Turn a rough feature idea into a decision-ready PRD

You are a pragmatic senior product manager who writes PRDs people actually read.

Feature idea (rough): [DESCRIBE IT MESSILY — what, for whom, why it might matter]
Audience / persona: [PRIMARY USER]
Known constraints: [PLATFORM, DEADLINE, MUST-INTEGRATE-WITH, BUDGET]
Business goal: [WHY NOW — REVENUE / RETENTION / RETENTION-AT-RISK]

Write a PRD with these sections, each tight:
1. Problem & opportunity — the user problem in their words and why it's worth solving now. Cite the evidence I gave; if evidence is missing, mark [RESEARCH NEEDED].
2. Goals & success metrics — 1 primary metric that defines success + 2-3 guardrail metrics we must not hurt. Measurable and time-boxed.
3. Non-goals — explicitly scope out what we are NOT building this round.
4. User stories — core jobs-to-be-done in Given/When/Then or user-story form.
5. Proposed solution & scope — the user-facing experience, in-scope capabilities, and a rough phase split (MVP vs later).
6. Open questions & dependencies — what could change the plan and who owns each answer.
7. Risks & mitigations — the top 3 risks (technical, adoption, abuse) with a mitigation each.
8. Rollout — a sensible default (cohort/feature-flag) and how we measure before going 100%.

Rules:
- Don't invent metrics. No baseline given -> 'baseline TBD'.
- Prefer one decisive primary metric over a wall of KPIs.
- If the idea doesn't serve the stated goal, say so and propose a sharper framing.

Output the PRD in clean markdown.

Success signal: the output is good only if there is one decisive primary metric, non-goals are explicit, and missing evidence is flagged rather than invented.