Build-plan deployed agent stays Sleeping after CreateDispatch; never joins room (outbound SIP) is almost the opposite problem, isn’t it - in that case an agent would not wake up and join a room, whereas you re asking about how to PREVENT agents joining a room.
Dispatches refer to agent dispatches, whether they are explicitly dispatched or dispatched automatically, the process is an agent in the project will join the specified room. Your project does not have any agents in it, so it would not be possible for an agent to join any room.
I’m basing that on your description of your project as “without AI agents”, and I also do not see any agents deployed to the project associated with your account.
To answer your specific questions though:
Yes, always provide your agents with an agent_name.
No, automatic dispatches require unnamed agents (and are not recommended)
Assuming you are referring to a malicious actor creating an externally hosted worker, they would need your API key in order to connect to your LiveKit instance. Your LiveKit cloud dashboard will list ALL agents (including externally hosted ones) at https://cloud.livekit.io/projects/p_/agents
You have complete control over whether agents get dispatched to your rooms, name all your agents and check your code base for agents being dispatched via API, agents being dispatched via token, or agents being dispatched by a SIP dispatch rule. The docs page at Agent dispatch | LiveKit Documentation is a good place to start.