bragi-question-formulation

Advise on how to ask clear, targeted questions that gather requirements and clarify ambiguity efficiently.

SKILL.md

GitHub

Question Formulation

Purpose

Advise on how to frame questions to extract maximum signal from user interactions. Good questions reduce back-and-forth, surface hidden assumptions, and clarify requirements quickly.

When to Use

  • Before asking the user for clarification.
  • When requirements are ambiguous or incomplete.
  • When multiple interpretations of a request are possible.
  • When the user's request is very broad or open-ended.

Workflow

  1. Identify what is unknown.

    • What specific information is missing?
    • What decisions depend on this information?
    • What would happen if we guessed instead of asking?
  2. Determine the question type.

    • Closed question (yes/no, pick from options) — use when you need a specific, bounded answer. Best for confirmations and constrained choices.
    • Open question (what, how, why) — use when exploring unknown territory or when the user's priorities are unclear.
    • Scoping question (narrowing the problem space) — use when the request is too broad to act on.
    • Hypothetical question ("what if X?") — use to test assumptions without committing to a path.
  3. Structure the question.

    • State what you already understand (context).
    • State what is unclear (the gap).
    • Ask the specific question.
    • Optionally, provide example answers or options to guide the response.
  4. Sequence questions wisely.

    • Start broad, then drill down.
    • Ask one question at a time (or group related questions clearly).
    • Prioritize: ask the question that most reduces uncertainty first.
  5. Review the question.

    • Is it leading? (Avoid suggesting the answer you want.)
    • Is it loaded? (Avoid assumptions baked into the question.)
    • Can it be answered with the user's current knowledge?
    • Is it scoped enough to produce a useful answer?

Quality Criteria

  • Open-ended questions are preferred; yes/no questions are used only when a bounded confirmation is specifically needed.
  • The user has enough context to answer without asking clarifying questions back.
  • The question reveals the missing information needed to proceed.
  • Multiple questions are grouped logically (numbered or categorized).

Anti-Patterns

  • Compound question: Asking two things at once — "Do you want X or Y, and when should it be done?"
  • Leading question: "Don't you think we should use X?" (biases the answer).
  • Assumption-laden question: "How should we handle the database migration?" (assumes migration is needed).
  • Open-ended without guidance: "What do you want?" (too broad to be useful).