Browser

**Quick judgment:** A solid, general-purpose headless browsing skill for OpenClaw that uses Puppeteer to load and scrape modern, JavaScript-heavy sites. Well-suited as a default "browser" for agents that need readable text from arbitrary URLs, but it is not a full web automation framework and inherits the usual fragility and overhead of Puppeteer-based tools. **Recommended scenarios** - When you need **clean text from dynamic or JS-rendered pages** that a simple HTTP fetch cannot handle. - For **research, summarization, or QA agents** that must browse arbitrary public webpages and pull back main content. - When you want a **single, simple entry point** (`openclaw browser read <url>`) for web content retrieval inside OpenClaw. **Key risks & limitations** - **Performance and resource cost:** Puppeteer is heavier than plain HTTP requests; repeated or parallel calls can be slow and resource-intensive. - **Reliability on hostile or complex sites:** Anti-bot measures, CAPTCHAs, login walls, and unusual frontends may cause failures or partial content extraction. - **Limited interaction surface (as described):** While the underlying tech supports DOM interactions, the documented primary interface (`read <url>`) focuses on one-shot page loads and text extraction, not complex multi-step workflows. - **Legal/ToS considerations:** As with any scraper, callers must ensure compliance with website terms of service, robots.txt, and applicable laws. Overall, this skill is a **good default browser integration** for OpenClaw agents needing modern web support. Use it when dynamic rendering is required, but prefer lighter HTTP-based skills for simple, static pages or high-volume scraping where performance is critical.