Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股多策略组合/策略配置分析。当用户说"多策略"、"multi strategy"、"策略组合"、"策略配置"、"怎么组合策略"、"策略相关性"时触发。基于 cn-stock-data 获取数据，量化分析多策略组合的协同效应。支持研报风格（formal）和快速分析风格（brief）。
.claude/skills/aifinlab-a-share-multi-strategy/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -60% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 26% | 0% |
| case-21 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -59% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -42% | 0% |
bashSCRIPTS="$SKILLS_ROOT/cn-stock-data/scripts" python "$SCRIPTS/cn_stock_data.py" kline --code [CODE] --freq daily --start [日期] python "$SCRIPTS/cn_stock_data.py" quote --code [CODE] python "$SCRIPTS/cn_stock_data.py" finance --code [CODE]
列出候选策略（动量/价值/均值回归/事件驱动等），明确各策略逻辑。
分别回测各策略的收益率序列、夏普比率、最大回撤。
计算策略间收益率相关矩阵，识别低相关/负相关策略组合。
| 维度 | formal | brief | |------|--------|-------| | 策略表现 | 各策略完整回测 | 夏普/回撤 | | 相关矩阵 | 完整相关性热力图 | 平均相关系数 | | 组合效果 | 多种配置方案对比 | 推荐配置 |
默认风格：brief。
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 | 33,381 | 22,044 | -34% | 1 | 1 | 0% | 4,942 | 3,629 | -27% | 0 | 0 | — |
case-02 | fail→fail | 19,801 | 14,453 | -27% | 1 | 1 | 0% | 3,175 | 1,368 | -57% | 0 | 0 | — |
case-03 | fail→fail | 24,940 | 25,984 | +4% | 1 | 1 | 0% | 4,468 | 4,329 | -3% | 0 | 0 | — |
case-04 | fail→pass | 19,709 | 4,861 | -75% | 1 | 1 | 0% | 3,411 | 1,357 | -60% | 0 | 0 | — |
case-05 | pass→pass | 23,797 | 21,205 | -11% | 1 | 1 | 0% | 3,174 | 3,493 | +10% | 0 | 0 | — |
case-06 | pass→pass | 22,599 | 17,177 | -24% | 1 | 1 | 0% | 3,653 | 4,228 | +16% | 0 | 0 | — |
case-07 | fail→fail | 36,110 | 91,141 | +152% | 1 | 1 | 0% | 6,172 | 8,732 | +41% | 0 | 0 | — |
case-08 | pass→pass | 22,623 | 16,853 | -26% | 1 | 1 | 0% | 2,885 | 2,986 | +4% | 0 | 0 | — |
case-09 | pass→pass | 25,186 | 28,068 | +11% | 1 | 1 | 0% | 3,549 | 3,832 | +8% | 0 | 0 | — |
case-10 | pass→pass | 21,724 | 20,542 | -5% | 1 | 1 | 0% | 3,138 | 3,594 | +15% | 0 | 0 | — |
case-11 | fail→pass | 16,327 | 13,656 | -16% | 1 | 1 | 0% | 2,224 | 2,797 | +26% | 0 | 0 | — |
case-21 | fail→pass | 10,600 | 2,400 | -77% | 1 | 1 | 0% | 1,645 | 936 | -43% | 0 | 0 | — |
case-12 | fail→pass | 21,550 | 4,709 | -78% | 1 | 1 | 0% | 2,893 | 1,192 | -59% | 0 | 0 | — |
case-13 | fail→pass | 15,303 | 5,616 | -63% | 1 | 1 | 0% | 2,642 | 1,524 | -42% | 0 | 0 | — |
case-14 | pass→pass | 18,735 | 18,501 | -1% | 1 | 1 | 0% | 2,546 | 2,806 | +10% | 0 | 0 | — |
case-15 | fail→pass | 28,594 | 18,133 | -37% | 1 | 1 | 0% | 2,808 | 3,052 | +9% | 0 | 0 | — |
case-16 | pass→pass | 18,485 | 16,110 | -13% | 1 | 1 | 0% | 2,764 | 2,765 | +0% | 0 | 0 | — |
case-17 | pass→pass | 21,418 | 18,102 | -15% | 1 | 1 | 0% | 2,830 | 2,848 | +1% | 0 | 0 | — |
case-18 | pass→fail | 28,489 | 9,783 | -66% | 1 | 1 | 0% | 5,430 | 1,051 | -81% | 0 | 0 | — |
case-19 | pass→pass | 22,487 | 23,970 | +7% | 1 | 1 | 0% | 3,165 | 3,724 | +18% | 0 | 0 | — |
case-20 | pass→pass | 27,471 | 30,500 | +11% | 1 | 1 | 0% | 5,424 | 5,349 | -1% | 0 | 0 | — |
case-22 | pass→pass | 20,996 | 13,107 | -38% | 1 | 1 | 0% | 3,029 | 2,506 | -17% | 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 +23 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.