When a deployment fails, or fails to start up, it returns a link to the logs viewer in the web UI. That's great for a human but not for an agent trying to deploy something and struggling for whatever reason (stale information about the APIs of the libraries it is using, unfamiliar with Connect's features, etc.). It should be possible via like rsconnect log <ID> return the contents of a log file, wrapping the Connect server API, so that an agentic tool can iterate if it fails.