cyberneticlibrary

Browse and extract web content

agent-browserskillsetup L229,740
nanocoai/nanoclaw
What it does

Interact with web pages via snapshot tree and element-ref commands

Best for

Automating form submission, data extraction, and web app testing when refs are more resilient than XPath.

Inputs
  • · target URL
  • · snapshot mode (full accessibility tree or interactive-only)
  • · interaction command (click/fill/type) + element @ref
Outputs
  • · accessibility tree as JSON
  • · element refs (@e1, @e2...)
  • · page title/URL/text content
Requires
  • · Playwright MCP
  • · browser automation
Preconditions
  • · URL accessible
  • · browser session open
Failure modes
  • · client-side rendered pages need explicit wait-for-load
  • · DOM changes without re-snapshot invalidate old @refs
Trust signals
  • · ref-based element addressing (more stable than XPath)
  • · wait-for-element/URL/text patterns
  • · save/load authenticated browser state
  • · full-page screenshot support