browser-automation

Unified browser automation for AI agents. Use when Codex or Claude needs to open websites, click buttons, fill forms, log in, take screenshots, scrape data, download files, verify UI behavior, test web apps, or automate browser actions. Default to `agent-browser` for general browsing, session reuse, authentication, screenshots, extraction, responsive checks, and parallel agent workflows. Use the `playwright-cli` path only when the task explicitly needs request mocking, tracing, console or network inspection, fine-grained mouse events, dialog handling, explicit Firefox/WebKit/Edge coverage, or custom Playwright code.