When working in workflows and using prompts for the AI agent to state something, they're initially responding to my prompt, for example "Yeah I can do that" or "Sounds good" and then proceeds to do as it's prompted. Is anyone familiar with resolving this kind of behavior?