Install any skill in seconds. Free to start, no credit card required.
Get Started Free →跨市场联动分析(A股/港股/美股/大宗商品)。当用户说"跨市场"、"联动"、"A股和美股"、"港股联动"、"AH溢价"、"中概股"、"美股跌了A股会跌吗"、"原油对A股影响"、"汇率影响"、"cross market"、"人民币贬值利好什么"、"跨市场联动"、"全球市场"时触发。MUST USE when user asks about cross-market correlation, AH premium, how US/HK markets affect A-shares, or commodity/FX impact on Chinese stocks. 分析A股与全球主要市场(美股/港股/商品/汇率)的联动关系、AH溢价、中概股映射、大宗商品传导。支持研报风格(formal)和快速参考风格(brief)。
.claude/skills/aifinlab-a-share-cross-market/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | 135% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 73% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 102% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 72% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -7% | 0% |
bash# cn-stock-data 统一层获取行情(跨市场用 snowball 路由) python -c " import subprocess, json # A股主要指数(通过 cn-stock-data → efinance 路由) # 上证指数 000001, 沪深300 000300, 创业板指 399006 # 用 cn-stock-data skill 的 quote/kline 接口 # 港股(snowball 路由,HK 前缀) # 恒生指数 HK00001, 恒生科技 HKHSTECH, H股指数 HK00011 # 美股(snowball 路由,US 前缀) # 标普500 US.SPX, 纳斯达克 US.IXIC, 道琼斯 US.DJI # 需要 pysnowball token 才能获取港美股数据 import pysnowball as ball # ball.set_token('your_token') # ball.quotec('SH000001,HKHSI,.DJI,.IXIC,.INX') # 跨市场报价 # ball.kline('SH000001', period='day', count=-120) # K线数据 " # AH 股比较数据(akshare 直接调用) python -c " import akshare as ak # AH 股比价 df = ak.stock_a_ah_comparison_em() print(df.head(20).to_string()) # 返回字段:代码/名称/A股价格/H股价格/比价(A/H)/溢价率 " # 大宗商品(akshare) python -c " import akshare as ak # 国际原油 df = ak.futures_foreign_hist(symbol='CL') # WTI原油 # 伦敦金 df = ak.futures_foreign_hist(symbol='GC') # COMEX黄金 # LME铜 df = ak.futures_foreign_hist(symbol='HG') # COMEX铜 print(df.tail(5).to_string()) "
补充 web 搜索:美股收盘情况、大宗商品实时价格、美元指数、离岸人民币汇率、美联储议息会议纪要、全球央行动态。
Step 1: 多市场数据获取 根据用户需求确定分析范围:
获取数据(近 60-120 个交易日): | 市场 | 标的 | 数据源 | |------|------|--------| | A股 | 上证/沪深300/创业板指 | cn-stock-data (efinance路由) | | 美股 | S&P500/纳斯达克/道琼斯 | cn-stock-data (snowball路由) | | 港股 | 恒生指数/恒生科技/H股指数 | cn-stock-data (snowball路由) | | 商品 | 原油/黄金/铜/铁矿石 | akshare futures | | 汇率 | 美元指数/离岸人民币 | web搜索 |
Step 2: 相关性与领先滞后分析
相关系数计算:
领先滞后关系:
联动强度变化:
Step 3: AH 溢价分析(如涉及)
通过 ak.stock_a_ah_comparison_em() 获取 AH 股数据:
| 溢价区间 | 市场状态 | 含义 | |---------|---------|------| | <110 | 罕见低溢价 | A股相对低估或港股相对高估 | | 110-130 | 正常区间 | 反映制度差异溢价 | | 130-150 | 偏高 | A股情绪偏热或港股承压 | | >150 | 极端高溢价 | A股显著高估,关注回归风险 |
Step 4: 传导机制分析
资金流传导:
情绪传导:
基本面传导:
政策传导:
Step 5: 输出
| 维度 | formal(机构跨市场研究) | brief(快速跨市场参考) | |------|----------------------|----------------------| | 篇幅 | 4-8 页 | 1-2 页 | | 市场覆盖 | A股+美股+港股+商品+汇率全覆盖 | 聚焦用户关心的2-3个市场 | | 相关性分析 | 完整相关系数矩阵+滞后分析 | 定性描述联动方向和强度 | | AH溢价 | 完整AH溢价表+历史对比 | 整体溢价水平+异常个股 | | 传导机制 | 详细多路径分析 | 核心传导链1-2条 | | 历史对比 | 类似历史情景回顾 | 不需要 | | 免责声明 | 需要 | 不需要 |
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-13 | pass→pass | 18,637 | 24,647 | +32% | 1 | 1 | 0% | 2,899 | 5,914 | +104% | 0 | 0 | — |
case-15 | pass→pass | 15,280 | 21,931 | +44% | 1 | 1 | 0% | 2,314 | 4,904 | +112% | 0 | 0 | — |
case-14 | pass→pass | 17,234 | 26,094 | +51% | 1 | 1 | 0% | 2,616 | 6,149 | +135% | 0 | 0 | — |
case-11 | pass→pass | 9,460 | 6,588 | -30% | 1 | 1 | 0% | 1,475 | 3,389 | +130% | 0 | 0 | — |
case-12 | fail→pass | 16,982 | 23,918 | +41% | 1 | 1 | 0% | 2,303 | 5,402 | +135% | 0 | 0 | — |
case-01 | fail→fail | 49,191 | 70,431 | +43% | 1 | 1 | 0% | 8,287 | 8,674 | +5% | 0 | 0 | — |
case-02 | fail→fail | 30,747 | 8,092 | -74% | 1 | 1 | 0% | 5,397 | 2,747 | -49% | 0 | 0 | — |
case-03 | fail→pass | 23,059 | 30,400 | +32% | 1 | 1 | 0% | 3,229 | 5,590 | +73% | 0 | 0 | — |
case-04 | fail→pass | 22,603 | 26,072 | +15% | 1 | 1 | 0% | 3,417 | 6,909 | +102% | 0 | 0 | — |
case-05 | fail→fail | 30,745 | 42,254 | +37% | 1 | 1 | 0% | 4,147 | 7,536 | +82% | 0 | 0 | — |
case-06 | pass→pass | 16,483 | 19,046 | +16% | 1 | 1 | 0% | 2,290 | 4,235 | +85% | 0 | 0 | — |
case-07 | pass→pass | 14,862 | 15,799 | +6% | 1 | 1 | 0% | 2,547 | 4,827 | +90% | 0 | 0 | — |
case-08 | fail→pass | 15,880 | 13,182 | -17% | 1 | 1 | 0% | 2,407 | 4,138 | +72% | 0 | 0 | — |
case-09 | fail→pass | 17,309 | 6,210 | -64% | 1 | 1 | 0% | 3,474 | 3,247 | -7% | 0 | 0 | — |
case-10 | fail→pass | 14,371 | 16,524 | +15% | 1 | 1 | 0% | 2,589 | 4,222 | +63% | 0 | 0 | — |
case-16 | pass→pass | 18,919 | 25,419 | +34% | 1 | 1 | 0% | 2,795 | 4,842 | +73% | 0 | 0 | — |
case-17 | pass→pass | 26,599 | 10,885 | -59% | 1 | 1 | 0% | 2,109 | 3,907 | +85% | 0 | 0 | — |
case-18 | pass→pass | 14,786 | 16,489 | +12% | 1 | 1 | 0% | 1,956 | 4,649 | +138% | 0 | 0 | — |
case-19 | pass→pass | 22,358 | 15,342 | -31% | 1 | 1 | 0% | 4,044 | 4,887 | +21% | 0 | 0 | — |
case-20 | pass→fail | 47,197 | 7,771 | -84% | 1 | 1 | 0% | 8,242 | 2,593 | -69% | 0 | 0 | — |
case-21 | pass→pass | 19,735 | 24,495 | +24% | 1 | 1 | 0% | 3,401 | 6,091 | +79% | 0 | 0 | — |
case-22 | pass→pass | 24,294 | 37,955 | +56% | 1 | 1 | 0% | 4,582 | 7,891 | +72% | 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. 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 | +9% |
Other measured skills in the registry, with their headline benchmark lift.