Unified web access skill for search, scraping, login-required operations, and real-time browser interaction.

• Supports three access methods: WebSearch for discovery, WebFetch/curl for static content, and CDP browser automation for login-protected or dynamically-rendered platforms (Xiaohongshu, WeChat, Twitter, etc.)
• Operates through user's existing Chrome/Edge browser via CDP Proxy, preserving login state without separate browser instance; all operations run in background tabs
• Includes 10+ core commands: /eval for DOM queries and JavaScript execution, /click for UI interaction, /screenshot for visual capture, /scroll for lazy-load triggering, /navigate for page transitions, and file upload support
• Emphasizes human-like browsing logic: clarify success criteria first, choose the most direct path, validate results against goals, and adjust strategy based on evidence rather than repeating failed approaches
• Supports parallel sub-agent execution for independent research targets, with shared browser instance and automatic tab cleanup