Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股AutoML策略/自动化量化建模。当用户说"AutoML"、"自动建模"、"自动机器学习"、"auto ML"、"自动化策略"、"一键建模"时触发。基于 cn-stock-data 获取数据，使用AutoML自动构建量化策略。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-automl-strategy/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 58% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 42% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 28% | 0% |
| case-22 | ✗→✓ | ▲ Improved | 18% | 0% |
通过 cn-stock-data skill 获取数据：
# AutoML量化策略报告
## 一、搜索结果
| 模型 | IC | Sharpe | 排名 |
|------|-----|--------|------|
## 二、最优模型
[模型配置、特征重要性]
## 三、策略表现
[回测结果、风险指标]
## 四、部署建议
[更新频率、监控指标]## AutoML速览
- 搜索50个模型，最优：LightGBM
- IC=0.045, Sharpe 2.0
- Top特征：动量+质量+资金流
- 建议：月度重训练，监控IC衰减参考 references/automl-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 | 35,948 | 38,630 | +7% | 1 | 1 | 0% | 6,295 | 6,453 | +3% | 0 | 0 | — |
case-02 | fail→pass | 12,728 | 17,013 | +34% | 1 | 1 | 0% | 2,090 | 3,308 | +58% | 0 | 0 | — |
case-03 | fail→fail | 35,824 | 33,427 | -7% | 1 | 1 | 0% | 5,662 | 5,607 | -1% | 0 | 0 | — |
case-04 | pass→pass | 51,245 | 30,680 | -40% | 1 | 1 | 0% | 4,893 | 6,108 | +25% | 0 | 0 | — |
case-05 | pass→pass | 37,613 | 27,560 | -27% | 1 | 1 | 0% | 6,250 | 5,101 | -18% | 0 | 0 | — |
case-06 | pass→pass | 24,291 | 25,764 | +6% | 1 | 1 | 0% | 3,241 | 4,365 | +35% | 0 | 0 | — |
case-07 | fail→pass | 23,641 | 26,110 | +10% | 1 | 1 | 0% | 3,547 | 3,800 | +7% | 0 | 0 | — |
case-08 | pass→pass | 20,133 | 28,393 | +41% | 1 | 1 | 0% | 3,128 | 4,454 | +42% | 0 | 0 | — |
case-09 | pass→pass | 21,486 | 20,975 | -2% | 1 | 1 | 0% | 2,953 | 3,758 | +27% | 0 | 0 | — |
case-10 | fail→pass | 19,198 | 20,618 | +7% | 1 | 1 | 0% | 2,689 | 3,814 | +42% | 0 | 0 | — |
case-11 | fail→fail | 21,859 | 22,735 | +4% | 1 | 1 | 0% | 2,796 | 3,696 | +32% | 0 | 0 | — |
case-12 | pass→pass | 21,512 | 22,984 | +7% | 1 | 1 | 0% | 3,270 | 4,125 | +26% | 0 | 0 | — |
case-13 | fail→fail | 16,062 | 16,724 | +4% | 1 | 1 | 0% | 2,598 | 3,156 | +21% | 0 | 0 | — |
case-14 | fail→fail | 22,606 | 19,688 | -13% | 1 | 1 | 0% | 3,175 | 3,804 | +20% | 0 | 0 | — |
case-15 | fail→fail | 13,035 | 8,530 | -35% | 1 | 1 | 0% | 1,955 | 1,950 | -0% | 0 | 0 | — |
case-16 | fail→pass | 18,314 | 24,425 | +33% | 1 | 1 | 0% | 2,868 | 3,671 | +28% | 0 | 0 | — |
case-17 | pass→fail | 21,148 | 22,252 | +5% | 1 | 1 | 0% | 3,266 | 3,617 | +11% | 0 | 0 | — |
case-18 | pass→pass | 34,264 | 20,047 | -41% | 1 | 1 | 0% | 3,209 | 3,569 | +11% | 0 | 0 | — |
case-19 | pass→pass | 19,015 | 18,577 | -2% | 1 | 1 | 0% | 3,099 | 3,333 | +8% | 0 | 0 | — |
case-20 | pass→pass | 22,108 | 30,582 | +38% | 1 | 1 | 0% | 3,245 | 4,740 | +46% | 0 | 0 | — |
case-21 | fail→fail | 18,230 | 20,047 | +10% | 1 | 1 | 0% | 2,891 | 3,736 | +29% | 0 | 0 | — |
case-22 | fail→pass | 19,376 | 20,412 | +5% | 1 | 1 | 0% | 2,839 | 3,354 | +18% | 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. The headline lift of +18 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/28/2026 | +22% |
Other measured skills in the registry, with their headline benchmark lift.