agent-browser

What This Skill Does

The agent-browser skill provides an advanced command-line interface for browser automation powered by Playwright. Designed for the OpenClaw AI agent, it allows the agent to navigate the web, interact with DOM elements, and extract information with high precision. By leveraging accessibility tree snapshots, it offers a more robust way to target web elements compared to traditional coordinate-based scraping, making it highly effective for complex, dynamic web applications.

Installation

To integrate this skill into your environment, use the OpenClaw CLI tool:

clawhub install openclaw/skills/skills/joshhuang123/agent-browser-cli

Ensure that Node.js is installed on your system. After installation, verify the environment by running the dependency setup command provided by the agent-browser package:

npm install -g agent-browser agent-browser install

Use Cases

This tool is ideal for repetitive tasks and automated research. Use it for automated check-ins on attendance or loyalty platforms, filling out complex web forms across multiple pages, scraping specific data points from websites that require JavaScript execution, and conducting automated visual regression or documentation tasks through page snapshots and annotated screenshots.

Example Prompts

1. "Open https://example.com/login, fill in the username field with 'my_user', enter 'secret_pass' in the password field, and click the submit button."
2. "Go to the company dashboard, capture a snapshot to identify the latest report link, click it, and save a screenshot of the results to my workspace."
3. "Check for new items on the product page and alert me if the 'Buy Now' button appears."

Tips & Limitations

• Snapshot First: Always run agent-browser snapshot before performing actions to ensure your references (@ref) are current and valid.
• Handle Latency: Web pages are dynamic. Always include sleep commands or explicit waits after navigation to allow for elements to render.
• Resource Management: Keep your browser sessions clean; always execute agent-browser close to free up system memory and terminate underlying browser processes after your task is finished.
• Complexity: While powerful, this tool requires the agent to be aware of the DOM structure. If a website changes its layout, you may need to re-verify your targeting logic.