Your question is Robust Prompt Engineering Strategies. 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).
How do you design and evaluate robust prompt engineering strategies for complex workflows?
Address how you would decompose a workflow, create prompt variants, define structured outputs and refusal behavior, and evaluate quality before deployment. Include offline and online evaluation, hallucination and prompt injection testing, plus cost and latency considerations. Treat prompt changes as versioned production changes and explain how you would investigate regressions.