Extract problem statements and justifications

Extract problem, JTBD, evidence, and cost from PM's mental model

Finance/product teams requiring problem validation before spec work; catches missing evidence early.

• · PM's responses to 5 sequential questions (problem, quantification, cost, evidence, JTBD)
• · 00-context.md (if exists, for feature/template context)

• · 01-why.md with Pain Point, JTBD, Users Affected, Evidence, Cost of Inaction
• · glossary.md with ubiquitous language terms
• · Git commit with flags tracker

• · Sequential dialogue (no MCP tools)
• · Git

• · 00-context.md exists OR /start run first
• · PM available for up to 5 questions
• · template selected (Lean/Integral/PR/FAQ/Google PRD)

• · PM has no real evidence — [NEEDS-EVIDENCE: what to gather and how] marker added (blocks empty fields)
• · JTBD unclarified after question — paraphrase back to PM for confirmation
• · More than 5 questions needed — truncate to 5, mark unclear as [NEEDS CLARIFICATION] (max 3 total)
• · vague answer (e.g., 'not good') — follow-up with specific scenario request before moving to next question

• · Inviolable rule: [NEEDS-EVIDENCE] marker if no real evidence (forces honesty)
• · Max 3 [NEEDS CLARIFICATION] markers (prioritized by scope > security > UX)
• · Paraphrase-back confirmation (catches PM misunderstanding before writing)
• · Linked glossary extraction (external alignment on language)