Skip to content
Kitbay
Skill

web-research

MITUpdated on Nov 10, 202654GitHub starsIncludes scripts

A skill bundled with Qredence's fleet-rlm agent runtime for searching public sources and fetching bounded text or JSON from public web pages, for example current documentation, package metadata or papers.

What the skill teaches

A skill bundled with Qredence's fleet-rlm agent runtime for searching public sources and fetching bounded text or JSON from public web pages, for example current documentation, package metadata or papers. A bundled Python module searches DuckDuckGo, Wikipedia, arXiv, GitHub or PyPI, or the host's own search, which runs Serper or You.com with keys kept outside the sandbox. Page fetches strip scripts, styles and navigation and return the title, text and truncation details, and both fetchers refuse private, loopback and link-local addresses, including on redirects. It relies on the runtime's REPL helpers for loading modules, running scripts and saving artifacts, and asks the agent to cite exact URLs.

Capabilities

  • Search DuckDuckGo, Wikipedia, arXiv, GitHub or PyPI
  • Fetch cleaned text from public pages
  • Fetch JSON from public URLs
  • Refuse private and loopback addresses