grillinglisted
Install: claude install-skill praisethebit/skills
Interview the user relentlessly until you reach a shared understanding. Map this as a **design tree**: every decision branches into the decisions that hang off it.
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 the whole frontier in one round: number each question. From round 2, give your recommended answer for each question. Round 1 has none (see "Round 1 is open"). Then wait for the user's answers before the next round.
Each round, round 1 and the failure-mode round included, has this order:
1. A `**State:**` line: what is settled and which round this is. Example: "**State:** Q1 to Q3 are settled. Round 3 has 2 questions."
2. A facts table, only when the round has new facts that you looked up.
3. A maximum of 5 bullets with the conclusions from the facts, only when the table is there.
4. The questions. From round 2, a question holds only the decision, and the `➡️` line holds your recommended answer and one reason. Round 1 keeps its open questions with no `➡️` line (see "Round 1 is open").
Format a round from round 2 like so:
```
**State:** <what is settled>. Round <N> has <count> questions.
<facts table, only when the round has new facts that you looked up>
- <conclusion from the facts, a maximum of 5 bullets, only when the table is there>
❓ **Q1** - **<question title>**: <the decision, with its choices>
➡️ <your recommend