code-ai-installer 4.0.0 → 4.0.1-a

package/domains/product/locales/en/.agents/skills/assumption-mapping/SKILL.md

@@ -1,307 +1,323 @@
----
-name: assumption-mapping
-description: Assumption map across value / usability / feasibility / viability axes — hypothesis prioritization for testing
----
-# Assumption Mapping
-
-> **Category:** Discovery  ·  **Slug:** `assumption-mapping`
-
-## When to Use
-
-- Before launching discovery or experimentation — to know which hypotheses to validate.
-- When reviewing new ideas from OST — which assumptions underpin this solution?
-- In decision reviews — which assumptions were made and validated / not validated?
-- As a risk instrument: high-uncertainty assumptions = focal risks.
-
-## Input
-
-| Field | Required | Description |
-|-------|:--------:|-------------|
-| Solution / initiative context | ✅ | What we plan to do |
-| Team brainstorm log | ✅ | Raw assumptions list |
-| Existing evidence | ⬚ | Support/refute for assumptions |
-
-## Data Sources
-
-1. Team brainstorm (minimum 5-10 participants ideal).
-2. Existing customer data (interviews, analytics).
-3. Market data (competitor behavior, industry reports).
-4. Technical docs (internal architecture constraints).
-
-### Related Skills
-
-| Skill | What we take | When to call |
-|-------|-------------|--------------|
-| `opportunity-solution-tree` | Solutions → assumptions | For each solution branch |
-| `hypothesis-template` | Assumption → testable hypothesis | For top-risky assumptions |
-| `user-interview-script` | Questions targeting assumptions | Validation via interviews |
-| `ab-test-design` | Experiment for high-risk assumption | For quantitative validation |
-
-## Structure: four quadrants
-
-1. **Value (Desirability)** — will customers want it?
-2. **Usability** — can they figure out how to use it?
-3. **Feasibility** — can we build it?
-4. **Viability** — does it make business sense (cost, scale, regulation)?
-
-Each assumption falls into one of the four.
-
-## Protocol
-
-### Step 0 — Brainstorm Raw Assumptions
-
-Workshop format:
-1. Present context (solution / initiative).
-2. Each participant individually writes assumptions (Post-it or digital).
-3. Target: 30-50 raw assumptions.
-4. No filtering at this step.
-
-Technique: "If [X] — then this works". Assumption = X.
-
-### Step 1 — Categorize into 4 Quadrants
-
-Distribute assumptions across Value / Usability / Feasibility / Viability.
-
-If it doesn't fit — rephrase or discard.
-
-### Step 2 — Risk Assessment per Assumption
-
-Two dimensions:
-
-**Uncertainty** — how unknown this is, true or not (1-5):
-- 1 = proven (data/experience)
-- 5 = complete unknown
-
-**Impact if wrong** — how bad if the assumption is false (1-5):
-- 1 = minor adjustment
-- 5 = kill the project
-
-Risk Score = Uncertainty × Impact. Top: Score ≥ 16.
-
-| # | Quadrant | Assumption | Uncertainty (1-5) | Impact (1-5) | Score |
-|---|----------|------------|:------------------:|:------------:|:-----:|
-| A1 | Value | Customers pay $X/seat | 4 | 5 | 20 |
-
-### Step 3 — 2×2 Matrix Visualization
-
-```
-Impact if wrong →
-High │  [Test NOW]    │  [Validated? If not, test]
-     │                │
-     │  [Document]    │  [Low priority]
-Low  │________________│________________
-      Low    ←       Uncertainty      →    High
-```
-
-Top-right quadrant = immediate testing priority.
-
-### Step 4 — Test Strategy per Top Assumption
-
-For each top-risky assumption — choose a validation method:
-
-| Method | Speed | Confidence | Cost |
-|--------|:-----:|:----------:|:----:|
-| User interviews | Fast | Low-medium | Low |
-| Prototype test | Medium | Medium-high | Medium |
-| Landing page test | Fast | Medium | Low |
-| A/B test on production | Slow | High | Medium |
-| Data analysis | Fast | Medium | Low |
-| Tech spike | Medium | High (feasibility only) | Medium |
-
-Match method to assumption type:
-- Value → interviews, landing page, smoke test
-- Usability → prototype test, usability test
-- Feasibility → tech spike, architecture review
-- Viability → data analysis, financial modeling
-
-### Step 5 — Test Plan
-
-For each top assumption — test card:
-- Assumption text
-- What we think will happen (prediction)
-- How we'll measure it
-- Success criteria (specific threshold)
-- Time box
-
-Via `$hypothesis-template`.
-
-### Step 6 — Iteration
-
-Assumption map — living document. After each validation:
-- Mark as ✅ validated / ❌ invalidated / ⚠️ partial
-- Update Risk Score (uncertainty goes down if data collected)
-- Promote new assumptions from surprises
-
-## Validation (Quality Gate)
-
-- [ ] ≥ 20 raw assumptions in brainstorm
-- [ ] All categorized in 4 quadrants
-- [ ] Each assumption has Uncertainty × Impact scoring
-- [ ] Top 3-5 high-risk assumptions identified
-- [ ] For each top assumption — test method + plan
-- [ ] Test cards via `$hypothesis-template`
-
-## Handoff
-
-The result is the input for:
-- **`hypothesis-template`** — top assumptions → formal hypotheses
-- **`ab-test-design`** — test method = A/B for high-certainty validation
-- **`user-interview-script`** — questions targeting specific assumptions
-- **PM**: top risks → PRD risk section
-
-Format: assumption map (table + 2×2 visualization) + test plan. Via `$handoff`.
-
-## Anti-patterns
-
-| Error | Why it's bad | How to do it right |
-|-------|-------------|-------------------|
-| Skip categorization | 4 quadrants give balanced risk view | Mandatory V/U/F/Vi |
-| Test low-risk first | Does not accelerate learning | Start with top-right (high U × high I) |
-| Not iterating | Stale map = useless | Update after each validation |
-| Over-testing feasibility | Engineers love tech spikes | Balance across quadrants |
-| Vague assumptions | "Customers will like it" | Specific: "Customers will pay $X/mo" |
-
-## Worked Example — TeamFlow Assumption Map (AI 1:1 Summarization)
-
-**Context:** TeamFlow (B2B SaaS HR-tech). After Discovery + OST the team planned work on AI summarization. Before engineering begins — assumption map session (workshop 90 min, 8 participants: PM, Eng Lead, Designer, CTO, VP Sales, CPO-customer-advocate, Data Lead, Legal counsel).
-
-### Raw Assumptions Brainstorm (30+ assumptions in 30 min)
-
-Categorized into 4 quadrants with Risk Score (Uncertainty × Impact, 1-5 each).
-
-#### Value (Desirability) — Will customers want this?
-
-| # | Assumption | Uncertainty | Impact | Risk | Status |
-|---|-----------|:-----------:|:------:|:----:|:------:|
-| V1 | Customers will pay **$8-12/seat/month premium** for AI summarization feature | 4 | 5 | **20** | 🔴 Critical |
-| V2 | End-user managers will **adopt** AI summarization at >60% active users within 90 days | 4 | 4 | **16** | 🔴 Critical |
-| V3 | Buyers (VP HR) see **aggregate dashboard** as tier-upgrade justification | 3 | 5 | **15** | 🔴 Critical |
-| V4 | Feature reduces manager churn signal ("tool helps me") by >10pp | 4 | 3 | 12 | 🟡 Major |
-| V5 | Competitors (Lattice, 15Five) **have not yet shipped** AI summarization in the next 6 months | 3 | 4 | 12 | 🟡 Major |
-| V6 | Customers prefer AI-in-TeamFlow vs standalone (Otter.ai etc.) — "privacy + integrated" | 2 | 4 | 8 | 🟢 Low |
-
-#### Usability — Can managers figure out how to use it?
-
-| # | Assumption | Uncertainty | Impact | Risk | Status |
-|---|-----------|:-----------:|:------:|:----:|:------:|
-| U1 | Managers will **trust AI summaries** enough to rely on them (not manually re-write) | 4 | 4 | **16** | 🔴 Critical |
-| U2 | First-time user clearly understands how to **enable** AI summarization (opt-in UX) | 2 | 3 | 6 | 🟢 Low |
-| U3 | Summaries written in **tone appropriate** for performance conversations (not too casual, not too corporate) | 3 | 3 | 9 | 🟡 Major |
-| U4 | Managers will **correct errors** via inline edit (not abandon feature) | 3 | 3 | 9 | 🟡 Major |
-| U5 | Mobile experience for on-the-go review / approve summaries acceptable | 3 | 2 | 6 | 🟢 Low |
-
-#### Feasibility — Can we build it?
-
-| # | Assumption | Uncertainty | Impact | Risk | Status |
-|---|-----------|:-----------:|:------:|:----:|:------:|
-| F1 | **LLM summarization quality** acceptable for HR conversations (no hallucinations, no misattributions) — >85% acceptable rate from managers | 4 | 5 | **20** | 🔴 Critical |
-| F2 | Streaming **transcription latency** acceptable (<3 sec lag) for real-time in-meeting use | 3 | 4 | 12 | 🟡 Major |
-| F3 | Engineering can ship MVP in 8-10 weeks with existing team (8 eng) | 3 | 3 | 9 | 🟡 Major |
-| F4 | **Data residency** requirements (EU, HIPAA in healthcare customers) solvable via regional LLM endpoints (Azure, Bedrock) | 3 | 4 | 12 | 🟡 Major |
-| F5 | **Cost per summary** economically viable at target price ($0.05-0.10 per 30-min meeting) | 3 | 4 | 12 | 🟡 Major |
-| F6 | Integration with existing TeamFlow 1:1 data model without major refactor | 2 | 3 | 6 | 🟢 Low |
-
-#### Viability — Does it make business sense?
-
-| # | Assumption | Uncertainty | Impact | Risk | Status |
-|---|-----------|:-----------:|:------:|:----:|:------:|
-| Vi1 | **Privacy / compliance** (GDPR, SOC2, CCPA) manageable via architecture choices (opt-in, user-controlled, no training on data) | 3 | 5 | **15** | 🔴 Critical |
-| Vi2 | **Legal risk** (wrongful termination suit referencing AI summary as evidence) not material — enterprise customers do not backlash | 3 | 4 | 12 | 🟡 Major |
-| Vi3 | Pricing premium **does not cannibalize** core tier ARR (existing customers upgrade, not swap) | 3 | 4 | 12 | 🟡 Major |
-| Vi4 | **GTM execution** ready (Sales trained, CS enabled, Marketing positioned) in target timeline | 3 | 3 | 9 | 🟡 Major |
-| Vi5 | Infrastructure + LLM API costs **scale economically** at 5,000+ active users | 3 | 3 | 9 | 🟡 Major |
-| Vi6 | Customer success in EU / regulated industries (healthcare, finance) **does not block** sales | 3 | 3 | 9 | 🟡 Major |
-
-### 2×2 Prioritization Matrix
-
-```
-Impact if wrong →
-High (5) │  V1 (20)  F1 (20)   │  [Validated: F6, U2]
-         │  U1 (16)  V2 (16)   │
-         │  V3 (15)  Vi1 (15)  │
-         │                     │
-         │  V4 (12)  F2 (12)   │
-         │  F4 (12)  F5 (12)   │  
-         │  Vi2 (12) Vi3 (12)  │
-         │  V5 (12)            │  
-         │                     │
-         │  U3 (9)  U4 (9)     │  [Document]
-Low (1)  │  F3 (9)  Vi4 (9)    │
-         │  Vi5 (9) Vi6 (9)    │
-         │                     │
-         │  V6 (8)  U5 (6)     │  [Low priority]
-         │  U2 (6)  F6 (6)     │
-         ──────────────────────→
-         Low (1)     Uncertainty     High (5)
-```
-
-### Top-Right Assumptions (test NOW) — Risk ≥ 15
-
-1. **V1** (Pay $8-12 premium, Risk 20) — biggest money question
-2. **F1** (LLM quality acceptable, Risk 20) — biggest technical question
-3. **U1** (Managers trust summaries, Risk 16) — adoption question
-4. **V2** (>60% adoption in 90 days, Risk 16) — usage scale
-5. **V3** (Dashboard tier upgrade, Risk 15) — buyer expansion
-6. **Vi1** (Privacy manageable, Risk 15) — existential risk
-
-### Test Strategy per Top Assumption
-
-| # | Assumption | Method | Duration | Cost | Confidence |
-|---|-----------|--------|:--------:|:----:|:----------:|
-| V1 | Pay $8-12 premium | Landing page test with pricing + "notify me" CTA + follow-up interview | 2 weeks | Low | Medium |
-| V1 | (secondary) | Commitment test: 20 customers sign LOI for beta at $10/seat | 3 weeks | Low | High |
-| F1 | LLM quality | **Wizard-of-Oz:** OpenAI API + manual review, 20 beta managers, blind quality rating | 4 weeks | Medium | High |
-| U1 | Managers trust | Observational: beta managers — do they edit summaries? How often? Interview delta | 4 weeks (concurrent with F1) | Low | Medium |
-| V2 | 60% adoption | A/B launch: 50% feature flag, measure 30-day adoption | 4 weeks (post-MVP) | Medium | High |
-| V3 | Dashboard tier upgrade | Concierge: offer 5 enterprise customers "preview" dashboard, measure willingness to upgrade tier | 3 weeks | Low | Medium |
-| Vi1 | Privacy architecture | **Spike** (engineering + legal): design + privacy review + counsel opinion | 2 weeks | Medium | High |
-
-### Test Plan per Top Assumption (through `$hypothesis-template`)
-
-#### Test Card — V1 (Willingness to Pay)
-
-```
-**We believe** that mid-market and enterprise TeamFlow customers
-**For** accounts with ≥10 active managers
-**Will pay** $8-12/seat/month premium for AI 1:1 summarization
-**We'll know it's true when** either:
-  (a) 30% visitors to landing page click "Pricing details" AND 15% of those click "Get beta access"
-  (b) 8 of 10 customer conversations confirm "we'd upgrade at that price"
-**Because** 6 of 8 Discovery interviews expressed pay-interest
-    + competitor pricing ($5-7 for ChatGPT Teams) suggests price elasticity exists
-    + existing TeamFlow seats price point $15-50/mo → 20-30% premium within precedent range.
-```
-
-#### Test Card — F1 (LLM Quality)
-
-```
-**We believe** that GPT-4 / Claude 3.5 level LLMs
-**For** 30-minute 1:1 performance conversations in English
-**Will generate summaries** with >85% acceptable rate from managers
-**We'll know it's true when** in Wizard-of-Oz test:
-  - Blind quality rating from managers: ≥4 out of 5 stars (average), across ≥100 meetings
-  - Hallucination rate (factually wrong info): <5% of summaries
-  - Misattribution rate (wrong person said X): <3%
-**Because** recent LLM benchmarks on summarization task show 85-90% acceptable rates
-    + our manually-crafted test prompts achieved 90% in internal QA
-    + similar-adjacent use cases (Fireflies.ai, Gong) report >80% customer satisfaction.
-```
-
-### Iteration Log (3 weeks later, hypothetical)
-
-**Status updates:**
-- **V1:** ✅ Validated. Landing page test: 38% click-through, 22% "Get beta access". 7 of 10 customer conversations confirmed $10/seat acceptable. 2 customers said "we'd pay $15 for enterprise features". → **Risk downgraded to 6**.
-- **F1:** ⚠️ Partial. Wizard-of-Oz: 78% quality acceptable (below 85% target). Hallucinations 8% (above 5% target). → Additional work needed: prompt engineering + human-review layer. Risk **still 15**.
-- **U1:** ✅ Validated. Beta managers edit summaries 35% of time (not abandon). Sentiment positive. → Risk 6.
-- **Vi1:** ✅ Validated. Privacy architecture approved by counsel (opt-in, no training on data, regional endpoints). Risk 3.
-- **V2 / V3:** Pending MVP launch.
-
-**New assumptions emerged:**
-- **NEW V7:** "Customers accept 24-hour summary delivery delay for enhanced quality" (Uncertainty 4, Impact 3). 
-- **NEW F7:** "Human review layer scalable to 5,000 summaries/day" (Uncertainty 4, Impact 4). Added to high-risk quadrant.
-
-### Promoted to PRD Risk Register
-
-Assumptions with lingering risk (F1 quality, F7 human review scale) → PRD Top 5 risks section + mitigation plan.
-
-> **Assumption mapping lesson:** Brainstorm with a cross-functional team (legal, sales, CS in addition to PM + eng) revealed **Vi1 (privacy)** as an existential risk that an engineering-only session would have missed. Landing page test for V1 cost $400 and gave high-signal validation in 2 weeks — saved 10 weeks of engineering investment in the wrong pricing model.
+---
+name: assumption-mapping
+description: Assumption map across value / usability / feasibility / viability axes — hypothesis prioritization for testing
+type: triggered
+domain: product
+owners:
+  - pm
+gates:
+  - PM
+tech: []
+topic: []
+triggers:
+  - assumption-mapping
+  - assumption mapping
+  - карта допущений
+  - гипотезы
+related: []
+budget_lines: 323
+schema_version: 1
+---
+# Assumption Mapping
+
+> **Category:** Discovery  ·  **Slug:** `assumption-mapping`
+
+## When to Use
+
+- Before launching discovery or experimentation — to know which hypotheses to validate.
+- When reviewing new ideas from OST — which assumptions underpin this solution?
+- In decision reviews — which assumptions were made and validated / not validated?
+- As a risk instrument: high-uncertainty assumptions = focal risks.
+
+## Input
+
+| Field | Required | Description |
+|-------|:--------:|-------------|
+| Solution / initiative context | ✅ | What we plan to do |
+| Team brainstorm log | ✅ | Raw assumptions list |
+| Existing evidence | ⬚ | Support/refute for assumptions |
+
+## Data Sources
+
+1. Team brainstorm (minimum 5-10 participants ideal).
+2. Existing customer data (interviews, analytics).
+3. Market data (competitor behavior, industry reports).
+4. Technical docs (internal architecture constraints).
+
+### Related Skills
+
+| Skill | What we take | When to call |
+|-------|-------------|--------------|
+| `opportunity-solution-tree` | Solutions → assumptions | For each solution branch |
+| `hypothesis-template` | Assumption → testable hypothesis | For top-risky assumptions |
+| `user-interview-script` | Questions targeting assumptions | Validation via interviews |
+| `ab-test-design` | Experiment for high-risk assumption | For quantitative validation |
+
+## Structure: four quadrants
+
+1. **Value (Desirability)** — will customers want it?
+2. **Usability** — can they figure out how to use it?
+3. **Feasibility** — can we build it?
+4. **Viability** — does it make business sense (cost, scale, regulation)?
+
+Each assumption falls into one of the four.
+
+## Protocol
+
+### Step 0 — Brainstorm Raw Assumptions
+
+Workshop format:
+1. Present context (solution / initiative).
+2. Each participant individually writes assumptions (Post-it or digital).
+3. Target: 30-50 raw assumptions.
+4. No filtering at this step.
+
+Technique: "If [X] — then this works". Assumption = X.
+
+### Step 1 — Categorize into 4 Quadrants
+
+Distribute assumptions across Value / Usability / Feasibility / Viability.
+
+If it doesn't fit — rephrase or discard.
+
+### Step 2 — Risk Assessment per Assumption
+
+Two dimensions:
+
+**Uncertainty** — how unknown this is, true or not (1-5):
+- 1 = proven (data/experience)
+- 5 = complete unknown
+
+**Impact if wrong** — how bad if the assumption is false (1-5):
+- 1 = minor adjustment
+- 5 = kill the project
+
+Risk Score = Uncertainty × Impact. Top: Score ≥ 16.
+
+| # | Quadrant | Assumption | Uncertainty (1-5) | Impact (1-5) | Score |
+|---|----------|------------|:------------------:|:------------:|:-----:|
+| A1 | Value | Customers pay $X/seat | 4 | 5 | 20 |
+
+### Step 3 — 2×2 Matrix Visualization
+
+```
+Impact if wrong →
+High │  [Test NOW]    │  [Validated? If not, test]
+     │                │
+     │  [Document]    │  [Low priority]
+Low  │________________│________________
+      Low    ←       Uncertainty      →    High
+```
+
+Top-right quadrant = immediate testing priority.
+
+### Step 4 — Test Strategy per Top Assumption
+
+For each top-risky assumption — choose a validation method:
+
+| Method | Speed | Confidence | Cost |
+|--------|:-----:|:----------:|:----:|
+| User interviews | Fast | Low-medium | Low |
+| Prototype test | Medium | Medium-high | Medium |
+| Landing page test | Fast | Medium | Low |
+| A/B test on production | Slow | High | Medium |
+| Data analysis | Fast | Medium | Low |
+| Tech spike | Medium | High (feasibility only) | Medium |
+
+Match method to assumption type:
+- Value → interviews, landing page, smoke test
+- Usability → prototype test, usability test
+- Feasibility → tech spike, architecture review
+- Viability → data analysis, financial modeling
+
+### Step 5 — Test Plan
+
+For each top assumption — test card:
+- Assumption text
+- What we think will happen (prediction)
+- How we'll measure it
+- Success criteria (specific threshold)
+- Time box
+
+Via `$hypothesis-template`.
+
+### Step 6 — Iteration
+
+Assumption map — living document. After each validation:
+- Mark as ✅ validated / ❌ invalidated / ⚠️ partial
+- Update Risk Score (uncertainty goes down if data collected)
+- Promote new assumptions from surprises
+
+## Validation (Quality Gate)
+
+- [ ] ≥ 20 raw assumptions in brainstorm
+- [ ] All categorized in 4 quadrants
+- [ ] Each assumption has Uncertainty × Impact scoring
+- [ ] Top 3-5 high-risk assumptions identified
+- [ ] For each top assumption — test method + plan
+- [ ] Test cards via `$hypothesis-template`
+
+## Handoff
+
+The result is the input for:
+- **`hypothesis-template`** — top assumptions → formal hypotheses
+- **`ab-test-design`** — test method = A/B for high-certainty validation
+- **`user-interview-script`** — questions targeting specific assumptions
+- **PM**: top risks → PRD risk section
+
+Format: assumption map (table + 2×2 visualization) + test plan. Via `$handoff`.
+
+## Anti-patterns
+
+| Error | Why it's bad | How to do it right |
+|-------|-------------|-------------------|
+| Skip categorization | 4 quadrants give balanced risk view | Mandatory V/U/F/Vi |
+| Test low-risk first | Does not accelerate learning | Start with top-right (high U × high I) |
+| Not iterating | Stale map = useless | Update after each validation |
+| Over-testing feasibility | Engineers love tech spikes | Balance across quadrants |
+| Vague assumptions | "Customers will like it" | Specific: "Customers will pay $X/mo" |
+
+## Worked Example — TeamFlow Assumption Map (AI 1:1 Summarization)
+
+**Context:** TeamFlow (B2B SaaS HR-tech). After Discovery + OST the team planned work on AI summarization. Before engineering begins — assumption map session (workshop 90 min, 8 participants: PM, Eng Lead, Designer, CTO, VP Sales, CPO-customer-advocate, Data Lead, Legal counsel).
+
+### Raw Assumptions Brainstorm (30+ assumptions in 30 min)
+
+Categorized into 4 quadrants with Risk Score (Uncertainty × Impact, 1-5 each).
+
+#### Value (Desirability) — Will customers want this?
+
+| # | Assumption | Uncertainty | Impact | Risk | Status |
+|---|-----------|:-----------:|:------:|:----:|:------:|
+| V1 | Customers will pay **$8-12/seat/month premium** for AI summarization feature | 4 | 5 | **20** | 🔴 Critical |
+| V2 | End-user managers will **adopt** AI summarization at >60% active users within 90 days | 4 | 4 | **16** | 🔴 Critical |
+| V3 | Buyers (VP HR) see **aggregate dashboard** as tier-upgrade justification | 3 | 5 | **15** | 🔴 Critical |
+| V4 | Feature reduces manager churn signal ("tool helps me") by >10pp | 4 | 3 | 12 | 🟡 Major |
+| V5 | Competitors (Lattice, 15Five) **have not yet shipped** AI summarization in the next 6 months | 3 | 4 | 12 | 🟡 Major |
+| V6 | Customers prefer AI-in-TeamFlow vs standalone (Otter.ai etc.) — "privacy + integrated" | 2 | 4 | 8 | 🟢 Low |
+
+#### Usability — Can managers figure out how to use it?
+
+| # | Assumption | Uncertainty | Impact | Risk | Status |
+|---|-----------|:-----------:|:------:|:----:|:------:|
+| U1 | Managers will **trust AI summaries** enough to rely on them (not manually re-write) | 4 | 4 | **16** | 🔴 Critical |
+| U2 | First-time user clearly understands how to **enable** AI summarization (opt-in UX) | 2 | 3 | 6 | 🟢 Low |
+| U3 | Summaries written in **tone appropriate** for performance conversations (not too casual, not too corporate) | 3 | 3 | 9 | 🟡 Major |
+| U4 | Managers will **correct errors** via inline edit (not abandon feature) | 3 | 3 | 9 | 🟡 Major |
+| U5 | Mobile experience for on-the-go review / approve summaries acceptable | 3 | 2 | 6 | 🟢 Low |
+
+#### Feasibility — Can we build it?
+
+| # | Assumption | Uncertainty | Impact | Risk | Status |
+|---|-----------|:-----------:|:------:|:----:|:------:|
+| F1 | **LLM summarization quality** acceptable for HR conversations (no hallucinations, no misattributions) — >85% acceptable rate from managers | 4 | 5 | **20** | 🔴 Critical |
+| F2 | Streaming **transcription latency** acceptable (<3 sec lag) for real-time in-meeting use | 3 | 4 | 12 | 🟡 Major |
+| F3 | Engineering can ship MVP in 8-10 weeks with existing team (8 eng) | 3 | 3 | 9 | 🟡 Major |
+| F4 | **Data residency** requirements (EU, HIPAA in healthcare customers) solvable via regional LLM endpoints (Azure, Bedrock) | 3 | 4 | 12 | 🟡 Major |
+| F5 | **Cost per summary** economically viable at target price ($0.05-0.10 per 30-min meeting) | 3 | 4 | 12 | 🟡 Major |
+| F6 | Integration with existing TeamFlow 1:1 data model without major refactor | 2 | 3 | 6 | 🟢 Low |
+
+#### Viability — Does it make business sense?
+
+| # | Assumption | Uncertainty | Impact | Risk | Status |
+|---|-----------|:-----------:|:------:|:----:|:------:|
+| Vi1 | **Privacy / compliance** (GDPR, SOC2, CCPA) manageable via architecture choices (opt-in, user-controlled, no training on data) | 3 | 5 | **15** | 🔴 Critical |
+| Vi2 | **Legal risk** (wrongful termination suit referencing AI summary as evidence) not material — enterprise customers do not backlash | 3 | 4 | 12 | 🟡 Major |
+| Vi3 | Pricing premium **does not cannibalize** core tier ARR (existing customers upgrade, not swap) | 3 | 4 | 12 | 🟡 Major |
+| Vi4 | **GTM execution** ready (Sales trained, CS enabled, Marketing positioned) in target timeline | 3 | 3 | 9 | 🟡 Major |
+| Vi5 | Infrastructure + LLM API costs **scale economically** at 5,000+ active users | 3 | 3 | 9 | 🟡 Major |
+| Vi6 | Customer success in EU / regulated industries (healthcare, finance) **does not block** sales | 3 | 3 | 9 | 🟡 Major |
+
+### 2×2 Prioritization Matrix
+
+```
+Impact if wrong →
+High (5) │  V1 (20)  F1 (20)   │  [Validated: F6, U2]
+         │  U1 (16)  V2 (16)   │
+         │  V3 (15)  Vi1 (15)  │
+         │                     │
+         │  V4 (12)  F2 (12)   │
+         │  F4 (12)  F5 (12)   │  
+         │  Vi2 (12) Vi3 (12)  │
+         │  V5 (12)            │  
+         │                     │
+         │  U3 (9)  U4 (9)     │  [Document]
+Low (1)  │  F3 (9)  Vi4 (9)    │
+         │  Vi5 (9) Vi6 (9)    │
+         │                     │
+         │  V6 (8)  U5 (6)     │  [Low priority]
+         │  U2 (6)  F6 (6)     │
+         ──────────────────────→
+         Low (1)     Uncertainty     High (5)
+```
+
+### Top-Right Assumptions (test NOW) — Risk ≥ 15
+
+1. **V1** (Pay $8-12 premium, Risk 20) — biggest money question
+2. **F1** (LLM quality acceptable, Risk 20) — biggest technical question
+3. **U1** (Managers trust summaries, Risk 16) — adoption question
+4. **V2** (>60% adoption in 90 days, Risk 16) — usage scale
+5. **V3** (Dashboard tier upgrade, Risk 15) — buyer expansion
+6. **Vi1** (Privacy manageable, Risk 15) — existential risk
+
+### Test Strategy per Top Assumption
+
+| # | Assumption | Method | Duration | Cost | Confidence |
+|---|-----------|--------|:--------:|:----:|:----------:|
+| V1 | Pay $8-12 premium | Landing page test with pricing + "notify me" CTA + follow-up interview | 2 weeks | Low | Medium |
+| V1 | (secondary) | Commitment test: 20 customers sign LOI for beta at $10/seat | 3 weeks | Low | High |
+| F1 | LLM quality | **Wizard-of-Oz:** OpenAI API + manual review, 20 beta managers, blind quality rating | 4 weeks | Medium | High |
+| U1 | Managers trust | Observational: beta managers — do they edit summaries? How often? Interview delta | 4 weeks (concurrent with F1) | Low | Medium |
+| V2 | 60% adoption | A/B launch: 50% feature flag, measure 30-day adoption | 4 weeks (post-MVP) | Medium | High |
+| V3 | Dashboard tier upgrade | Concierge: offer 5 enterprise customers "preview" dashboard, measure willingness to upgrade tier | 3 weeks | Low | Medium |
+| Vi1 | Privacy architecture | **Spike** (engineering + legal): design + privacy review + counsel opinion | 2 weeks | Medium | High |
+
+### Test Plan per Top Assumption (through `$hypothesis-template`)
+
+#### Test Card — V1 (Willingness to Pay)
+
+```
+**We believe** that mid-market and enterprise TeamFlow customers
+**For** accounts with ≥10 active managers
+**Will pay** $8-12/seat/month premium for AI 1:1 summarization
+**We'll know it's true when** either:
+  (a) 30% visitors to landing page click "Pricing details" AND 15% of those click "Get beta access"
+  (b) 8 of 10 customer conversations confirm "we'd upgrade at that price"
+**Because** 6 of 8 Discovery interviews expressed pay-interest
+    + competitor pricing ($5-7 for ChatGPT Teams) suggests price elasticity exists
+    + existing TeamFlow seats price point $15-50/mo → 20-30% premium within precedent range.
+```
+
+#### Test Card — F1 (LLM Quality)
+
+```
+**We believe** that GPT-4 / Claude 3.5 level LLMs
+**For** 30-minute 1:1 performance conversations in English
+**Will generate summaries** with >85% acceptable rate from managers
+**We'll know it's true when** in Wizard-of-Oz test:
+  - Blind quality rating from managers: ≥4 out of 5 stars (average), across ≥100 meetings
+  - Hallucination rate (factually wrong info): <5% of summaries
+  - Misattribution rate (wrong person said X): <3%
+**Because** recent LLM benchmarks on summarization task show 85-90% acceptable rates
+    + our manually-crafted test prompts achieved 90% in internal QA
+    + similar-adjacent use cases (Fireflies.ai, Gong) report >80% customer satisfaction.
+```
+
+### Iteration Log (3 weeks later, hypothetical)
+
+**Status updates:**
+- **V1:** ✅ Validated. Landing page test: 38% click-through, 22% "Get beta access". 7 of 10 customer conversations confirmed $10/seat acceptable. 2 customers said "we'd pay $15 for enterprise features". → **Risk downgraded to 6**.
+- **F1:** ⚠️ Partial. Wizard-of-Oz: 78% quality acceptable (below 85% target). Hallucinations 8% (above 5% target). → Additional work needed: prompt engineering + human-review layer. Risk **still 15**.
+- **U1:** ✅ Validated. Beta managers edit summaries 35% of time (not abandon). Sentiment positive. → Risk 6.
+- **Vi1:** ✅ Validated. Privacy architecture approved by counsel (opt-in, no training on data, regional endpoints). Risk 3.
+- **V2 / V3:** Pending MVP launch.
+
+**New assumptions emerged:**
+- **NEW V7:** "Customers accept 24-hour summary delivery delay for enhanced quality" (Uncertainty 4, Impact 3). 
+- **NEW F7:** "Human review layer scalable to 5,000 summaries/day" (Uncertainty 4, Impact 4). Added to high-risk quadrant.
+
+### Promoted to PRD Risk Register
+
+Assumptions with lingering risk (F1 quality, F7 human review scale) → PRD Top 5 risks section + mitigation plan.
+
+> **Assumption mapping lesson:** Brainstorm with a cross-functional team (legal, sales, CS in addition to PM + eng) revealed **Vi1 (privacy)** as an existential risk that an engineering-only session would have missed. Landing page test for V1 cost $400 and gave high-signal validation in 2 weeks — saved 10 weeks of engineering investment in the wrong pricing model.