browser-use

name	browser-use
description	Web browser automation skill for navigating pages, filling forms, clicking elements, taking screenshots, and extracting web data using AI-driven browser control.

browser-use

Automate web browser interactions using AI-driven browser control.

When to use

• The user needs to interact with a web page (fill forms, click buttons, navigate).
• Web scraping or data extraction from dynamic pages is required.
• Visual verification of web applications (screenshots, element inspection).
• Testing web UIs or checking deployment status.

Instructions

1. Tool selection: Prefer the browser-use MCP server if available. Fallback to the built-in web-browse tool for simple page fetching.
2. Navigation: Use browser-use open <url> to navigate to pages. Wait for load completion.
3. Interaction sequence: Plan the interaction as a series of steps:
   • open → navigate to URL
   • click → click on element by selector or text
   • type → enter text into input fields
   • screenshot → capture current page state
   • extract → pull structured data from page
4. Element identification: Use CSS selectors, ARIA labels, or visible text to identify elements. Prefer semantic selectors over brittle XPath.
5. Error handling: If an element isn't found, take a screenshot to diagnose, then retry with an alternative selector.
6. Data extraction: Output extracted data as structured JSON or Markdown tables.
7. Rate limiting: Add reasonable delays between actions to avoid rate limiting or bot detection.
8. Security: Never enter real credentials. Use test/demo accounts only. Never screenshot sensitive information.
9. Reference: browser-use/browser-use — 88k+ stars, MIT license, Python + Chromium required.