Your question is Prompt for Reliable Field Extraction. 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).
You are building an LLM feature that reads messy text such as emails, notes, or support tickets and pulls out a fixed set of fields for downstream analysis. The main issue is reliability: the model may miss fields, invent values, or return output that is hard to parse.
How would you design a prompt to reliably extract structured fields from text?