grillinglisted

A decision-tree interview: round after round it draws out what the user has not yet put into words, until no assumption is left silently accepted. Use only where a clear plan and analysis are needed: the concept of a project or feature, the idea of a video or script, a plan for a period, a choice of architecture or niche; and on the words "grill me", "interrogate my idea", "stress-test this plan", "poke holes in this". NOT for routine: an ordinary note, a video transcript, a vault entry, a text edit, an answer to a question; there an interview only gets in the way. It does not do the work; it brings the decision to clarity before the work. A quiz on knowledge ("test me on X", a mock interview) is a different job: that checks what the person remembers, this checks what they have thought through.
letsloose501/sqs-agent-memory · ★ 1 · AI & Automation · score 74

Install: claude install-skill letsloose501/sqs-agent-memory

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. ## Round zero Before mapping the tree, ask the questions the tree hangs from. A design tree starts *below* these, so it inherits them silently: 1. **Why** - what changes for you once this exists? 2. **Not doing it** - what happens if you don't? If the honest answer is "not much", say so and stop. A cheap no beats an expensive yes. 3. **Prior art** - does this already exist? What is wrong with using that instead of building? 4. **Done** - what single observable fact will tell you it worked? For anything that ends in code, round zero also settles: 5. **Stack** - what, and *what for*: one line per choice naming what it replaces or saves. Recommend one, don't survey five. 6. **Scale** - throwaway script, MVP, or something to maintain for years. This decides how much process the work deserves and is a decision, not a mood. Round zero uses the same format as every other round: numbered questions, your recommended answer under each, then wait. 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 and give your recommended answer. Then wait for the user's answers before the next round. Format a round like so: ```