OpenCLI Browser Automation
Drives real Chrome browser sessions via OpenCLI to inspect pages, fill forms, and extract data using agent-optimized structured outputs.
This skill enables Claude to interact directly with a live Chrome browser using the OpenCLI framework, providing a robust interface for complex web tasks. It excels at navigating logged-in flows, handling dynamic DOM changes through advanced fingerprinting, and capturing network traffic, all while returning machine-readable JSON envelopes that minimize context usage. Designed specifically for agent-native interaction, it offers persistent session management and resilient element targeting, making it ideal for ad-hoc browsing tasks where standard APIs are unavailable.
主要功能
01Persistent session management for complex, multi-step browser workflows
029 GitHub stars
03Resilience to DOM drift using unique element fingerprinting and stable matching levels
04Semantic locators and accessibility-tree (AX) inspection for precise element identification
05Integrated network traffic capture to inspect API responses and JSON payloads
06Agent-native structured envelopes for reliable interaction feedback and error handling
使用场景
01Debugging web interactions by inspecting live browser state and capturing annotated screenshots
02Extracting structured data from dynamic single-page applications (SPAs) and React-heavy sites
03Automating multi-step form submissions and logged-in flows on sites without public APIs