Install
https://github.com/wind-information-co-ltd/wind-skillsWho this is for
Open this monorepo only when you need to connect Wind financial data to an AI Agent over the MCP protocol and use Wind’s own data plus community analysis workflows. Officially it targets AI Agents such as Claude / OpenClaw / Hermes, with 34 finance skills in total.
It fits people who already have a finance research question and accept going through the “financial capability entry” first before deciding what to install. The body names Claude, OpenClaw, and Hermes as integration targets, then adds “and others”—do not read that as support for only those three. If you do not need Wind data, this skill set is a mismatch.
Page: https://github.com/wind-information-co-ltd/wind-skills
Finish a first try in 3 minutes
The body does not give npx, plugin, or any other copyable installer. The official recommended path is written on the financial capability entry: it enumerates every skill on the platform, recommends by the user’s question, and guides install / upgrade. For the first try, only open the official page. Do not invent install commands, and do not name directories that never appear.
Precondition: your question is about Wind financial data or Alice analysis, not generic chat.
Do a read-only check with the agent: what the financial capability entry does; which Wind data it can reach; which sub-capabilities the Alice entry will run. Success means the agent can restate “recommend by the user’s question and guide install / upgrade”, and can name stocks (A-share/Hong Kong/US quotes and financials), funds, indexes/sectors, bonds, company announcements and news, macroeconomic indicators, plus Alice’s company one-pager, research question list, fact verification, earnings commentary, thematic stock picking, fund analysis, macro/bond/credit analysis, market-size estimation, and comparable-company analysis. If the agent invents a skill folder name or install command that never appeared, stop.
Usage cases
Case 1: Reach Wind financial data. The body covers stocks (A-share/Hong Kong/US quotes and financials), funds (quotes and full-dimension data), indexes/sectors, bonds, company announcements and news, macroeconomic indicators. Give the agent a concrete quote or announcement question and require it to say first which of those data classes it will use. This is outbound data access; before running, confirm the instrument you want sits in those classes. The page does not write an MCP connection string or field names—do not add them.
Case 2: Use Alice specialist financial analysis. The Alice entry uses the A2A protocol + SSE streaming and runs the named sub Skills. When you want a company one-pager, earnings commentary, thematic stock picking, or comparable-company analysis, first have the agent restate which sub-capabilities it will enable, then continue. Success means sub-capability names come from the body, not newly invented analysis modules. The page gives no A2A address or SSE debug steps; stop the check at the protocol names.
Usage limits
There is no copyable installer command. The financial capability entry only promises to guide install / upgrade; do not rewrite that sentence into an install command that never appeared.
The 34 finance skills are not named one by one in the body. Apart from the data classes and Alice sub-capabilities already written, do not add names.
Integration is written as the MCP protocol, plus Claude / OpenClaw / Hermes and others; do not frame it as an exclusive client, and do not invent other protocols.
Do not add account signup, permissions, billing, or local cache details that were never written. Fact verification is an Alice sub-capability name; it does not mean extra release steps.