External Search

Use this skill when a request needs external web evidence or content from specific URLs.

Scripts

• external_info_search.py: uses the existing Moonshot Formula tools/fibers loop, and writes Kimi's final message.content as the primary answer observation.
• url_content_reader.py: reads user-provided URLs; it does not discover new pages.

Routing Notes

Use external_info_search.py for web discovery, current facts, event dates, news, announcements, and non-AInvest sources. Downstream reasoning must consume payload answer first. articles only contains citations/references extracted from Kimi's final response when available.

Do not call legacy search engines, crawl returned pages, or rebuild Kimi's answer from tool arguments. Execute requested Formula fibers, submit their outputs as tool messages, continue the model loop, and consume the final answer.

Use url_content_reader.py only when the user provides one or more URLs and asks to read, summarize, extract, or analyze them.

The Kimi script is pinned through kimi-search.lock.json to the public weiweidesigner/kimi-search repository. Credentials remain in the host ignored .env or process environment. url_content_reader.py is separately pinned through url-content-reader.lock.json to the public weiweidesigner/url-content-reader repository. Its logical host entry and URL-reading output contract remain unchanged.