Handling Ambiguity with Clarification

See when a model should ask questions instead of assuming.

What is a good prompting strategy when user requests may be ambiguous?