Browser Automation

Browser Automation

Overview

Automate browser interactions: navigate pages, click elements, fill forms, take screenshots, extract data. Use whatever browser automation tool is available in your environment (e.g., playwright-cli, Playwright MCP, WebFetch, curl).

Instructions

1. Detect Tool

   • Check what browser tools are available. Prefer full browser (JS rendering, interaction) over HTTP-only tools.
   • If no browser tool is found, fall back to WebFetch or curl for read-only page access. Inform the user about limitations.

2. Navigate

   • Open the target URL. Wait for the page to load.

3. Observe Before Acting

   • Always capture page state (accessibility snapshot, DOM, or HTML) before interacting. Use structured data (accessibility tree, element refs) over raw screenshots when possible — structured data is better for reasoning.

4. Interact

   • Click, fill, type, select using element references from the snapshot.
   • After each interaction, re-capture page state to verify the result.

5. Capture Output

   • Take screenshots or extract data as requested.
   • Save files to the working directory.

6. Clean Up

   • Close browser sessions when finished.

Best Practices

1. Snapshot First: Always capture page state after navigation or interaction before taking next action.
2. Isolation: Use separate sessions for independent tasks.
3. Clean Up: Always close browser sessions when done.
4. Prefer Structured Data: Accessibility snapshots > screenshots > raw HTML for element identification.