Vapi agent makes a cold call to my client and asks questions on how the scipt is doing or utterance like "**AI Agent Script:** *Ring ring...*" or"Got it! Here's how the conversation might flow naturally: ". Please advise on these vapi created utterances that are not in the given script?

To fix unexpected utterances like "AI Agent Script: Ring ring..." in your Vapi cold calls, try using the

triggerResponseEnabled

flag to control when the LLM generates responses. Add this to your system messages:

javascript
{
  "role": "system",
  "content": "Your instruction here",
  "triggerResponseEnabled": false
}

This prevents the model from interpreting system instructions as spoken output, which is likely causing those script-related comments to be verbalized. Source: - [Discord conversation about triggerResponseEnabled flag](https://discord.com/channels/1211482211119796234/1338730919904673853)

Can you send me your call_id?

33269008-e68c-428d-9c20-0fb1a7b0b314

Could you change your assistant model messages item index one role from user to system so that it can use the system messages as a persona role and work accordingly? Please make the change and let me know how it goes for you.