us-stock-datasourcelisted
Install: claude install-skill starkiller-stucky/us-stock-datasource
# US Stock Datasource(美股数据源)/ US Stock Datasource
一个**元技能** / A **meta-skill**:它不直接提供行情,而是负责「**分层架构 → 向导配置 → 运行时路由**」,
让 Agent 在任何美股数据需求下自动选对数据源、并如实标注数据新鲜度。
It does not serve quotes directly; it owns **layered architecture → wizard setup → runtime routing**,
so the agent always picks the right source and honestly labels data freshness.
> 范围约定 / Scope:本技能**只做取数,不做交易下单**。/ This skill **only fetches data, never places orders**.
---
## 一、十层数据源架构 / 1. Ten-Layer Architecture
数据源必须分层,因为每一层解决不同问题、时延和成本不同,**上层不能替代下层**。
Data sources must be layered: each layer solves a different problem with different latency/cost,
and **a higher layer can never substitute a lower one**.
完整清单见 / Full source list:[references/layers.md](references/layers.md)。
```
L1 实时行情 Real-time quotes 券商 feed → Finnhub → yfinance(15min delay)
L2 财报原文 SEC filings SEC EDGAR → sec-api.io → yfinance(aggregated)
L3 基本面聚合 Fundamentals yfinance → SimFin
L4 分析师观点 Analyst views yfinance estimates → TipRanks
L5 新闻事件 News Finnhub news → Marketaux → Yahoo news
L6 散户情绪 Retail sentiment Adanos → StockTwits
L7 宏观指数 Macro FRED → yfinance ^VIX
L8 期权微观结构 Options yfinance options → Polygon
L9 筛选与广度 Screening/breadth Finviz free → yfinance Screener
L10 历史回测 Historical/backtest yfinance history → SimFin
```
## 二、两种工作模式 / 2. Two Modes
### 模式 A:向导安装模式 / Mode A — Setup Wizard(首次配置 / 重配)
```bash
python wizard.py # 交互式完整向导 / interactive full wizard
py