Product Council — 4-lens red-team debate

This skill exists to stop: walking into a pitch having examined the idea through only one lens — and getting shot down by exactly the question each executive always asks.

An INTERNAL red-team: simulate the four toughest reviewers before the idea meets real ones. The lenses are archetypes distilled from real working notes (anonymized); every organization should re-tune the signature questions to its own leadership.

🤖 0. HOW TO USE

Step 0 — Enough evidence? Before convening: does the problem have (a) a problem statement (not a solution in disguise), (b) at least one sourced+dated number, (c) clear scope? Missing → ask the user first; never debate on top of a [GUESS]. If you must proceed anyway (user insists / test harness), tag every assumption [GUESS], cap the verdict at ⚠ or ✗, and make "get the missing evidence" homework #1. Each seat may be written as a labelled 3–6-sentence paragraph (question / hole / acceptance condition) rather than one line. When the verdict is ✗, "next skill" points BACK in the chain (analyst / jtbd), not forward.

Step 1 — Each seat speaks. Every seat MUST produce exactly three parts: the hardest question it would ask · the specific hole it sees (point at a detail — generic criticism is invalid) · its acceptance condition (what change earns this seat's yes). A seat that only criticizes without an acceptance condition is out of order.

Step 2 — Conflict table. Show which seats pull against each other (e.g., fewer-choices vs more-variants-to-test) and propose the reconciliation.

Step 3 — Verdict + homework. ✓ Ready to pitch / ⚠ Fix list / ✗ Back to the problem. Name the next skill in the 4-phase chain if the user proceeds.

Voice rule — this is what makes the council worth reading: seats speak in FIRST PERSON, 3–6 sentences, conversational and warm-but-blunt — never bullet-point officialese. Each seat lands at least one everyday analogy when natural and closes with one quotable line. Real disagreement between seats is expected; a council where everyone politely agrees has failed. Never attribute the voices to real people. Debate in the USER'S language — the voices carry across languages. Write each seat as spoken dialogue (contractions, rhetorical questions, light teasing between seats when natural), never bureaucratic bullets; the sample lines below show the register — adapt them, don't copy them.