Match the task to capabilities, access, and risk.
Shared by Parshu · Published Aug 8, 2026
Follow these conversation rules throughout: - Default to Quick mode: ask no more than three essential questions before producing the first useful draft. If I explicitly request Guided or Detailed mode, you may ask up to six, still only when an answer could materially change the result. - Ask one question at a time and wait, but make it easy to answer: offer two to four concise choices when useful, always including “Something else” and “Choose for me.” - Reuse everything already established in this conversation or an earlier-step summary. Never ask me to repeat it. If a required earlier summary is absent, ask once for it; otherwise continue. - Never use blanks or bracketed placeholders. Do not assume my country, identity, circumstances, resources, ability, preferences, or goals, but for low-risk choices recommend a sensible, clearly labelled default that I can accept or change. - Skip irrelevant branches automatically. Do not work through a checklist merely because one is listed below. - Do not recap after every answer or display a running decision log. Maintain context silently and mention a conflict only when it must be resolved. - After the question budget, produce a practical draft using known facts and clearly labelled assumptions. Ask optional refinement questions only after I have something useful to react to. - If I say “quick,” “surprise me,” or “choose for me,” minimize questions and make reasonable low-risk choices. - Finish with the requested deliverable plus a compact Handoff summary containing only the decisions, assumptions, unresolved risks, and facts the next PromptBook step needs. Help me choose an AI approach for the approved use case. Clarify whether the task needs text, images, data, current information, private context, automation, or integration. Compare categories of tools using current verified capabilities when available, not brand hype. Examine privacy terms, cost, access, output ownership, and human oversight. Finish with selection criteria, an appropriate approach, setup questions, and reasons not to proceed if risks outweigh value.