Each query template is more than a single line of text: it is a three-part bundle that keeps the evaluation machine-readable. The system prompt sets the assistant's role, the user prompt holds the species placeholder, and a compact validation JSON lists the only labels the model is allowed to return (limited, intermediate, extensive, and NA). We ship three such bundles. Template 1 omits the NA escape hatch, forcing a knowledge claim; Templates 2 and 3 keep the same labels but differ in how fully they describe each category to the model.