The agent asks what it needs. You answer in a click.

When the Obvious agent needs information before starting meaningful work, it surfaces a Questions card in the chat window. Each question is formatted to match the kind of answer it needs — a yes/no toggle, a single-choice picker, a multiple-choice list, a 1–10 scale, an ordering interface, or a short text field. Recommended answers come pre-filled: the agent's best guess based on what it already knows. You accept, adjust, or skip the card entirely and let the agent proceed on its own assumptions.

Short text for open-ended responses. Textarea for longer context. Yes/no for binary decisions. Single-choice when exactly one option applies. Multiple-choice when several might. Numeric scale for ratings or priorities. Ordering for ranking items. Each type is chosen because it matches how people actually think about different kinds of decisions — not because it's comprehensive, but because it's fast.

Prompting is a learned skill. It creates a two-tier system: expert users get precisely calibrated results; everyone else gets something approximate that takes a revision round to fix what a better upfront question would have prevented. Questions Tool collapses that gap. The agent takes on the burden of knowing what it needs. You only have to answer.