Challenge the strongest material assumption or failure mode and recommend the highest-value next question or action.

Do not invent problems. If more analysis is unlikely to change the decision, say so.