SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillswind-financial-data-skill

wind-financial-data-skill

Data & analytics

View repository

Updated 10-15 17:27

About this skill

Wind Financial fundamental data & RAG retrieval

This skill focuses on Wind Financial fundamental data + RAG retrieval (corresponding MCP server:vserver_wind_financial_data). Single responsibility, zero redundancy.

When to use

✅ Trigger scenarios:

  • Financial report numerical indicators (revenue / profit / assets and liabilities / cash flow / financial ratios / growth rates)
  • Valuation multiples (PE / PB / PS / dividend yield), cross-instrument / cross-index comparison
  • Macroeconomic indicators, industry indicators (natural language query)
  • Listed company announcements / periodic reports / annual reports / prospectuses / regulatory filings
  • Financial news summaries, research report excerpts (semantic retrieval)

❌ Non-trigger scenarios:

  • Real-time quotes, K-lines, minute-level data, sector constituents → please use wind-quote-skill
  • US stocks / European stocks / non-China concept stocks → not supported
  • Exchange rates / futures order book / cryptocurrencies → not supported
  • Non-financial data → not triggered

Workflow (3 steps)

All commands are prefixed with node ~/.claude/skills/wind-financial-data-skill/scripts/cli.mjs .

Step 1: List available tools

node ~/.claude/skills/wind-financial-data-skill/scripts/cli.mjs list-tools

Returns all tools of this server (search_financial_data / get_financial_documents), including the full inputSchema。

Step 2: Choose a tool + construct parameters

Based on the user's question and the tool descriptions, choose the most suitable tool and construct parameters according to inputSchema :

User intent Recommended tool Main parameters
Query specificvalues/indicators(financial reports, macro, valuation, growth rates, cross-instrument comparison) search_financial_data question(natural language)
Querydocument excerpts(news, research reports, announcements, annual reports, prospectuses) get_financial_documents query(natural language) + optional docType(1=news / 3=announcements, etc.)

Step 3: Call

node ~/.claude/skills/wind-financial-data-skill/scripts/cli.mjs call <tool_name> '<params_json>'

Typical examples

Use case A: Numerical query — "Kweichow Moutai 2024 revenue"

node ~/.claude/skills/wind-financial-data-skill/scripts/cli.mjs call \
  search_financial_data \
  '{"question":"贵州茅台 2024 年营业收入"}'
# → 结构化数值（营收金额）

Use case B: Document retrieval — "Kweichow Moutai 2024 annual report discussion on capacity expansion"

node ~/.claude/skills/wind-financial-data-skill/scripts/cli.mjs call \
  get_financial_documents \
  '{"query":"贵州茅台 2024 年报 产能扩张","docType":"3","top_k":5}'
# → 公告类文档片段

Actual parameters are subject to the list-tools returned by inputSchema (docType: 1=news / 2=research reports / 3=announcements / ...).

API Key configuration

Inquiry-based acquisition process (must be followed when the Key is missing）

When the user's request requires this skill, but cli.mjs reports ❌ WIND_API_KEY 未配置:

  1. First ask the user, something like: "You haven't configured the Wind API Key yet. Would you like me to open the developer center for you now to get a Key?"
  2. User agrees → run the open-portal subcommand: 
    node ~/.claude/skills/wind-financial-data-skill/scripts/cli.mjs open-portal
    The command returns JSON, in which:
    • url field is the link to visit (if logged in, it goes directly to overview; if not logged in, the SPA will automatically redirect to /#/login）
    • fallback_message show the user a fallback message (especially important when spawn fails in headless / remote environments)
  3. User refuses → directly tell the user to visit:https://aimarket.wind.com.cn/#/user/overview, obtain the Key yourself after logging in
  4. After the user gets the Key, guide them to choose one of the following three configuration methods:
# A. 环境变量
export WIND_API_KEY=ak_xxx

# B. 本 skill 目录 config.json（仅本 skill 用）
echo '{"wind_api_key":"ak_xxx"}' > ~/.claude/skills/wind-financial-data-skill/config.json

# C. 全局 config（推荐：所有 wind 系列 skill 共享一份）
mkdir -p ~/.wind-aimarket && echo "WIND_API_KEY=ak_xxx" > ~/.wind-aimarket/config

Three-level fallback priority:env > skill 目录 config.json > 全局 config

Common errors

Error Handling
WIND_API_KEY 未配置 Follow the aboveinquiry-based acquisition process: first ask the user → run open-portal → guide configuration
MCP 调用失败 HTTP 401 / 403 Key invalid → run open-portal to have the user regenerate it in the developer center
MCP 调用失败 HTTP 5xx Server error → retry later or check status.wind.com.cn
tool not found Tool name misspelled, first run list-tools to confirm
params JSON 解析失败 Note JSON escaping (double quotes, Chinese requires UTF-8)

Data source attribution

When presenting results to the user, you must include: 数据来源于 Wind 万得金融终端。

Self-check (before responding)

  • Is the user's question anumeric metric / document retrieval type(not real-time quotes, not US stocks, not exchange rates)? Yes → continue; No → do not trigger this skill
  • Has WIND_API_KEYbeen configured? Not configured → after asking the userrun open-portal, do not pop up the browser without notice
  • Is the data source attributed in the results?