Where do persmisison request from agent to agent land?

[Kwisss]

Guys I might be stupid but Claude is not giving me a clear answer, I have exhausted /doctor /debug and /manage-channels:

When an agent spawns a subagent, wired to nothing but just for him to talk to,
Where are permission request suppose to land? permanent sub-agents seem to get stuck on single turns when trying to do something?

Temporary sub-agents do not seem to have the same trait, During setup I choose telegram as the first channel to wire.

[Kwisss]

Closest to an explanation is this, I might not be experiencing a permission issue:

  1. Chatting (Requires Wiring)

  To send a message to your Telegram DM, an agent must be wired to that channel. Since TestAgent is unwired, it can never send you a chat message. If Nano is talking to TestAgent internally, that happens via
  the database/containers, not through your chat app.

  2. Permission Requests (The "Side-Channel")

  When an agent (even an unwired one like TestAgent) tries to do something restricted—like installing a package or using a secret—it doesn't "send a message" to you. Instead:

  1. The Request: TestAgent tells the Host: "I want to do X."
  2. The Gate: The Host sees that "X" requires approval.
  3. The Routing: The Host doesn't look at TestAgent's wiring. It looks at the Owner of the system (which is you).
  4. The Delivery: The Host finds a channel that reaches the owner (your Telegram DM) and sends the approval request there.

  So, TestAgent doesn't need to be wired to you to ask for permission. The Host handles the delivery. The request arrives in your DM as a system notification ("Approval requested by TestAgent..."), not as a
  standard chat message from a bot.