| name | agent-browser |
| description | Automate a real browser with the agent-browser CLI — open pages, snapshot interactive elements, click/fill by ref, extract content. Use when the user wants to browse, open a page, visit a URL, navigate a site, click or fill a form element, scrape, take a screenshot of a page, test a web app, or says "agent-browser". Works as a general-purpose browser automation tool in any repo, even ones without a local browser skill installed. |
agent-browser
General-purpose browser automation via the agent-browser CLI. Run agent-browser --help for the full command list.
Core workflow
agent-browser open <url> — navigate to a page
agent-browser snapshot -i — get interactive elements with refs (@e1, @e2, ...)
agent-browser click @e1 / agent-browser fill @e2 "text" — interact using refs
- Re-snapshot after the page changes (navigation, dynamic content, dialogs)
Local skills for heavy use
When a repo uses browser automation extensively (e.g. scripted test flows, recurring scraping jobs, repo-specific login dances), prefer installing a repo-local agent-browser skill or command that encodes the project's conventions. This global skill is the fallback for ad-hoc automation in repos without one.