agent-browser

To automate browser tasks like form filling and data extraction, this Skill drives an external CLI to navigate pages, interact with elements, and extract data using compact refs to reduce context exchanged with AI agents.

Yes, you can take full-page screenshots for web testing by driving the external CLI to navigate a URL, snapshot elements, and execute a screenshot command with the full-page flag enabled.

You need a Bash-compatible runtime with the agent-browser binary installed and accessible in PATH, because the Skill relies on this external tool rather than embedded executable logic to perform browser automation.

The best way to automate navigation across internal dashboards and public sites is using this Skill to coordinate a CLI that opens pages, waits for network idle states, and performs element interactions with minimal context exchange.

To fill out web forms automatically, you use the CLI to open a page, snapshot interactive elements to get ref IDs, and then pass the ref ID and desired text to the fill command to complete the form entry.

A key limitation is the strict dependency on a Bash-compatible runtime with the specific external binary in PATH, meaning it cannot run browser automation tasks in environments lacking this external tool installation.