Prompt structure for my swarm agent

Hey everyone :waving_hand:

I’m building a multi-agent setup in n8n, where my Main Agent delegates tasks to various Sub Agents (like Email Agent Tool, Calendar Agent Tool, etc.).

The system works great so far for the functionality but there’s one issue:

Whenever I give the Main Agent a task (e.g. “create an event for XYZ and send a draft email for XYZ”), it routes the task correctly to the Sub Agents – but then they keep asking me follow-up questions, like:

  • “Do you want to proceed?”
  • “Are you sure you want to create this event?”
  • “Should I add anything else?”
  • “Please confirm before I send it.”

I don’t want that.

:repeat_button: I want the agents to just do the task immediately – no confirmation, no extra dialogue. If the data is already available, they should just execute and say something like:
:white_check_mark: “Calendar event created” or :e_mail: “Draft email ready”.


:red_question_mark: My question to the community:

What exact prompt structure or phrases do you recommend for:

  1. Main Agent prompt – so that it passes clear instructions and intent to the Sub Agents?
  2. Sub Agent prompt – so that they automatically execute the task without asking for confirmation or more input?

Ideally, I’d like to include fallback logic only if important data is missing, e.g.:
:cross_mark: “Missing date for event creation” – but not vague or open-ended follow-up questions.


Bonus: Here’s what I already tried

I added things like:

  • “Don’t ask for confirmation.”
  • “Execute directly when information is complete.”
  • “Avoid unnecessary dialogue.”

Still, the agents often act cautious and default to a more helpful/chatty mode.

Any suggestions for bulletproof phrasing or prompt structuring would be amazing!

Thanks in advance :folded_hands:

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.