What problem does it solve? Automated web tasks require a browser, but this environment exposes no native browser tool. This Skill provides a CLI-driven way to open pages, click elements, fill forms, scroll, take screenshots, and inspect page state in a persistent Chromium session. ## Core Features & Use Cases - Full Browser Interaction: Open URLs, click elements by snapshot refs, fill inputs, press keys, hover, scroll, and wait for network idle states. - Page Inspection & Evidence Capture: Take DOM snapshots with element refs, read URLs and titles, evaluate JavaScript, and save viewport or full-page screenshots to specified paths. - Persistent Daemon Session: A browser daemon starts on first command and persists across calls, supporting multi-step flows like form submissions and batch JSON command sequences. - Use Case: Complete a web form submission task by opening the page, snapshotting to get element refs, filling fields, clicking submit, waiting for navigation, and saving a screenshot of the confirmation page as evidence. ## Quick Start Use the agent-browser CLI to open the target website, snapshot the page to find element refs, fill in the form fields, click submit, and save a screenshot of the result to /tmp_workspace/results/screenshots.
