agent-browser - Browser Automation for AI Agents

When to use this skill

• Open websites and automate UI actions
• Fill forms, click controls, and verify outcomes
• Capture screenshots or PDFs, or extract page content
• Run deterministic web checks with accessibility refs
• Execute parallel browser tasks via isolated sessions
• Prefer this over playwriter when you want disposable browser state instead of the user's already-running Chrome session

Instructions

Step 1: Pick the right browser surface

• Use agent-browser when the task should run in a fresh or isolated browser session, when browser state should be disposable, or when the flow needs to be reproducible in CI or automation.
• Route to playwriter instead when the task depends on the user's current Chrome state, saved logins, cookies, extensions, or already-open tabs.

Step 2: Refresh runtime guidance when version-specific behavior matters