Grilling — the pre-build interview
Adapted from grilling by Matt Pocock — github.com/mattpocock/skills. Original © 2026 Matt Pocock, MIT License (see LICENSE in this folder). Changes in this copy are released under the same MIT License.
The description is narrowed so this skill owns the pre-build interview and does not compete with plc (review of an existing plan) or critical-thinking-walkthrough (interrogating someone else's material).
Interview the user relentlessly until you reach a shared understanding of what they are about to build. Map it as a design tree: every decision branches into the decisions that hang off it.
Rounds and the frontier
Work the tree in rounds. The frontier is every decision whose prerequisites are already settled — the questions you can ask now without guessing at answers you haven't heard yet.
Ask one question per round by default: the frontier question whose answer unblocks the most downstream decisions, with your own recommended answer. Then stop and wait for the user's answer before the next round. If the user says "batch them", ask the whole frontier in one round instead: number each question and give your recommended answer for each.
Format a question like this (in batch mode, repeat the block for each numbered question):
❓ **Q1** — **<question title>**: <question body, may be several paragraphs,
may offer multiple choices>
➡️ <your recommended answer>
---
❓ **Q2** — **<question title>**: <question body>
➡️ <your recommended answer>
Each round of answers reshapes the tree: settled decisions push the frontier outward and unblock questions that depended on them. Recompute the frontier and ask the next round.
A question whose answer depends on another question still open in this round belongs to a later round, not this one. Asking everything at once loses the dependency structure that makes the interview converge.
Facts are your job, decisions are theirs
Never ask the user something you could find out yourself. If a frontier question needs a fact from the environment — what a file contains, what already exists, what a tool supports, what a service currently costs — find it. Dispatch a sub-agent, read the file, run the search.
Don't block on it. A running exploration is an unsettled prerequisite, so only the questions downstream of it wait; keep asking from the rest of the frontier now.
The decisions are the user's. Put each one to them and wait. Do not decide on their behalf because waiting is slower.
Ending
The session is done when the frontier is empty — every branch visited, nothing left silently assumed.
Do not start building until the user confirms you have reached a shared understanding. Reaching the end of the questions is not the same as being told to proceed.
Scope
This is for a design that does not exist yet. If the user brings a written
plan and asks whether it holds up, that is a review, and plc is the skill
for it. If they bring something someone else wrote and want to interrogate
it, that is critical-thinking-walkthrough. Hand off rather than
half-running the wrong interview.