### Answer Your agent is saying "NO_RESPONSE_NEEDED" out loud because your prompt is telling it to reply with that exact text in quotes [(1) ]() [(2) ]() . The issue is in how you're instructing the LLM. When you write something like:

- if user says "hold on", reply exactly the following: "NO_RESPONSE_NEEDED".

[(1) ]() The quotes cause the LLM to output it as spoken text [(2) ]() . Solution: Update your instructions to be more explicit about outputting ONLY the stop sequence without any additional text, punctuation, or formatting . Try this instead:

## Silence Rules
- If the user says "hold on", "give me a second", "wait", or similar, respond with ONLY this exact text with nothing before or after: NO_RESPONSE_NEEDED
- If you detect the user is still thinking, pausing, or not ready to respond, output ONLY: NO_RESPONSE_NEEDED
- When you output NO_RESPONSE_NEEDED, do not add ANY other text, punctuation, or explanation before or after it.
- NO_RESPONSE_NEEDED must be the complete and only content of your response.

The key is making it crystal clear that the LLM should output the stop sequence without quotes, without "Okay," before it, and without any formatting .

​ To learn more, see the sources I considered: [[1] Make agent hold and respond nothing](https://docs.retellai.com/build/no-response) [[2] no_response_needed](https://discord.com/channels/1186424376300679209/1433483896179261480) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer The issue is that your instruction "output: NO_RESPONSE_NEEDED" still frames it as something the agent should say [(1) ]() . The LLM interprets "output" as text to speak. According to the official documentation, you need to tell the agent to "reply exactly the following: 'NO_RESPONSE_NEEDED'" . The key word is "exactly" - this signals to the LLM that it should output the stop sequence precisely as written, with nothing added . Try this exact phrasing from the docs:

- if user says "hold on", reply exactly the following: "NO_RESPONSE_NEEDED".

The documentation shows this specific format works because

NO_RESPONSE_NEEDED

is a hard-coded stop sequence in the LLM . When this sequence is met, response generation stops . If you're using Conversational Flow instead of Single Prompt Agent, note that one Discord user reported this worked in Single Prompt but not in Conversational Flow [(2) ]() . They found that being even more explicit helped - emphasizing that NO_RESPONSE_NEEDED must be the complete and only content of the response, with nothing before or after it .

​ To learn more, see the sources I considered: [[1] Make agent hold and respond nothing](https://docs.retellai.com/build/no-response) [[2] no_response_needed](https://discord.com/channels/1186424376300679209/1433483896179261480) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.