The Web Browser Automation skill empowers Claude with powerful browser control and scraping capabilities by leveraging accessibility-tree snapshots and element references (@e1, @e2) rather than bulky raw DOM structures. This AI-optimized approach reduces context usage by up to 93% while enabling deterministic clicking, form filling, text extraction, visual screenshots, and isolated multi-session workflows. It is ideal for end-to-end web testing, authenticated scraping workflows, multi-agent swarm tasks, and general browser automation across complex web interfaces.
主要功能
01Context-efficient accessibility tree snapshots with deterministic element refs (@e1, @e2)
02Dynamic waiting mechanisms for network idle, selectors, specific text, and URL patterns
03Comprehensive interaction suite including click, fill, type, hover, select, and scroll
041 GitHub stars
05Isolated multi-session support for parallel multi-agent swarm navigation and auth sharing
06Seamless integration with Claude Flow MCP tools and persistent memory patterns
使用场景
01Running parallel scraping agents using isolated browser sessions and shared auth states
02Extracting structured data and capturing full-page screenshots for content verification
03Automating multi-step login flows and form submissions across authenticated web portals