Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股算法执行/智能拆单策略。当用户说"算法执行"、"智能拆单"、"execution algo"、"算法交易"、"拆单策略"、"大单拆分"时触发。基于 cn-stock-data 获取数据，设计算法执行方案。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-execution-algo/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-18 | ✓→✗ | ▼ Worse | -60% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 4% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 63% | 0% |
通过 cn-stock-data skill 获取数据：
# [标的] 算法执行方案
## 一、方案配置
| 参数 | 设置 |
|------|------|
| 算法 | VWAP |
| 总量 | 50万股 |
## 二、执行计划
[分时段拆单明细]
## 三、风控参数
[价格保护、参与率]
## 四、预期成本## [标的] 执行方案速览
- VWAP算法，50万股，执行窗口9:30-14:30
- 参与率上限15%，预期冲击0.08%
- 价格保护：±1%
- 预期总成本：0.15%参考 references/execution-algo-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-05 | pass→pass | 19,590 | 18,599 | -5% | 1 | 1 | 0% | 2,975 | 3,094 | +4% | 0 | 0 | — |
case-01 | fail→fail | 36,347 | 41,994 | +16% | 1 | 1 | 0% | 6,010 | 7,927 | +32% | 0 | 0 | — |
case-02 | fail→fail | 16,137 | 12,089 | -25% | 1 | 1 | 0% | 2,564 | 1,179 | -54% | 0 | 0 | — |
case-03 | fail→fail | 28,878 | 13,946 | -52% | 1 | 1 | 0% | 3,732 | 1,918 | -49% | 0 | 0 | — |
case-04 | pass→pass | 20,873 | 23,427 | +12% | 1 | 1 | 0% | 2,800 | 4,562 | +63% | 0 | 0 | — |
case-06 | pass→pass | 15,230 | 14,187 | -7% | 1 | 1 | 0% | 2,241 | 2,748 | +23% | 0 | 0 | — |
case-07 | pass→pass | 9,811 | 10,402 | +6% | 1 | 1 | 0% | 1,437 | 2,143 | +49% | 0 | 0 | — |
case-08 | pass→pass | 16,338 | 13,648 | -16% | 1 | 1 | 0% | 2,116 | 2,441 | +15% | 0 | 0 | — |
case-09 | pass→pass | 19,803 | 7,051 | -64% | 1 | 1 | 0% | 2,592 | 1,700 | -34% | 0 | 0 | — |
case-10 | pass→pass | 15,531 | 16,054 | +3% | 1 | 1 | 0% | 2,220 | 2,976 | +34% | 0 | 0 | — |
case-11 | pass→pass | 22,785 | 22,823 | +0% | 1 | 1 | 0% | 2,967 | 3,926 | +32% | 0 | 0 | — |
case-12 | pass→pass | 19,927 | 20,876 | +5% | 1 | 1 | 0% | 2,808 | 3,783 | +35% | 0 | 0 | — |
case-13 | fail→pass | 11,444 | 5,963 | -48% | 1 | 1 | 0% | 1,644 | 1,236 | -25% | 0 | 0 | — |
case-14 | pass→pass | 17,827 | 8,867 | -50% | 1 | 1 | 0% | 2,527 | 1,992 | -21% | 0 | 0 | — |
case-15 | pass→pass | 9,680 | 5,317 | -45% | 1 | 1 | 0% | 1,542 | 1,220 | -21% | 0 | 0 | — |
case-16 | pass→pass | 14,201 | 6,048 | -57% | 1 | 1 | 0% | 1,939 | 1,503 | -22% | 0 | 0 | — |
case-17 | pass→pass | 11,887 | 2,818 | -76% | 1 | 1 | 0% | 1,852 | 1,078 | -42% | 0 | 0 | — |
case-18 | pass→fail | 16,855 | 10,621 | -37% | 1 | 1 | 0% | 2,642 | 1,055 | -60% | 0 | 0 | — |
case-19 | fail→pass | 15,746 | 6,609 | -58% | 1 | 1 | 0% | 2,223 | 1,684 | -24% | 0 | 0 | — |
case-20 | pass→pass | 59,239 | 40,479 | -32% | 1 | 1 | 0% | 5,771 | 8,868 | +54% | 0 | 0 | — |
case-21 | pass→pass | 25,753 | 29,461 | +14% | 1 | 1 | 0% | 3,984 | 4,789 | +20% | 0 | 0 | — |
case-22 | pass→pass | 23,032 | 19,683 | -15% | 1 | 1 | 0% | 3,326 | 4,094 | +23% | 0 | 0 | — |
case-23 | pass→pass | 9,318 | 4,205 | -55% | 1 | 1 | 0% | 1,265 | 1,156 | -9% | 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 20 counted toward the lift figure. The other 3 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 +4 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.
Other measured skills in the registry, with their headline benchmark lift.