| name | browse-and-extract |
| description | Navigate to websites, interact with page elements, and extract structured data
using Playwright browser automation in isolated sandboxes. Use when the user asks
to visit a website, get content from a page, fill a form, or interact with a web app.
|
| metadata | {"author":"Troy Taylor","version":"1.0","pattern":"discovery"} |
Browse and Extract
When to Activate
- User asks "go to [URL]", "what's on [website]", "extract [data] from [URL]"
- User wants to interact with a web page (click, fill, submit)
- User asks to "scrape", "get the content", or "read from [site]"
Workflow
- Create session if none exists:
create_browser_session()
- Navigate:
navigate(session_id, url, wait_until: "domcontentloaded")
- Interact if needed:
click(session_id, selector) or fill(session_id, fields)
- Extract:
extract(session_id, selector, extract: "textContent", all: true)
- Present results as markdown table or summary
- Destroy session when done:
destroy_session(session_id)
Output Format
For structured data, present as a table. For text, present as a clean summary citing the source URL.
Missing Tools
If a Work IQ tool is needed but unavailable (e.g., saving results to OneDrive):
- Tell the user: "To save this to OneDrive, please enable Work IQ OneDrive in Cowork Settings > Tools, then try again."
