Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股领先滞后关系/板块传导分析。当用户说"领先滞后"、"lead lag"、"谁先涨"、"传导"、"板块传导"、"龙头带动"时触发。量化分析股票/板块间的领先滞后关系。支持formal和brief风格。
.claude/skills/aifinlab-a-share-lead-lag/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -57% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -29% | 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]
计算标的A在t期收益率与标的B在t+k期收益率的相关性（k=-5到+5）
检验A是否Granger因果引起B（或反向）
构建多标的间的领先-滞后关系网络
| 维度 | formal | brief | |------|--------|-------| | 相关矩阵 | 多lag完整矩阵 | 最强领先关系 | | 因果检验 | Granger检验结果 | 领先/滞后天数 | | 关系图谱 | 完整网络图 | Top 3领先者 | 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-06 | pass→pass | 15,404 | 13,131 | -15% | 1 | 1 | 0% | 2,392 | 2,805 | +17% | 0 | 0 | — |
case-01 | fail→fail | 26,849 | 16,417 | -39% | 1 | 1 | 0% | 3,535 | 1,378 | -61% | 0 | 0 | — |
case-02 | fail→fail | 21,753 | 14,683 | -33% | 1 | 1 | 0% | 2,559 | 1,199 | -53% | 0 | 0 | — |
case-03 | fail→fail | 23,787 | 16,012 | -33% | 1 | 1 | 0% | 2,981 | 1,257 | -58% | 0 | 0 | — |
case-04 | fail→pass | 18,201 | 3,214 | -82% | 1 | 1 | 0% | 2,602 | 1,109 | -57% | 0 | 0 | — |
case-05 | pass→pass | 15,989 | 4,341 | -73% | 1 | 1 | 0% | 2,551 | 1,176 | -54% | 0 | 0 | — |
case-11 | fail→fail | 29,354 | 24,338 | -17% | 1 | 1 | 0% | 3,950 | 4,940 | +25% | 0 | 0 | — |
case-07 | pass→pass | 18,426 | 14,932 | -19% | 1 | 1 | 0% | 2,317 | 2,708 | +17% | 0 | 0 | — |
case-08 | pass→pass | 20,861 | 11,331 | -46% | 1 | 1 | 0% | 2,589 | 2,233 | -14% | 0 | 0 | — |
case-09 | pass→pass | 20,476 | 12,188 | -40% | 1 | 1 | 0% | 2,568 | 2,120 | -17% | 0 | 0 | — |
case-10 | pass→pass | 15,814 | 11,318 | -28% | 1 | 1 | 0% | 2,450 | 1,835 | -25% | 0 | 0 | — |
case-12 | fail→fail | 15,179 | 17,788 | +17% | 1 | 1 | 0% | 2,275 | 1,795 | -21% | 0 | 0 | — |
case-13 | pass→fail | 24,950 | 7,183 | -71% | 1 | 1 | 0% | 3,565 | 887 | -75% | 0 | 0 | — |
case-14 | pass→pass | 37,919 | 45,335 | +20% | 1 | 1 | 0% | 5,505 | 7,181 | +30% | 0 | 0 | — |
case-15 | pass→pass | 28,242 | 42,200 | +49% | 1 | 1 | 0% | 3,932 | 8,544 | +117% | 0 | 0 | — |
case-16 | fail→pass | 12,897 | 5,682 | -56% | 1 | 1 | 0% | 1,975 | 1,513 | -23% | 0 | 0 | — |
case-17 | fail→pass | 6,439 | 2,283 | -65% | 1 | 1 | 0% | 1,005 | 825 | -18% | 0 | 0 | — |
case-18 | fail→pass | 18,789 | 4,164 | -78% | 1 | 1 | 0% | 1,686 | 880 | -48% | 0 | 0 | — |
case-19 | fail→pass | 18,478 | 7,717 | -58% | 1 | 1 | 0% | 2,706 | 1,910 | -29% | 0 | 0 | — |
case-20 | fail→pass | 14,745 | 4,406 | -70% | 1 | 1 | 0% | 2,299 | 1,253 | -45% | 0 | 0 | — |
case-21 | pass→pass | 17,483 | 24,879 | +42% | 1 | 1 | 0% | 2,236 | 2,872 | +28% | 0 | 0 | — |
case-22 | fail→fail | 24,078 | 16,358 | -32% | 1 | 1 | 0% | 3,314 | 1,256 | -62% | 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 16 counted toward the lift figure. The other 6 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 16 comparable cases. 4 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.