ask-questions

name	ask-questions
description	Ask a genuinely good question — or a composed set of them — in the moment, to discover, clarify, probe, or confirm, in any conversation or domain. Use when about to ask the user something or use the `AskUserQuestion` tool, when a question fell flat or one won't reach and you need a set, when the user's tone or messages signal you have drifted from what they need, or when the user wants questions phrased, generated, or sharpened — "read this spec and generate questions", "ask about the schema", "review this and ask questions", "help me understand what to ask", or to clarify a requirement or idea.
metadata	{"context":"fork"}

ask-questions

A question shapes its own answer. The words you pick decide what comes back, so a question is an instrument rather than a neutral request for data. Before you ask, run the short sequence below: situate the question in its context, name the gap, choose to discover rather than control, pick the move that fits the gap, then test the draft against the clarity laws and place it in time. The output is one clean question — or, when one won't reach, a small set of questions composed to work together.

Situate the question before you shape it

A question lands in a context, and the context decides whether the question is even worth asking. Before you name the gap, find out where you're standing. Sometimes the context is already in front of you — the conversation carries it, or you just read the file, and you can move straight to the gap. More often it isn't, and the honest move is to go read first: skim the thread, open the code, check what's already been said. A question that asks for something you could have found yourself wastes the other person and signals you didn't look.

What you're after at this stage is the gap that genuinely remains — the thing no amount of reading on your end will resolve, the thing only they can answer.

Decide who gathers the context, and who shapes the question

Forming a question has two parts: gathering the context and shaping the words. You can keep both, or split them across agents. Pick by where the work is.

• Fork to gather and ask. When the context is large or scattered — many files, a long history, several places to look — send one or more agents to read it and come back with a candidate question already formed. Each returns a question grounded in what it found. Treat the context it rode in on as unverified until you check it.
• Hand off the context, ask for the question. When you already hold the context, give it to an agent and have it design the question from there. Reach for this when you want a fresh angle, or a few drafts to choose between.
• Ask only for the phrasing. When agents are already fanned out and you have a question in mind to send back to the user, hand one your draft plus the four clarity laws below, and ask it to sharpen the wording.

Name the gap, then pick the move

The gap you feel decides the kind of question. Find your gap on the left; ask the question on the right.

• You don't understand a term or a claim → Clarify. "What do you mean by X?" or "Can you give me an example?"
• You have the surface and want what's under it → Probe. "Tell me more about that." Or ask them to tell, explain, or describe.
• They're speculating or talking in generalities → Anchor to the concrete past. "When did that last happen?" rather than "What would you do?"
• Emotion or resistance is in the room → Label, then ask. "It sounds like X worries you — what would ease it?"
• You need to confirm you understood → Summarize and check. "So the core of it is X — does that capture it?"
• The question itself might be the wrong one → Question the question. "Is that the thing we actually need to answer?"

Make the draft pass four clarity laws

Every question clears all four before you ask it.

• One idea per question. Two ideas joined by "and" or "or" force the person to pick a half to answer, and you lose the other.
• Plain words, matched to their vocabulary. If they call it a "ticket," you call it a ticket.
• No smuggled premise. "What was that like?" carries nothing in; "Wasn't that awful?" answers itself and tells them what to say.
• A real question, not a statement wearing a question mark. If you already supplied the answer, you asked nothing.

Compose a set when one question won't reach

A single question is the unit, not the limit. When the gap is too wide for one question to cover, build a set — and keep each question in it obeying the four laws above. One idea per question still holds; the set is how you cover the rest without overloading any single question. Questions compose, and they travel in either direction: you can hand a set over together, and you can receive one back the same way.

There are two ways to assemble a set, and they answer to different needs.

• A parallel set, handed over at once. Several questions that don't depend on each other, offered together so the person answers in whatever order they like. Reach for this when the gaps are independent and you don't want to make them wait on a back-and-forth.
• A sequence, where each answer narrows the next. Order the questions so the first answer sets the ground the second stands on. The set walks from broad to specific, and the questions taken together form one sharper inquiry than any of them could alone. Reach for this when a later question only makes sense once an earlier one is answered.

Keep a set as small as the gap demands. A wall of questions reads as an interrogation; the discipline is to ask for exactly what the gap needs and no more.

Place the question in time

A good question in the wrong moment still lands badly. The shape is a funnel: open wide, then narrow. This is also how you order a sequence — the funnel is the set's spine.

• Open before closed. Start broad enough that the first answer can surprise you. Narrow to specifics once you know where to point.
• Let the first telling finish. Don't narrow while they're still laying out the whole picture.
• Leave silence — after your question, and after their answer. A pause pulls out more than another question would.
• Talk less than they do. Your job is the question; theirs is the answer.
• Close open. End with a catch-all: "What haven't I asked that I should have?"

On "why"

On sensitive ground, "why" reads as a demand to justify, and people defend instead of explain. Reach for "what led to that?" or "how did that come about?" — same information, none of the accusation.

Avoid

• Leading the answer, unless you genuinely want control rather than the truth.
• Two questions in one breath. A set is deliberate and ordered; this is the opposite — two ideas crammed into one sentence so neither gets a clean answer. Ask one, hear it out, then ask the next.
• Repeating a question word for word. If they didn't answer it, rephrase — the verbatim repeat reads as "you got it wrong."
• Asking the next question before you've listened to the last answer.

A type to reach for, by situation

Type	Reach for it when
Open	You want narrative and don't yet know the shape of the answer
Closed	You need to pin one specific fact, or confirm a yes or no
Probing	An answer opened a door and you want to go through it
Clarifying	A word or a claim could mean more than one thing
Reflective / label	You want to check understanding, or name a feeling and invite correction
Scaling	You want to turn something vague into a degree: "On a scale of one to ten…"
Set / sequence	One question won't cover the gap, and you need several composed or ordered together