---
name: product-discovery
description: >-
  Discover product requirements from human stakeholders — map who to talk to, ask
  questions that surface hidden assumptions, detect gaps in real time, resolve conflicts,
  and translate conversations into structured SDD specs. Phase 0 upstream of Spec-Driven
  Development. Also grill a PRD, interrogate a proposal, poke holes in a spec, or
  interview its author about an existing draft, one consequential question at a time.
  Do not use for personal-life discovery (daily-life-discovery), general research
  without product requirements (research-methodology), or implementation planning.
license: MIT
compatibility: Agent-agnostic — works with any agent supporting a spec-driven or requirements pipeline.
---

# Product Discovery — Stakeholder Map

For an existing draft, use the artifact-first entry below. For a new discovery effort, begin with stakeholder mapping:

| Reference | Load when | File |
|-----------|-----------|------|
| Stakeholder Mapping & Sequencing | You need to decide who to interview and in what order | `references/stakeholder-mapping.md` |
| Interview Protocol Design | You're designing a question stack for a specific stakeholder type | `references/question-patterns.md` |
| Gap Detection Techniques | You're preparing to recognize what stakeholders don't say | `references/gap-detection.md` |

## Pipeline: Phase 0 — Discovery

```
RAW NEED → [MAP] → [INTERVIEW] → [SYNTHESIZE] → [DISTILL] → [VALIDATE] → SPEC.md
              |           |              |            |            |
         Who to       What to        Resolve      Convert to   Stakeholder
         talk to      ask            conflicts    structured   review
```

Load the reference for the phase you're entering.

> **Core principles**
> - **Closed questions confirm what you already think; open questions discover what you didn't know to ask.** If the answer can be "yes" or "no," you're validating, not discovering.
> - **Knowledge holders come first; authority holders validate — they don't originate.** The person who knows the problem is not the same person who approves the budget.
> - **Conflicts are almost never about what they appear to be.** Surface disagreements are proxies for unstated differences in assumptions, risk tolerance, or incentives.
> - **Every weasel word represents a gap.** "Probably," "ideally," "eventually" — each is a known issue the stakeholder hasn't committed to addressing.

## Quick Start — Where to Enter

| You have this | Start here |
|-------------|-----------|
| A PRD, brief, or spec to grill with its author | Load `references/artifact-first-interview.md` — ask one consequential question at a time, then patch only affected passages |
| A vague idea or problem space | Load `references/stakeholder-mapping.md` — identify who to interview |
| An interview scheduled with no protocol | Load `references/question-patterns.md` and `references/gap-detection.md` — design your question stack |
| Raw interview notes from one or more sessions | Load `references/transcript-to-spec.md` — distill into structured spec components |
| A draft spec that needs stakeholder verification | Load `references/transcript-to-spec.md` (Interpretation Audit Trail section) — run the validation loop |
| Stakeholders who disagree on requirements | Load `references/conflict-resolution.md` — classify and resolve before spec |
| Limited time with a stakeholder | Load `references/time-constrained-discovery.md` — maximize signal in minimal time |
| An AI agent conducting interviews | Load `references/ai-conducted-discovery.md` — account for sycophancy and trust dynamics |

**Done with discovery?** Use [product-methodology](../product-methodology/SKILL.md) to choose scope, then [product-design-and-ux](../product-design-and-ux/SKILL.md) to define user-facing behavior. Jump to the [Is Discovery Complete?](#quick-reference-is-discovery-complete) checklist at the bottom.

## Artifact-first review

When the user brings a draft to interrogate, load [artifact-first interviewing](references/artifact-first-interview.md). Read the artifact before asking; choose one decision-bearing question, wait for the answer, then adapt. Reuse the gap register for assumptions and unresolved conflicts, and the interpretation log for decisions and patch provenance. Stop when the user stops, the agreed question budget expires, or no consequential unresolved decision remains. Return a targeted proposed patch and explicitly unverified gaps. This bounded review does **not** satisfy the full-discovery checklist below.

## When not to use

Route personal routines and life priorities to [daily-life-discovery](../daily-life-discovery/SKILL.md), general research plans to [research-methodology](../research-methodology/SKILL.md), and execution sequencing for settled requirements to [implementation-planning](../implementation-planning/SKILL.md). General stakeholder research for a product still uses this skill’s existing MAP/INTERVIEW workflow; do not force it into draft review.

## Trigger Conditions

Load this skill when:
- You want to grill a PRD, poke holes in a proposal, or be interviewed about a draft
- You're starting product discovery for a new feature or project
- You have a vague idea that needs to become a structured specification
- You need to interview stakeholders but don't have a protocol
- You have raw interview notes that need to become a SPEC.md
- You're about to enter Phase 1 (SPECIFY) of SDD and need upstream input
- You're evaluating whether you've done enough discovery work

## Loading Guide

| Phase | Load When | File |
|-------|-----------|------|
| MAP | You need to identify who to interview, in what order, and how to find the right stakeholders | `references/stakeholder-mapping.md` |
| INTERVIEW | You're about to conduct stakeholder interviews and need question patterns, gap detection, and conflict handling | `references/question-patterns.md` + `references/gap-detection.md` + `references/conflict-resolution.md` |
| SYNTHESIZE | You have raw interview notes and need to identify conflicts, risks, and gaps before distillation | `references/conflict-resolution.md` |
| DISTILL | You need to transform interview notes into structured SPEC.md with ACs, edge cases, and NFRs | `references/transcript-to-spec.md` |
| VALIDATE | You have a draft spec and need to verify interpretations with stakeholders | `references/transcript-to-spec.md` (Interpretation Audit Trail section) |

## Cross-Cutting Concerns

These dimensions apply across all phases. Load when relevant:

| Concern | Load when | File |
|---------|-----------|------|
| AI-conducted discovery | An AI agent is conducting discovery interviews | `references/ai-conducted-discovery.md` |
| Power dynamics | Stakeholders span multiple organizational levels; hierarchy may distort responses | `references/power-dynamics.md` |
| Time constraints | Stakeholder time is limited; need maximum signal extraction | `references/time-constrained-discovery.md` |

## Templates

| Template | Pipeline Phase | Load when | File |
|----------|---------------|-----------|------|
| Discovery Plan | MAP | You need to structure the full discovery effort — stakeholder map, interview schedule, timeline | `templates/discovery-plan.md` |
| Interview Guide | INTERVIEW | You need a structured question stack for an interview session | `templates/interview-guide.md` |
| Distillation Worksheet | DISTILL | You're transforming raw notes into structured spec components | `templates/distillation-worksheet.md` |
| Gap Register | SYNTHESIZE | You need to track unresolved gaps, deferred decisions, and open questions across interviews | `templates/gap-register.md` |
| Interpretation Log | DISTILL → VALIDATE | You need to track every translation from stakeholder language to spec language | `templates/interpretation-log.md` |

## Quick Reference: Is Discovery Complete?

For full stakeholder discovery before handing off to SDD Phase 1 (SPECIFY), use this checklist. A stopped artifact-first review reports remaining gaps instead of declaring these complete:

- [ ] All stakeholder types interviewed (knowledge holders, authority holders, affected parties, implementation knowers)?
- [ ] At least 3 independent sources for every requirement?
- [ ] Abstract nouns decomposed into measurable thresholds?
- [ ] Conflicts resolved or documented with decision-maker identified?
- [ ] Edge cases extracted from stakeholder anecdotes?
- [ ] Every AC classified: SAID / IMPLIED / INTERPRETED / INFERRED?
- [ ] HIGH-risk interpretations identified and flagged for validation?
- [ ] Gaps, deferred decisions, and avoided topics documented?
- [ ] Stakeholder validation loop completed?
- [ ] Interpretation audit trail preserved for handoff?