agent-browser provides an AI-friendly browser automation CLI that uses a compact snapshot+refs model to let agents navigate pages, identify interactive elements, and perform actions with far less context than full Playwright traces.
Use agent-browser when an agent needs to drive a real browser for end-to-end flows: form submissions, screen capture, automated testing, or debugging with video. It's especially helpful for long autonomous sessions where keeping context size small matters.
Designed for agent runtimes that can spawn subprocess CLIs or call shell tools (Claude Code and similar agent CLIs). Works well where token efficiency matters compared to Playwright-style traces.
This skill has not been reviewed by our automated audit pipeline yet.