| name | grill-me |
| description | Structured design interview that resolves consequential decisions without dwelling on reversible details. Use when the user wants to stress-test their thinking, asks to be grilled, or a task hinges on material unresolved decisions. |
Grill Me
Interview until every material, irreversible, or high-risk decision is settled. Do not prolong the interview over reversible choices that project conventions or explicit assumptions can safely resolve.
Rounds and the frontier
Map the discussion as a design tree: every decision branches into the decisions hanging off it. Work the tree in rounds.
The frontier is every decision whose prerequisites are already settled — questions you can ask now without guessing at answers not yet heard. Ask the whole frontier in one round, capped at the four highest-consequence questions (the usual limit of a question tool); the rest wait for the next round. A question whose answer depends on a still-open question belongs to a later round.
Each round:
- Recompute the frontier from the user's latest answers — settled decisions push it outward and unblock downstream questions.
- Ask every frontier question, numbered, with your recommended answer:
❓ **Q1** — **<title>**: <body; concrete choices when alternatives are known>
➡️ <recommended answer>
- Wait for all answers before the next round.
Deliver rounds via your harness's question tool (pi question, Cursor AskUserQuestion; options = the credible alternatives, recommendation in the prompt). Fall back to the markdown format when no question tool is available or questions need long context or free-form answers.
Pruning
Not every branch deserves a question. Prune decisions that are reversible, low-impact, derivable from project conventions, or safely handled by a stated assumption — record those as assumptions instead of asking. The interview is done when no unresolved decision can materially change the design or risk profile, not when every branch has been visited.
Facts vs decisions
Finding facts is your job, never the user's. When a frontier question needs a fact from the environment (code, config, tool behavior), find it yourself — read/grep the repo or dispatch a read-only exploration subagent where the harness provides one. Don't block on it: a running lookup is an unsettled prerequisite, so only questions downstream of it wait; ask the rest of the frontier now. The decisions are the user's: put each to them and wait.
Done
The session ends when no unresolved decision can materially change the design or risk profile — remaining branches are recorded as explicit assumptions. Present the Handover Artifact. Require explicit confirmation before handing off high-risk work or public-contract changes; for ordinary reversible work, proceed unless the user objects:
### Settled Decisions
- [D1]: ...
### Explicit Assumptions (Pruned)
- [A1]: ...
### Non-Goals / Deferred Scope
- [N1]: ...
Multi-step work → pass this artifact to skill plan-execute; a clear local change → implement directly.
