claw-browser_skill

Overview

This skill provides a visual, human-like Chromium browser interface for OpenClaw agents. It renders JavaScript, handles dynamic single-page apps, captures screenshots, and can run in visible (non-headless) mode for demos or visual verification. It turns an agent from simple HTTP requester into a browser-powered assistant.

How this skill works

The skill launches a real Chromium instance (via Playwright) to load and render pages, execute client-side scripts, and interact with dynamic content. It supports two primary actions: read (extract page text/content) and screenshot (save a full-page image). Visible mode opens a real browser window when you need to see or demo the session.

When to use it

• Page requires JavaScript rendering or client-side execution
• Site is a SPA (React, Vue, Angular) or otherwise dynamic
• You need accurate visual output or full-page screenshots
• Traditional HTTP requests or API calls fail to return expected content
• User asks to open or visually inspect a web page

Best practices

• Always activate the virtual environment and install dependencies before running the tool
• Run playwright install chromium to download browser binaries; the browser will not launch without it
• Use visible mode for demos or debugging; use headless for automated runs or CI
• Limit navigation depth and set sensible timeouts to avoid long waits on slow pages
• Save screenshots with clear filenames and include timestamps for evidence or auditing

Example use cases

• Extract rendered article text from a news site that loads content via JavaScript
• Capture evidence.png of a web page for reporting or audit purposes
• Demo a live web interaction by launching a visible Chromium window for stakeholders
• Access and read dynamic dashboards or single-page apps that standard requests cannot load
• Automate visual checks of UI changes by taking periodic screenshots