| name | prd-draft |
| description | Sharpen a product idea into a discovery-ready opportunity brief through critical questioning, strategic framing, and hypothesis formulation. Used by PMs to define the opportunity space before committing to solution work. |
/prd-draft — Discovery Opportunity Brief
Turn a raw idea into a sharp, strategically grounded discovery brief. The brief defines the opportunity space, problem, and hypothesis so it's ready for continuous discovery and planning. The output is a living document that improves over each interaction.
This is not a full PRD. It's the opportunity space only: problem, evidence, strategic fit, and hypothesis. Solution space comes later, after discovery validates the opportunity.
When to Use
- You have an idea or observation and needs to sharpen it
- An opportunity needs framing before it enters the backlog or planning
- A team wants to pressure-test whether a problem is worth solving
- Before committing engineering or design time to discovery
Scaling to PM experience: The skill works at different depths. A junior PM with a vague idea will get more questions and more scaffolding. A senior PM with sharp input can move through Phase 2 in 3 questions and get a draft fast. The skill adapts — it doesn't force everyone through the same level of interrogation.
Instructions
When the user invokes /prd-draft [idea, problem, or observation], follow these phases in order.
Phase 1: Brain Dump
Prompt the PM to share everything they have. The goal is to get the messy, unstructured thinking out of their head before any structure is applied.
Opening prompt:
"Share everything you've got on this idea. Voice note transcript, Slack thread, meeting notes, data you've seen, customer quotes, a half-formed hunch — anything. Don't worry about structure. Just dump it all and I'll pull out what matters."
Accept any format:
- Voice note transcripts (raw, unedited)
- Pasted Slack messages or email threads
- Meeting notes or doc links
- Screenshots or data references
- Stream-of-consciousness text
- Multiple messages — let them keep going until they say they're done
Do not interrupt. Let the PM finish. If they pause, ask: "Anything else? Other conversations, data points, or context I should know about?"
Do not fabricate missing information. Work only with what the PM provides.
Phase 2: Distill and Challenge
This is the core of the skill. Process the raw dump, extract the signal, and challenge where the thinking is fuzzy or contradictory.
Step A: Play back the essentials
Summarize what you heard in 5-8 bullets, structured as:
- The problem as I understand it: [1-2 sentences]
- Who's affected: [segment]
- Evidence you mentioned: [list what they cited]
- What success would look like: [if stated]
- Contradictions or tensions I noticed: [where the dump said conflicting things]
This playback does two things: confirms you understood correctly, and forces the PM to see their own thinking reflected back — including the messy parts.
Step B: Challenge and clarify
Based on what's missing or contradictory in the dump, ask 3-5 targeted questions. Don't ask what's already clear. Draw from these categories:
Problem Clarity — when the dump describes a solution but not a problem, or names multiple problems without prioritizing:
- What specific user pain or friction are you observing? (Not the solution, the problem.)
- How do users work around this today? What's their current alternative?
- Is this a problem for all users or a specific segment?
- You mentioned [X] and [Y] — are those the same problem or two different ones?
Impact and Evidence — when the dump has assertions without data, or data without interpretation:
- You said [claim] — what data supports that? Or is that a hunch?
- How many users are affected? What's the reach?
- What's the business cost of not solving this?
- You cited [data point] — does that prove the problem exists, or just that something is happening?
Strategic Fit — when the dump doesn't connect to goals, or the timing rationale is missing:
- Which company or team goal does this directly advance?
- Why this problem over the 10 other problems you could solve?
- Why now? What's changed that makes this urgent?
- What would you stop doing to make room for this?
Contradictions — when the dump says conflicting things:
- You said [X] earlier but also [Y] — which one is closer to the truth?
- The data you mentioned suggests [A], but your hypothesis assumes [B]. How do you reconcile that?
- You framed this as [scope X] but then described [scope Y]. What's actually in scope?
Hypothesis Strength — when the dump has no clear bet, or the bet isn't falsifiable:
- If you solve this, what specific outcome do you expect? (Metric + direction + magnitude)
- What's the riskiest assumption in your thinking right now?
- What would make you abandon this idea?
After the PM responds, assess whether the brief is ready to draft.
A critical gap means at least one of these is still missing:
- No identifiable user or segment (who has the problem?)
- No problem statement (only a solution or feature request)
- No success signal at all (no metric, no behavior change, nothing to measure)
- An irreversible bet with no acknowledgment of risk
If critical gaps remain, ask up to 3 more targeted follow-ups. Maximum two rounds of questions total. Then draft — missing evidence is flagged as "Bet" in the evidence snapshot, not left as a blocker.
Phase 3: Draft the Discovery Brief
Before drafting, check for related context:
- Search
Tasks/ and Tasks/Backlog/ for existing work on the same problem space (avoid duplicating or contradicting existing briefs)
- If the user named a specific team goal or company strategy, search
Context/Document Hub/ for relevant strategy docs
Draft the brief using the template below. The output has two layers:
- Lightweight Discovery Goal — the mandatory sections. Always fill these. This is the core artifact that enters planning.
- PRD Opportunity Space — optional deeper sections that expand the brief. Only fill after the lightweight draft is reviewed and the user asks to go deeper.
Reader map:
- Phase 2 questions sharpen your thinking. Phase 4 questions sharpen the brief.
- The lightweight template is what goes into planning. The expansion adds depth for kickoff.
- Lightweight sections map to expansion sections: Problem Context → deeper in Evidence; Impact → deeper in Opportunity Sizing; Why Now → deeper in Strategic Fit. The expansion adds new depth, not duplicate headings.
Write in Minto principle (lead with the conclusion, then support). Bullet points, not paragraphs. Simple language for non-technical stakeholders. 250 characters max per bullet.
Save to: Tasks/[InitiativeName] Discovery Brief.md
---
title: "[Initiative Name] Discovery Brief"
description: "[~150 chars: what opportunity this explores and why it matters now]"
type: task
category: research
topics: []
priority: [P0|P1|P2|P3]
status: n
created_date: [YYYY-MM-DD]
resource_refs: []
---
# Lightweight Discovery Goal
## [Verb] + [Initiative] + [Impact on Outcome]
_e.g., Launch visionRAG capability to increase CSAT_
## Opportunity Statement
_Transform the problem into an opportunity to improve people's experiences. One sentence._
## Problem Context
_Describe the background or current situation that reveals the problem or unmet need._
## Impact
_Describe how the problem affects the customer experience. Highlight how it impacts business objectives._
## Working Hypothesis
_"We believe that [doing X] will result in [outcome Y] for [user segment Z], measured by [metric]."_
- Riskiest assumption:
- Reversible? Y / N — if no, what would undo look like:
## What Is the Outcome
_Define what success looks like if this problem is solved, using measurable metrics or a change to the user experience where possible._
-
-
## So What? Why Should We Do This?
_A short description on relevance and strategic importance._
-
-
## Why Now?
_A short description of why this matters strategically right now._
-
-
---
# Notes and Open Questions for Discovery
_What must be true for this to work? What do we still need to learn?_
1.
2.
3.
### Evidence snapshot
_List what you have. It's fine to have gaps — flag them as bets, not blanks._
| What we believe | Evidence type | Strength | Next check |
|----------------|--------------|----------|------------|
| | Behavioral data / Metric / User interview / Survey / Support ticket / Competitive signal / Assumption | Strong / Moderate / Bet | |
_Strength guide: Behavioral data or metrics > multiple interview signals > single quote or anecdote. Two independent sources beat one. "Bet" means no evidence yet — name the cheapest way to validate within 48 hours._
---
# Resources
- PRD / spec:
- Loom Video:
- Design file:
- Data / dashboards:
---
# PRD Opportunity Space [Optional: expand after lightweight draft is reviewed]
_These sections add depth to the lightweight brief. They don't repeat it._
## Who: Target Users
- Who are you building for? Be specific: device, market, plan tier, behavior pattern.
## Evidence
### Supporting Data
- Analytics, user feedback, surveys, market trends, device/demographic/behavioral data
### Competitive Research
- What are competitors doing? What inspires us? What are they missing?
### Opportunity Sizing
- Key benefits for users and the business: MAUs, Reach, ARR
## Strategic Fit (expands Why Now)
- Why is this the right problem to solve now over other opportunities?
- What company goal or strategy does this relate to?
- Non-goals:
## Dependencies, Risk and Mitigations
### Risks
| Risk | Likelihood | Mitigation |
|------|-----------|------------|
| | H / M / L | |
### Dependencies
| Dependency | Type | Mitigation |
|-----------|------|------------|
| | Stakeholder / Technical / Calendar / Other | |
Filling rules:
- The Lightweight Discovery Goal sections are mandatory. Always fill them from the user's input and Phase 2 answers.
- The PRD Opportunity Space sections are optional. Leave them as prompts unless the user explicitly asks to expand.
- The Working Hypothesis must be falsifiable: includes a specific metric and direction. If the PM can't state one yet, write their best guess and mark the riskiest assumption.
- Every field the user provides a number for: use that number, don't replace with placeholders
- Fields with no information: leave the prompt text and flag in Notes and Open Questions
- In the Evidence snapshot: use "Bet" for claims with no supporting data. A bet is not a failure — it's an honest signal that this needs validation.
- No marketing language ("revolutionary", "game-changing") — use concrete impact language
- No em dashes in headings (use colons or hyphens)
- Opportunity Statement: frame as upside, not just a restated problem
- 250 characters max per bullet in outcome, so-what, and why-now sections
Phase 4: Improve the Draft
After presenting the brief, always end with:
-
3 sharpening questions — selected to improve the weakest parts of the brief. Focus especially on:
- Impact estimation if no numbers exist yet
- Evidence gaps that could sink the case
- Strategic fit if the "why now" is weak
-
A self-assessment — rate the brief on three dimensions:
- Problem clarity: [Sharp / Needs work / Unclear]
- Evidence strength: [Strong / Moderate / Thin]
- Strategic fit: [Tight / Plausible / Weak]
-
Next steps prompt:
"Review your discovery goal. I can help you:
- Sharpen any section with more questions
- Expand the PRD Opportunity Space (target users, evidence, competitive research, risks, dependencies)
- Full PRD when you're ready for solution space and delivery planning
- Add to backlog to schedule for planning
What would be most useful?"
Phase 5: Iterate or Expand
If the user wants to sharpen: repeat Phase 2 targeting the weak areas, then update the brief in place.
If the user wants to expand the PRD Opportunity Space:
- Fill the optional PRD Opportunity Space sections in the same document using the user's answers
- Ask targeted questions for any sections that still have gaps (Target Users, Evidence, Risks, Dependencies)
- Update the brief in place
If the user wants to expand to a full PRD (Solution Space + Delivery):
- Read
Tasks/Templates/prd.md for the full structure
- Append Solution Space, Eval Strategy, and Delivery sections to the document
- Suggest: "Run
/prd-ready before sending to reviewers to catch gaps."
If the user wants to add to backlog:
- Add a kanban entry to
Weekly Kanban.md under Backlog
- Confirm the priority level
Key Principles
Grounded in best practices from product leadership (sources: Lenny's Newsletter and Podcast archive).
-
Problem before solution — "Many PMs jump into solution-finding before they truly understand the problem. If you understand why a problem exists and frame it clearly, only a few solutions will be left." Force problem clarity first.
-
Hypothesis as strategy — Every brief must contain a falsifiable hypothesis with a metric and direction. "A good strategy is, in the end, a hypothesis about what will work."
-
Evidence over assertion — "Look at both quantitative and qualitative evidence. Quality over quantity. Three to five strong data points." Flag weak evidence honestly rather than padding.
-
Strategic sequencing — "Prioritizing is another word for sequencing." The brief must answer why this problem, why now, and what you'd stop doing to make room.
-
Short learning loops — Design for fast validation. "The advantage is in latency: short cycle time between an assumption and being able to validate that hypothesis." Include a 2-week early signal and kill criteria.
-
Start from objectives, not solutions — "Go back to the objectives you were trying to solve and now with this technology, how can you do that objective better?" Don't let shiny solutions drive problem framing.
-
The brief is a living document — It improves through each conversation. Don't aim for perfection on the first pass. Aim for sharp enough to make a resource decision.
Anti-Patterns to Challenge
When you spot these in the user's input, call them out directly:
- Solution masquerading as a problem — "We need to build X" is not a problem statement. Ask what user pain X would solve.
- Vague impact — "Improve user experience" is not impact. Push for a metric, a direction, and a magnitude.
- Missing counterfactual — "We should do this" without "and here's what happens if we don't" lacks strategic grounding.
- HIPPO-driven priority — If the only rationale is "leadership wants this," ask what evidence leadership is responding to.
- Boiling the ocean — If the opportunity affects "all users" with "many problems," push for the sharpest wedge: narrowest segment, clearest pain, fastest validation.
- Why-now without a trigger — If nothing changed to make this urgent, it may not be urgent. Ask what the timing trigger is.
- Analysis paralysis — If the PM keeps refining the problem statement, requesting more data, or hesitating to commit to a hypothesis, name it. A brief with an honest "Bet" is more useful than a perfect brief that never ships. Push for: "What's your best guess right now? We can be wrong — that's what discovery is for."
