agent-browser

Browser tasks for AI agents are automated by driving a headless browser to navigate, interact, and extract data. This skill provides a CLI to perform actions like opening websites, filling forms, and taking screenshots with minimal context.

Snapshot-based refs capture compact element references like @eN for deterministic browser interactions. After navigating to a URL and rendering the page, you snapshot elements to obtain these refs and use them to click, fill, or select controls reliably.

Data extraction and screenshot capture are performed by driving a headless browser to render dynamic pages. The automation script navigates to the target URL, interacts with page elements via refs, and extracts text, attributes, screenshots, or PDFs.

Browser automation supports stateful sessions and persistence for AI agents. You can maintain session isolation across multiple interactions, allowing continuous form automation and data extraction tasks without losing the current page state.

Headless browser automation for web scraping relies on snapshot-based refs, which require re-snapshotting if the page DOM changes dynamically. It is best suited for deterministic interactions rather than highly volatile web pages with constantly shifting elements.

Browser automation is fully applicable for form filling, QA testing, and data extraction. By driving a headless browser, you can navigate to forms, use element refs to fill and submit data, and compare page states to verify application behavior.