Record a Browser Workflow

Browser beats capture semantic application states and interactions rather than arbitrary screen pixels.

1. Start the application in setup or hand a URL from a terminal action to a named browser pane.
2. Open the page with an explicit readiness condition.
3. Prefer stable semantic targets such as test_id, role, label, or text.
4. Use browser primitives such as click, type_text, drag, and move_pointer.
5. Wait for the state caused by an interaction before continuing.
6. Add checks for the visible or persisted result.

Use component-relative percentages only when the page has no meaningful DOM destination. Authentication values belong in a recording-local app.secret.env or declared CI source; do not put them in visible command configuration.

See the recording schema for exact browser targets, completion conditions, and timing fields.