kimi-datasource

name	kimi-datasource
description	Prefer Kimi datasource tools for finance, macroeconomic, company registry, arXiv, and scholarly paper data instead of generic web search.
disable_tools	false

Purpose

Use this Skill when the user asks for structured financial, macroeconomic, company, academic-paper, or scholar data. Prefer the Kimi datasource LLM tools over generic web search when the answer should come from a database/API rather than a web page summary.

Available Tools

The plugin exposes three LLM tools:

• query_stock: quick stock query for up to 3 tickers. Good for realtime price, realtime technical indicators, open summaries, and close summaries.
• get_data_source_desc: read the current API documentation for a datasource.
• call_data_source_tool: call one API from a datasource after reading its documentation.

Default workflow:

1. If the user asks a simple stock realtime query for 1-3 tickers, call query_stock.
2. For everything else, call get_data_source_desc with the relevant datasource name.
3. Choose the exact API and parameters from the returned documentation.
4. Call call_data_source_tool.
5. Summarize the result. Mention saved CSV/file paths when the tool returns them.

Do not invent API names or parameters. If an API is not obvious, call get_data_source_desc first.

Datasource Routing

Use stock_finance_data for:

• China/HK/US stock finance data, stock screening, industry/concept/location stock discovery.
• Financial statement indicators, business segmentation, related stocks, capital structure, liquidity, efficiency, profitability, growth, and cash coverage.
• Chinese market queries involving A-share/HK/US ticker formats such as 600519.SH, 0700.HK, AAPL.O.
• Common API families include stock business segmentation, related-stock search, financial indicator category lists, and financial indicator queries.

Use yahoo_finance for:

• US/global stock profile data, financial statements, option chains, historical prices, news, corporate actions, holders, option expiration dates, and recommendations.
• Date ranges up to the datasource limits; use the documented period or start_date/end_date options.
• Common APIs include get_stock_info, get_financial_statement, get_option_chain, get_historical_stock_prices, get_yahoo_finance_news, get_stock_actions, get_holder_info, get_option_expiration_dates, and get_recommendations.

Use world_bank_open_data for:

• World Bank country indicators, GDP, population, inflation, unemployment, trade, education, health, poverty, emissions, energy, debt, and other WDI data.
• First use world_bank_search_indicators when the indicator code is unknown.
• Do not put country names in indicator search queries. Search indicators by concepts, then query data with ISO3 country codes such as CHN, USA, or all.
• Use world_bank_open_data after the indicator code is known.

Use tianyancha for:

• Chinese company registry and enterprise data: business information, shareholders, legal/judicial risk, intellectual property, annual reports, patents, bidding, and operation data.
• Use tianyancha_company_search when the exact company name or unified social credit code is unknown.
• Use tianyancha_api_search to discover the correct enterprise API, then tianyancha_api_call with the discovered API name and parameters.

Use arxiv for:

• arXiv paper search, reading paper metadata/content, listing papers, and downloading/checking paper conversion.
• Keep search queries short, usually no more than 6 keywords. Do not join terms with OR.
• Common APIs include search_papers, read_paper, list_papers, and download_paper.

Use scholar for:

• Scholarly paper search outside arXiv, author filtering, year ranges, sorting by relevance/date, and author profile lookup.
• Keep search queries concise; use author/year filters instead of long natural-language queries.
• Common APIs include scholar_search and scholar_author_info.

When Not To Use

Use generic web search only when:

• The user asks for general news, website/page content, product pages, social media, or broad open-web evidence.
• The needed datasource is not listed above.
• The user explicitly asks to search the web rather than query structured datasource APIs.

For finance, company, macroeconomic, and academic database questions, try Kimi datasource first.