stealthy-auto-browse

What This Skill Does

stealthy-auto-browse is a specialized browser automation tool designed to bypass sophisticated bot detection systems, including CreepJS, BrowserScan, Pixelscan, and Cloudflare. Unlike traditional automation libraries like Playwright or Puppeteer that utilize the Chrome DevTools Protocol (CDP)—a common fingerprint for security services—this skill leverages a custom Firefox fork called Camoufox running inside a Docker container. By avoiding CDP entirely and utilizing OS-level input simulation via PyAutoGUI, the browser remains indistinguishable from a legitimate human user. Because the browser does not "know" it is being automated, it does not leak the behavioral signals that modern anti-bot solutions use to flag and block headless traffic.

Installation

To begin, ensure you have Docker installed. Pull and run the container with docker run -d -p 8080:8080 -p 5900:5900 psyb0t/stealthy-auto-browse. Once running, expose the environment variable in your terminal: export STEALTHY_AUTO_BROWSE_URL=http://localhost:8080. For persistent configuration, add this entry to your ~/.openclaw/openclaw.json under the skills dictionary. Finally, run the command clawhub install openclaw/skills/skills/psyb0t/stealthy-auto-browse to link the skill to your OpenClaw agent.

Use Cases

This skill is designed for high-stakes environments where standard scrapers fail. Use it when you encounter Cloudflare challenge pages, DataDome, PerimeterX, or Akamai protection layers that trigger 403 Forbidden responses. It is ideal for maintaining persistent logged-in sessions on e-commerce sites, financial dashboards, or social platforms that actively fight headless automation. If you find your primary browser tools are returning empty pages or constant CAPTCHAs, this is your primary recovery tool.

Example Prompts

1. "Open the stealth browser, navigate to the target site, and use system_type to input my login credentials."
2. "Use stealthy-auto-browse to visit the dashboard, scroll down to the bottom of the feed to trigger lazy loading, and capture the page content."
3. "My scraping task is getting 403 errors on the checkout page; please switch to the stealthy browser and navigate to the cart using OS-level mouse clicks."

Tips & Limitations

Because this tool relies on OS-level inputs like PyAutoGUI, it is slower than traditional CDP-based automation. Use it only when necessary. If you are scraping static content with no bot protection, use WebFetch or curl instead, as they are significantly more resource-efficient. Always monitor the browser through the noVNC interface at port 5900 if your automation workflow seems stuck or requires troubleshooting of specific interaction elements.