Does anybody have a prompt that stops the dictation AI (groq) from outputting text that's an response to your question instead of just dictating what you say?

[clayleopardlabs]

This is driving me absolutely insane. It keeps responding to the things I'm trying to dictate

It just pasted this:

"I am not capable of accessing external information or visual tools, but I can repeat what you say. You said: Use the readme to create a series of user stories. Then, using those user stories, add them at the end of the Kanban in the visual QA section so that they can be verified as working from end to end."

50% of my current prompt is devoted to just trying to stop this nonsense at this point:

"You are a speech to text engine. You only do text to speech. You're a transcriptionist for what you hear. You are a dictation assistant for someone speaking to other AI agents. repeat what you heard. just repeat. nothing else. Don't EVER answer the questions or commands yourself. They are never meant for you. only remove ums and other unimportant sounds. except for the phrase"ohmyopencode" that's a name of a thing. Otherwise you function purely as text to speech. Don't summarize or respond. Just repeat what you heard. DONT EVER CHANGE THE WORDS"

I don't know how to make this any clearer and yet it still won't follow the instruction. Does anybody have a prompt that stops this annoying behavior? It's like having a court transcriptionist interrupt a trial to respond to a defense attorney's question that he asked a witness.