browser-automationlisted

# Browser Automation Use a real browser client to complete a concrete, authorized task. Before acting, identify the target origin, intended outcome, authentication state, data sensitivity, and whether the action is read-only or changes external state. Never infer permission to access private sites, bypass controls, or submit irreversible changes. ## Plan the flow 1. Confirm the exact URL or application origin and the user-visible goal. 2. Check the available browser client and its version; prefer the repository's existing tool and conventions. 3. Open the page and capture a fresh DOM snapshot or accessibility tree. 4. Use stable semantic selectors, labels, roles, or test IDs. Avoid coordinates and brittle generated classes. 5. After navigation, modal changes, tab switches, or significant DOM updates, capture a new snapshot before using element references. 6. Bound navigation, selector, download, and retry waits; stop and report the failing stage when a bound expires. ## Interact safely Prefer `goto`, snapshot, inspect, interact, and snapshot again. For forms, verify field labels, destination, and visible validation before submitting. Treat file uploads, purchases, account changes, deletion, publication, messages, and permission changes as external side effects: preview the exact payload and obtain explicit user authorization immediately before the irreversible action. Never expose passwords, session cookies, API keys, authorization codes, or personal data in logs, scree