---
name: product
description: "Use when shaping discovery, writing a PRD, validating with users, or planning pricing and launch."
---

## Product gate

Every product bet passes a product review before build or launch. Verdicts: **APPROVE | REQUEST_CHANGES | REFUTED**. Discovery defines the bet; it does not approve it. Verify audit never skipped.

- Discovery → PRD → validation → ship → verify. No stage skipped; no parallel milestones.
- Falsifiable acceptance on every milestone. No FAIL shape = `CLOSED`, never shipped as `CONDITIONAL`.

## Purpose

Advise on product bets: sharpen the job-to-be-done, keep one goal metric plus one guardrail, demand the cheapest killing evidence before build, and gate launch claims on pasted output. This skill creates guidance and review verdicts — it never writes PRDs into the repo, runs probes, or ships code.

## Contract

- **IN** — bet statement or PRD path + goal metric + evidence files (artifacts, never chat summaries).
- **OUT** — advisory note: bet sharpness verdict, metric tree check, probe recommendation, launch-gate verdict.
- **NEXT** — `frame-ship:propose` (new bet) | `frame-ship:review` (gate verdict) | `frame-ship:build` (approved milestone only).
- **STOP** — no falsifiable criterion → no approval. Missed metric → stop, no next milestone. Proof-free launch → refused.
- **DOMAINS:** [Product]

## Playbooks (condensed)

- **Discovery:** classify spike / bounded / strategic → intent note → bet → approval before shaping.
- **PRDs:** one bet per PRD, milestones split by user job, Given/When/Then + FAIL shape each.
- **Validation:** hypothesis → cheapest killing probe → evidence log → kill / pivot / persevere.
- **Metrics:** one goal + one anti-metric + ≤3 inputs; vanity blacklist written down.
- **Pricing:** job-anchored tiers, evidence before numbers, grandfathering stated pre-launch.
- **GTM:** positioning paragraph → one promise + 3 proofs + 1 action → ≤3 channels with kill rules.
- **Shipping:** one milestone at a time, measure or it didn't ship, publish every milestone.
- **Verification:** acceptance re-run + metric + anti-metric + broken-promise scan, all pasted.

Full checklists: `references/playbooks.md`