browser-controllisted

# browser-control Use this skill when a task requires real browser interaction rather than plain HTTP fetch. Tool method: - `browser.control` Supported actions: - `open` - `navigate` - `click` - `fill` - `press` - `select` - `check` - `scroll` - `wait` - `content` - `snapshot` - `screenshot` - `close` Workflow: 1. Start with `action=open` (optional `url`). 2. Move page with `action=navigate`. 3. Interact with `action=click` / `action=fill` / `action=scroll`. Prefer semantic targets (`label`, `placeholder`, `field_name`, `role`, `target_text`) over brittle CSS selectors when possible. 4. If the page is still loading or reacts asynchronously, use `action=wait`. 5. For review or inspection tasks, prefer `action=snapshot` because it returns a structured page summary plus saved artifacts. 6. Use `action=content` when raw HTML/text is specifically needed. 7. Always finish with `action=close`. Session lifecycle: - Browser session is sticky for the current `profile_id + session_id`. - If a browser is already open for this chat session, reuse it instead of opening a new one. - Successful browser actions persist storage state for the same `profile_id + session_id`, so reopening the same chat session can restore cookies/local storage after process restarts. - Use `action=close` when the browser is no longer needed. - Use `action=close` with `clear_state=true` when the user wants to reset login state / cookies / local storage for that session. Required fields by action: - `navig