Include a risk when there is a meaningful uncertainty that could change the recommendation.

You do not need to invent a generic risk simply to complete a memorized template.

Prioritize the answer and supporting evidence first, then add the most decision-relevant risk or next step if one exists.

Formulaic synthesis sounds weaker than judgment. The structure should serve the decision, not the other way around.

My recommendations got better when I stopped forcing a random 'risk' at the end.