Agent Browser Automation
Automates browser-based tasks including web testing, form filling, and data extraction through an interactive CLI.
The upstream-agent-browser skill empowers Claude to interact with the web as a functional agent. By providing a comprehensive suite of tools for navigation, element interaction, and state management, it allows Claude to perform complex workflows like end-to-end testing, scraping dynamic content, and automating administrative web tasks. It utilizes a unique referencing system to map interactive elements to simple IDs, ensuring high accuracy when clicking, typing, or extracting data from modern web applications.
主要功能
01Comprehensive Interaction Suite: Supports clicking, typing, hovering, scrolling, and drag-and-drop actions.
02Interactive Element Mapping: Generates snapshots with reference IDs for precise element targeting.
03Parallel Session Management: Run multiple browser instances simultaneously with saved authentication states.
04Visual Evidence & Debugging: Capture full-page screenshots, PDFs, and high-quality video recordings of sessions.
050 GitHub stars
06Advanced Network & State Control: Intercept network requests, mock API responses, and manage cookies or local storage.
使用场景
01Automated end-to-end testing and visual regression checks for web applications.
02Automating repetitive web-based workflows such as form submissions and report generation.
03Data extraction and web scraping from JavaScript-heavy or authenticated websites.