What problem does it solve? AI answers about external facts—library versions, API signatures, CLI flags, pricing, error messages—are often stale or hallucinated when recalled from memory. This Skill replaces guessing with live retrieval: it fetches the actual page as agent-readable text and can drive a real browser for interactive tasks. ## Core Features & Use Cases - Fact-finding via read: Fetch any URL as markdown, get heading outlines, filter sections, and discover docs through llms.txt indexes—ideal for checking current versions, changelogs, config keys, and spec wording. - Browser automation: Navigate pages, snapshot interactive elements as refs, fill forms, click buttons, take screenshots, and capture console logs or JS errors for testing web apps. - Authentication support: Import browser state, use persistent profiles, or named sessions to work with logged-in pages. - Use Case: Before answering "what is the latest version of library X and how do I configure its timeout option", run agent-browser read on the official docs, quote the exact option name, and cite the URL instead of relying on memory. ## Quick Start Ask the agent to look up the current documentation for a library or open a website and fill out a form, and it will fetch the live page or drive the browser with agent-browser commands.
