Your question is Self-Correcting Loop Prevention. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
Imagine a scenario where an LLM agent gets stuck in an infinite loop of API calls. How would you design a self-correcting mechanism to prevent this in Mirakl agent workflows?