2,083股，参与率约8%
- 预期执行偏差 ±0.05%
- 建议：使用Volume-weighted TWAP参考 references/twap-strategy-guide.md 获取详细方法论与 A股实证研究。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 44,627 | 11,629 | -74% | 1 | 1 | 0% | 6,300 | 1,184 | -81% | 0 | 0 | — |
case-02 | fail→fail | 14,784 | 18,230 | +23% | 1 | 1 | 0% | 2,474 | 2,959 | +20% | 0 | 0 | — |
case-03 | fail→fail | 29,876 | 9,290 | -69% | 1 | 1 | 0% | 4,861 | 1,141 | -77% | 0 | 0 | — |
case-04 | pass→fail | 22,242 | 7,806 | -65% | 1 | 1 | 0% | 3,501 | 1,083 | -69% | 0 | 0 | — |
case-05 | pass→fail | 45,820 | 10,724 | -77% | 1 | 1 | 0% | 3,245 | 1,249 | -62% | 0 | 0 | — |
case-06 | pass→fail | 20,992 | 7,350 | -65% | 1 | 1 | 0% | 3,287 | 1,024 | -69% | 0 | 0 | — |
case-11 | pass→pass | 26,769 | 15,354 | -43% | 1 | 1 | 0% | 3,638 | 3,131 | -14% | 0 | 0 | — |
case-07 | pass→pass | 26,198 | 35,049 | +34% | 1 | 1 | 0% | 3,850 | 4,631 | +20% | 0 | 0 | — |
case-08 | pass→pass | 24,055 | 24,682 | +3% | 1 | 1 | 0% | 3,023 | 4,071 | +35% | 0 | 0 | — |
case-09 | fail→pass | 17,970 | 15,174 | -16% | 1 | 1 | 0% | 3,024 | 3,122 | +3% | 0 | 0 | — |
case-10 | pass→pass | 16,857 | 26,961 | +60% | 1 | 1 | 0% | 2,576 | 4,021 | +56% | 0 | 0 | — |
case-12 | pass→pass | 26,735 | 21,361 | -20% | 1 | 1 | 0% | 3,371 | 3,794 | +13% | 0 | 0 | — |
case-13 | pass→pass | 21,180 | 24,952 | +18% | 1 | 1 | 0% | 3,198 | 4,199 | +31% | 0 | 0 | — |
case-14 | fail→pass | 18,355 | 15,480 | -16% | 1 | 1 | 0% | 3,059 | 3,249 | +6% | 0 | 0 | — |
case-15 | pass→pass | 21,040 | 17,866 | -15% | 1 | 1 | 0% | 3,066 | 3,425 | +12% | 0 | 0 | — |
case-16 | pass→pass | 21,763 | 22,338 | +3% | 1 | 1 | 0% | 3,048 | 4,036 | +32% | 0 | 0 | — |
case-17 | pass→pass | 19,986 | 19,890 | -0% | 1 | 1 | 0% | 2,669 | 3,306 | +24% | 0 | 0 | — |
case-18 | pass→pass | 24,304 | 22,637 | -7% | 1 | 1 | 0% | 3,371 | 4,029 | +20% | 0 | 0 | — |
case-19 | pass→pass | 14,056 | 16,119 | +15% | 1 | 1 | 0% | 2,544 | 3,450 | +36% | 0 | 0 | — |
case-20 | pass→pass | 33,572 | 41,393 | +23% | 1 | 1 | 0% | 5,834 | 7,246 | +24% | 0 | 0 | — |
case-21 | pass→pass | 26,278 | 31,282 | +19% | 1 | 1 | 0% | 3,752 | 4,584 | +22% | 0 | 0 | — |
case-22 | pass→pass | 22,391 | 22,024 | -2% | 1 | 1 | 0% | 4,167 | 3,732 | -10% | 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 17 counted toward the lift figure. The other 5 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 -5 percentage points is the difference between those two pass rates over the 17 comparable cases. 5 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.
Other measured skills in the registry, with their headline benchmark lift.