remote-browser

The `remote-browser` skill enables agents running on sandboxed or headless environments to control a local browser remotely. This is particularly useful in environments where a graphical interface is unavailable, such as cloud virtual machines, continuous integration pipelines, or other remote coding agents. By providing a structured interface to navigate websites, interact with page elements, fill forms, take screenshots, and manage local development servers via tunnels, it addresses the challenge of automating web tasks in restricted execution contexts.

Key features of this skill include comprehensive browser automation capabilities such as navigation, tab management, scrolling, and element interaction through clicks, typing, and selections. Users can extract data by evaluating JavaScript, retrieving text, HTML, attributes, or bounding boxes of elements. Advanced interactions include uploading files, hovering, double-clicking, right-clicking, and sending keyboard commands. The skill also supports waiting for elements or text to appear and managing cookies. Multiple browser modes are available, including headless Chromium, cloud browser connections, and connecting via CDP URLs. Screenshots can be captured either as base64 or saved to disk, and browser sessions persist between commands to optimize workflow efficiency.

This skill is ideal for developers, QA engineers, or automation specialists who need to perform browser-based automation without direct GUI access. Common use cases include testing web applications in CI/CD pipelines, automating repetitive web tasks, scraping or extracting structured data from websites, and exposing local development servers to remote environments. Its flexibility and comprehensive command set allow users to perform almost any interaction typically possible in a full browser, making it an essential tool for automation in sandboxed or remote environments.