Phase 2 - Discovery
/pm-personas▊Four parallel research tracks that converge into a Problem Validation Summary. The goal is to understand the market, technology, domain, and customer deeply enough to validate (or invalidate) the problem worth solving.
Duration: 1-3 weeks · Gate type: Soft gate (gaps acknowledged, can proceed) · Playbooks: Greenfield
How to read this page
The four tracks run in parallel - there's no forced order between them, only within each. Per-track blocks below give the trigger, what to bring, and what you get. "Bring your data" rule applies everywhere here: Claude structures and formalizes what you bring - it does not hallucinate market data or interview insights. Every skill degrades gracefully if you have no data (marks output [ASSUMED - validate with real users] instead of blocking).
When to enter this phase
Enter Phase 2 after Phase 1 (or directly with /pureinn discover). Phase 3a can start in parallel with late Phase 2, once the Problem Validation Summary exists.
What you need before entering: Phase 1 artifacts (Project Charter, basic scope), and external research material per track - or willingness to proceed assumption-based.
Track A - Technology
- When to run / skip: run for every product - technical feasibility is never assumed. Never skip.
- Gather first: research on stack options, third-party APIs, technical constraints (Perplexity or your Tech Lead).
- Command:
/pm-tech-feasibility - What you get: Tech Feasibility Report - stack viability, key constraints, risks.
- What it does NOT give you: a final architecture decision - that's
/common-groundlater, closer to build. - Done when: the report names a viable stack path and its risks, not just "yes it's feasible."
Track B - Domain
- When to run / skip: run for every product with real domain/regulatory complexity. Skip only if the domain is trivial (no regulation, no specialized business logic).
- Gather first: domain and regulatory research (Perplexity or domain expert).
- Command:
/pm-domain-analysis - What you get: Domain Analysis + Legal Requirements.
- What it does NOT give you: the entity model itself - that's Phase 4 (
pm-entity-registry). This is the research that informs it, not the register. - Done when: regulatory obligations relevant to the product are named, not just "check with legal later."
Track C - Market
- When to run / skip: run for every product. Never skip - even a rough TAM/SAM/SOM beats none.
- Gather first: competitor research - or use Path C (AI-powered web research via Perplexity/OpenAI, requires
OPENAI_API_KEYinpureinn-variables.md) if you have no prep. - Command:
/pm-market-analysis - What you get: Market Size (TAM/SAM/SOM), Competitor Analysis, SWOT.
- What it does NOT give you: a pricing decision - that comes later, in Phase 3b business model work.
- Done when: TAM/SAM/SOM has real numbers or explicit
[ASSUMED]markers, not blanks.
Track D - Customer
- When to run / skip: run for every product. Never skip - this is the track most other decisions depend on.
- Gather first: ≥10 customer interviews for
pm-personas(or proceed assumption-based).jtbd-buildingrequirespm-personasto exist first - do not run it before. - Commands:
/pm-personas # first
/jtbd-building # after personas exist
- What you get: Customer Segments, Personas, Early Adopters Profile (from
pm-personas); JTBD Analysis + Forces Diagram per persona (fromjtbd-building). - What it does NOT give you: a validated problem - JTBD tells you the job, not whether the market will pay to have it done. That's Phase 3a.
- Done when: at least one persona has a JTBD Forces Diagram with real (or clearly assumed) forces.
Convergence - Problem Validation Summary
- When to run / skip: run only after all four tracks (A-D) are complete. Do not run early - it synthesizes them, it doesn't stand alone.
- Gather first: Tracks A-D outputs.
- Command:
/pm-problem-validation - What you get: the Phase 2 exit artifact - synthesizes all four tracks into a single validated (or flagged-as-unvalidated) problem statement.
- What it does NOT give you: a GO/NO-GO verdict - that requires real-world experiments, which is Phase 3a.
- Done when: the Summary exists and Phase 3a can start from it.
Client discovery layer (commissioned builds)
When a client, sponsor, or exec commissioned the work, discovery gains a plane above the end users - the commissioner's business, trigger, benefit, budget, constraints, references. Same four tracks, added tooling:
- When to run / skip: run only for commissioned builds (someone other than you is the customer of this discovery). Skip for your own product.
| Step | Skill / Material | Output |
|---|---|---|
| Before each session | pm-discovery-interview | Session agenda targeting the biggest remaining gaps |
| After each session | /pm-meeting (Client Discovery type) | Structured findings - references classified directive/hypothesis, constraints, [CANDIDATE-BR] exceptions |
| Continuously | pm-discovery-report | Client-facing "what we heard, what we recommend" - incremental, re-runnable |
Client claims about their own users enter Track D tagged [CLIENT-ASSERTED] - input, not evidence. When the mandate is already given (the client decided to build), Phase 3a can be skipped: the Discovery Report feeds /pm-scope-brief directly (Phase 3b alternative to the PRD).
What you exit with
- Tech Feasibility Report - stack viability, key constraints, risks
- Domain Analysis - domain structure, regulatory requirements, legal landscape
- Market Analysis - TAM/SAM/SOM, competitor landscape, SWOT
- Customer Segments + Personas - who the customer is, what they need
- JTBD Analysis - Jobs-to-be-Done + Forces Diagram per persona
- Problem Validation Summary - synthesizes all tracks; Phase 2 exit artifact
Phase exit gate
Run /pureinn after completing Phase 2 to check the exit gate.
Gate type: Soft. The engine checks all four tracks and the Problem Validation Summary. You can acknowledge gaps and proceed, or loop back to strengthen weak tracks.