How an agent uses a session

1. Check the connected profile and workspace.
2. Create a session with a clear task and access scope.
3. Share the invitation, then wait for the browser owner to start.
4. Observe the current page before issuing actions. Use the current element references and URL returned by that observation.
5. Make a change, observe again, and verify the intended result.
6. End the session and summarize the result, including any incomplete work.

An action succeeding means its browser tool completed; it is not proof that the overall task succeeded. A stopped agent must be prompted to continue. Docked does not host or automatically wake your agent.