a-stock-data is a self-contained AI Skill file, published under Apache-2.0, that consolidates raw China A-share market data from 22 sources into 60 endpoints an AI coding assistant can call directly, and it is aimed at quantitative researchers, investment analysts and developers who want A-share data inside an LLM workflow without writing their own scraping and authentication code.

What it is

The project is a single skill file that combines structured Markdown with embedded Python. It lives in the ecosystem of AI coding assistants rather than in the ecosystem of Python data libraries: the README states that it is compatible with Claude Code, Codex and OpenClaw, and that any assistant supporting context injection can use it. The file organises its coverage as a twelve-layer architecture at version V3.8.0, spanning quotes, research reports, signals, fund flows and chip distribution, news, fundamentals, announcements, limit-up boards, options, investor sentiment, macro data and index calendars, and it counts 60 endpoints made up of 55 main entries plus 5 fallbacks.

It solves a concrete integration problem. The README frames the work it removes as memorising mootdx K-line parameters, the Eastmoney PDF Referer header, and iwencai X-Claw authentication. Rather than stitching sources together per project, the skill exposes them behind named endpoints, prioritises sources that do not block by IP such as mootdx and Tencent, restricts Eastmoney to data only it provides with built-in rate limiting, and ships a fallback source quick-reference and degradation strategy for when a primary source is blocked.

Key capabilities

• Twelve-layer quote and fundamentals coverage combining mootdx, Tencent, Baidu K-line, Sina K-line with MA5/MA10/MA20, order book, PE/PB, market capitalisation, indices, ETFs and qfq/hfq adjustment factors.
• 60 endpoints documented in the README, with an explicit counting convention: 55 main entries plus 5 fallbacks, where the Eastmoney industry report and individual report share one endpoint distinguished by the qType parameter.
• Fund-flow and chip data, including 120-day daily net inflow by order size and locally computed CYQ chip distribution returning profit ratio, average cost, the 90-70 cost range and chip peaks.
• A board-trading layer built on Eastmoney push2ex and 同花顺, covering the limit-up pool, failed limit-ups, limit-downs, prior-day limit-ups, reasons for limit-ups, and the consecutive-board ladder.
• Announcement coverage across all Shanghai, Shenzhen and Beijing exchange filings through cninfo, alongside a news layer combining Eastmoney, Cailianshe telegraph and 7×24 global news.
• A macro layer drawing on the People's Bank of China and the National Bureau of Statistics for monthly social financing increments across twelve columns and PMI series for manufacturing, non-manufacturing, composite and firms by size.
• An options layer that returns Sina ETF option T-quote data with greeks and implied volatility, plus an index calendar with constituents, weights, PE and dividend yield and the official trading calendar.

Who uses it and how

• Claude Code users install the file to ~/.claude/skills/a-stock-data and activate it by asking a natural-language question, such as requesting a valuation check on ticker 688017.
• Codex and OpenClaw users paste the contents of SKILL.md into a system prompt or a project context file, where the embedded Python can be executed directly.
• Quantitative finance and investment research teams use it to pull research report lists, EPS forecasts and PDF downloads from Eastmoney, 同花顺 and iwencai without implementing per-source authentication.
• Analysts working with volatile upstream sources rely on the fallback quick-reference, which covers substitutes for the dragon-tiger list, fund flows and announcements, plus official margin trading data and Beijing Stock Exchange quotes and order book fallbacks.

Getting started

The README gives a three-step, two-minute install: create ~/.claude/skills/a-stock-data, download SKILL.md from the repository's raw URL into it, then run pip install mootdx requests pandas stockstats numpy baostock xlrd openpyxl. Since V3.0 the akshare dependency is no longer required.

How it compares

Among the libraries named in the facts, akshare was a former dependency that V3.0 dropped, while mootdx, baostock and stockstats remain pip dependencies that expose raw data access. This project sits above them as a packaging layer, wrapping 22 sources behind 60 endpoints and removing the need to recall per-source parameters.

When to use it — and when not to

A self-hoster installs only a skill file and a set of Python packages, with no database, storage or mail service to operate, and the zero-authentication design means no keys to manage. Limits are documented rather than hidden: the valuation history endpoint does not support the Beijing Stock Exchange, Shenzhen Connect northbound disclosure has been tightened so that stream is reference-only with HKEX as the authoritative fallback, and every layer depends on third-party upstream sources that can throttle or change. Anyone who needs a vendor with a contractual service level, licensed redistribution rights or Beijing Stock Exchange valuation history should look elsewhere.

A 股全栈数据 · 十二层架构 · V3.8.0
│  （优先级：mootdx/腾讯 不封IP 优先用；东财仅用于独有数据，已内置限流防封）
├── 行情层    mootdx + 腾讯 + 百度K线 + 新浪  K线(带MA5/10/20) + 五档盘口 + PE/PB/市值 + 指数/ETF
│                                           + 复权因子 qfq/hfq  ★V3.7
├── 研报层    东财 reportapi + 同花顺 + iwencai  个股研报 / 行业研报 / PDF下载 / 一致预期 / NL搜索
├── 信号层    同花顺 + 东财                  强势股 + 题材归因 + 北向资金 + 板块归属
│                                           + 资金流向(push2) + 龙虎榜 + 全市场龙虎榜 + 解禁 + 行业对比 + 板块资金流
├── 资金面    东财 datacenter + push2        融资融券 + 大宗交易 + 股东户数 + 分红送转 + 资金流(分钟+120日)
│   /筹码     本地计算                       筹码分布 CYQ：获利比例 / 平均成本 / 成本区间 / 筹码峰  ★V3.7
├── 新闻层    东财 + 财联社                  个股新闻 / 财联社电报(✅V3.4复活) / 全球资讯（互备）
├── 基础数据  mootdx + 东财 + 新浪           季报37字段 / F10九大类 / 财报三表
│           + baostock + 申万                估值历史(PE/PB/PS+换手率+停牌+ST) / 上市退市日 / 申万行业变迁史  ★V3.7
├── 公告层    巨潮 cninfo + mootdx           沪深北全量公告
├── 打板层    东财 push2ex + 同花顺          涨停池 / 炸板 / 跌停 / 昨涨停 / 涨停原因题材 / 连板梯队
│                                           + 重点监控池 + 日内异动池  ★V3.6
├── 期权层    新浪 hq.sinajs                ETF期权 T型报价 / 希腊字母 / 隐含波动率 IV  ★V3.3
├── 舆情互动  巨潮互动易 + 同花顺 + 东财     互动易问答 / 同花顺热榜 / 东财人气榜 / 概念命中  ★V3.3
├── 宏观层    人民银行 + 国家统计局          社会融资规模增量(月度12列) / PMI(制造业·非制造业·综合·大中小型)
└── 指数日历  中证指数 + 国证指数 + 深交所    指数成分 / 权重 / PE与股息率 / 官方交易日历

# 1. 创建 skill 目录
mkdir -p ~/.claude/skills/a-stock-data

# 2. 把 SKILL.md 放进去
curl -o ~/.claude/skills/a-stock-data/SKILL.md \
  https://raw.githubusercontent.com/simonlin1212/a-stock-data/main/SKILL.md

# 3. 安装依赖（V3.0 不再需要 akshare）
pip install mootdx requests pandas stockstats numpy baostock xlrd openpyxl