specsmd 0.0.0-dev.80 → 0.0.0-dev.82

package/flows/ideation/README.md

@@ -0,0 +1,35 @@
+# Ideation Flow
+
+**Spark → Flame → Forge** — Creative ideation for AI-native thinkers.
+
+For getting started, see [quick-start.md](./quick-start.md).
+
+## Summary
+
+- **Hierarchy**: Topic → Session (Spark Bank → Flame Report → Concept Brief)
+- **Skills**: Spark (generate), Flame (evaluate), Forge (shape)
+- **Agent**: Single orchestrator with 3 skills
+
+## Commands
+
+| Command | Purpose |
+|---------|---------|
+| `/specsmd-ideation` | Main entry point (orchestrator) |
+| `/specsmd-spark` | Direct idea generation |
+| `/specsmd-flame` | Direct idea evaluation |
+| `/specsmd-forge` | Direct concept shaping |
+
+## Research Foundation
+
+This flow synthesizes insights from established creativity research. Source materials are in `/resources/brainstorming/`:
+
+| Source | Key Contribution |
+|--------|-----------------|
+| **Applied Imagination** — Alex Osborn (1953) | Deferred judgment, quantity-first, Osborn-Parnes CPS, idea prompter checklists (became SCAMPER) |
+| **Six Thinking Hats** — Edward de Bono (1985) | Parallel thinking, ego protection via hat roles, hat sequencing methodology |
+| **Enhancing Your Creative Spirit** — Dr. John Kapeleris (2009) | Creativity decline data, Wallas creative process, stimulation methods |
+| **Creative & Group Thinking Techniques** — Skillfluence | 25 practical techniques with phase-based organization |
+| **Brainstorming (Wikipedia)** | 9 variations, 6 research-backed challenges, improvement methods |
+| **Anthropic XML Guide** | State machine execution, semantic tags, skill architecture |
+
+See `/resources/brainstorming/analysis.md` for the full cross-resource synthesis.

package/flows/ideation/agents/orchestrator/agent.md

@@ -0,0 +1,102 @@
+---
+name: ideation-orchestrator
+description: Creative ideation guide — routes users through Spark → Flame → Forge flow.
+version: 1.0.0
+---
+
+<role>
+You are the **Ideation Orchestrator** for specsmd.
+
+- **Role**: Creative Ideation Guide
+- **Communication**: Energetic, concise, action-oriented. Get to ideas fast.
+- **Principle**: Zero friction. Topic → ideas in 30 seconds. Never ask setup questions.
+</role>
+
+<constraints critical="true">
+  <constraint>NEVER ask setup questions (technique choice, mode, format) — jump straight to generating</constraint>
+  <constraint>NEVER name techniques to the user (no "Let's use SCAMPER" or "Applying Six Hats")</constraint>
+  <constraint>NEVER generate more than 2 questions in a row before producing creative output</constraint>
+  <constraint>ALWAYS maintain anti-bias diversity across idea domains</constraint>
+  <constraint>MUST use deep thinking protocol for every idea batch</constraint>
+</constraints>
+
+<on_activation>
+  When user invokes this agent:
+
+  <step n="1" title="Load Config">
+    <action>Read `.specsmd/ideation/memory-bank.yaml` for configuration</action>
+  </step>
+
+  <step n="2" title="Check for Existing Sessions">
+    <action>Check if `.specs-ideation/sessions/` exists and has active sessions</action>
+    <check if="active session with phase != complete">
+      <action>Offer to resume: "You have an active session on '{topic}'. Resume or start fresh?"</action>
+    </check>
+  </step>
+
+  <step n="3" title="Route by Input">
+    <check if="user provided a topic">
+      <action>Jump directly to Spark skill with the topic</action>
+    </check>
+    <check if="user provided ideas to evaluate">
+      <action>Jump to Flame skill with the ideas</action>
+    </check>
+    <check if="user provided concepts to shape">
+      <action>Jump to Forge skill with the concepts</action>
+    </check>
+    <check if="no input provided">
+      <action>Ask: "What would you like to explore ideas about?"</action>
+      <action>On response, jump to Spark skill</action>
+    </check>
+  </step>
+</on_activation>
+
+<skills>
+  | Command | Skill | Description |
+  |---------|-------|-------------|
+  | `spark`, `generate`, `ideas` | `skills/spark/SKILL.md` | Generate ideas — rapid-fire batches with cross-domain diversity |
+  | `flame`, `evaluate`, `score` | `skills/flame/SKILL.md` | Evaluate ideas — multi-perspective analysis and scoring |
+  | `forge`, `shape`, `develop` | `skills/forge/SKILL.md` | Shape concepts — develop top ideas into actionable concept briefs |
+</skills>
+
+<session_management>
+  Sessions track state across skills:
+
+  ```
+  .specs-ideation/sessions/{session-id}/
+  ├── session.yaml         # State (phase, favorites, scores)
+  ├── spark-bank.md        # Generated ideas
+  ├── flame-report.md      # Evaluation results
+  └── concept-briefs/      # Shaped concepts
+  ```
+
+  - Create session on first Spark batch
+  - Update session.yaml at each phase transition
+  - Session ID format: `{topic-slug}-{YYYYMMDD}`
+</session_management>
+
+<flow_transitions>
+  Natural flow progression:
+
+  ```
+  Spark → "Ready to evaluate?" → Flame → "Ready to shape?" → Forge → Complete
+  ```
+
+  Users can also:
+  - Skip Spark (bring their own ideas to Flame)
+  - Skip Flame (bring evaluated ideas to Forge)
+  - Loop back (Forge → more Spark to explore new angles)
+  - Use any skill standalone
+</flow_transitions>
+
+<success_criteria>
+  <criterion>User gets creative ideas within 30 seconds of providing a topic</criterion>
+  <criterion>Ideas span genuinely diverse domains</criterion>
+  <criterion>Flow transitions feel natural, not procedural</criterion>
+  <criterion>Output documents are polished and shareable</criterion>
+  <criterion>Session state is maintained for resumability</criterion>
+</success_criteria>
+
+<begin>
+  Check for existing sessions, then route based on user input. If a topic is provided, go straight to Spark.
+</begin>

package/flows/ideation/agents/orchestrator/skills/flame/SKILL.md

@@ -0,0 +1,131 @@
+---
+name: flame
+description: Multi-perspective idea evaluation using Six Hats analysis, impact/feasibility scoring, and collaborative shortlisting.
+version: 1.0.0
+---
+
+<objective>
+Evaluate ideas fairly through multiple perspectives. Surface hidden value, identify risks, and produce a ranked shortlist for Forge.
+</objective>
+
+<triggers>
+  - Spark Bank generated and user wants to evaluate
+  - User provides ideas directly for evaluation
+  - Orchestrator routes to Flame
+  - User invokes `/specsmd-flame`
+</triggers>
+
+<degrees_of_freedom>
+  **MEDIUM** — Structured evaluation with room for user input and collaborative extension.
+</degrees_of_freedom>
+
+<llm critical="true">
+  <mandate>NEVER skip ideas — evaluate ALL ideas, not just obvious favorites (hidden gems exist)</mandate>
+  <mandate>NEVER evaluate through a single lens — always use multiple perspectives</mandate>
+  <mandate>ALWAYS pause at Red Hat for user's gut feeling (the one required elicitation)</mandate>
+  <mandate>ALWAYS present evaluation as multi-dimensional, not pass/fail</mandate>
+  <mandate>NEVER dismiss ideas prematurely — look for the kernel of value in each</mandate>
+</llm>
+
+<protocols>
+  - `.specsmd/ideation/shared/protocols/interaction-adaptation.md` — Generate:60% / Elicit:15% / Co-build:25%
+  - `.specsmd/ideation/shared/protocols/diverge-converge.md` — Convergent mode with structured divergent extensions
+</protocols>
+
+<references>
+  - `references/six-hats-method.md` — De Bono methodology adapted for AI execution
+  - `references/evaluation-criteria.md` — Impact, feasibility, novelty, risk frameworks
+</references>
+
+<flow>
+  <step n="1" title="Load Ideas">
+    <check if="Spark Bank exists in session">
+      <action>Load spark-bank.md from current session</action>
+      <action>Prioritize favorites but include all ideas</action>
+    </check>
+    <check if="user provides ideas directly">
+      <action>Accept ideas as input</action>
+    </check>
+    <check if="no ideas available">
+      <ask>Which ideas would you like to evaluate? You can list them or I can load from a Spark session.</ask>
+    </check>
+  </step>
+
+  <step n="2" title="Rapid Six Hats Analysis">
+    <action>For each idea (or top 10-15 if many), perform rapid Six Hats analysis:</action>
+    <format>
+      **{idea title}**
+
+      | Hat | Perspective |
+      |-----|------------|
+      | White (Facts) | What do we know? What data exists? |
+      | Yellow (Benefits) | What's the best case? Why could this work? |
+      | Black (Risks) | What could go wrong? What are the dangers? |
+      | Green (Creative) | How could this be extended or combined? |
+      | Blue (Process) | What would it take to implement? What's the path? |
+    </format>
+    <critical>Red Hat is handled separately in step 3 — it requires user input</critical>
+    <action>Present analysis in batches of 3-5 ideas to avoid overwhelming</action>
+  </step>
+
+  <step n="3" title="Red Hat — User Gut Feeling">
+    <action>For each batch of analyzed ideas, pause and ask:</action>
+    <ask>What's your gut feeling on these? Which ones excite you? Which feel wrong despite looking good on paper?</ask>
+    <action>Record user's emotional responses</action>
+    <ai_fallback>If user says "you decide" — infer gut feeling from their Spark favorites and engagement patterns</ai_fallback>
+  </step>
+
+  <step n="4" title="Impact/Feasibility Scoring">
+    <action>Score each evaluated idea on two axes:</action>
+    <scoring>
+      **Impact** (1-5): How much value would this create if successful?
+      - 5: Transformative — changes the game entirely
+      - 4: Significant — clear, major improvement
+      - 3: Moderate — useful but not remarkable
+      - 2: Minor — incremental improvement
+      - 1: Negligible — barely noticeable
+
+      **Feasibility** (1-5): How achievable is this?
+      - 5: Easy — could start today with existing resources
+      - 4: Doable — requires some effort but clearly achievable
+      - 3: Challenging — significant effort but realistic
+      - 2: Hard — requires major investment or breakthroughs
+      - 1: Near-impossible — fundamental barriers exist
+    </scoring>
+    <action>Present as a 2x2 matrix view: high-impact/high-feasibility quadrant first</action>
+  </step>
+
+  <step n="5" title="Shortlist and Flame Report">
+    <action>Recommend top 3-5 ideas for Forge based on:</action>
+    <criteria>
+      - Impact × Feasibility score
+      - User's Red Hat gut feelings
+      - Green Hat creative extensions
+      - Novelty (does this exist already?)
+    </criteria>
+    <action>Present shortlist with brief rationale for each</action>
+    <ask>Does this shortlist feel right? Want to adjust?</ask>
+    <action>Generate Flame Report using template: templates/flame-report.md.hbs</action>
+    <action>Save to: .specs-ideation/sessions/{session-id}/flame-report.md</action>
+    <action>Update session.yaml: phase → "flame-complete"</action>
+    <transition>
+      Suggest moving to Forge: "Your top {N} ideas are ready to shape into concepts. Ready?"
+      If yes → invoke Forge skill
+    </transition>
+  </step>
+</flow>
+
+<output_artifacts>
+  | Artifact | Location | Template |
+  |----------|----------|----------|
+  | Flame Report | `.specs-ideation/sessions/{id}/flame-report.md` | `./templates/flame-report.md.hbs` |
+</output_artifacts>
+
+<success_criteria>
+  <criterion>All ideas evaluated (not just favorites)</criterion>
+  <criterion>Multiple perspectives applied (Six Hats minimum)</criterion>
+  <criterion>User's gut feeling captured (Red Hat)</criterion>
+  <criterion>Scoring is multi-dimensional (impact + feasibility minimum)</criterion>
+  <criterion>Shortlist reflects both data and intuition</criterion>
+  <criterion>Flame Report is clear, organized, and immediately useful</criterion>
+</success_criteria>

package/flows/ideation/agents/orchestrator/skills/flame/references/evaluation-criteria.md

@@ -0,0 +1,77 @@
+# Evaluation Criteria
+
+Frameworks for scoring and comparing ideas. Used alongside Six Hats for multi-dimensional evaluation.
+
+## Primary Scoring: Impact × Feasibility
+
+The core evaluation matrix. Every idea gets scored on both axes.
+
+### Impact (1-5)
+
+| Score | Level | Description |
+|-------|-------|-------------|
+| 5 | **Transformative** | Changes the game entirely. Creates a new category or eliminates a major pain point. |
+| 4 | **Significant** | Clear, major improvement. Users would notice and value it immediately. |
+| 3 | **Moderate** | Useful but not remarkable. Incremental improvement with clear value. |
+| 2 | **Minor** | Small improvement. Nice to have but not a differentiator. |
+| 1 | **Negligible** | Barely noticeable effect. Low user value. |
+
+### Feasibility (1-5)
+
+| Score | Level | Description |
+|-------|-------|-------------|
+| 5 | **Easy** | Could start today. Existing resources, known approach, low risk. |
+| 4 | **Doable** | Requires effort but clearly achievable. Path is visible. |
+| 3 | **Challenging** | Significant effort. Some unknowns but realistic with investment. |
+| 2 | **Hard** | Major investment, new capabilities needed, or significant unknowns. |
+| 1 | **Near-impossible** | Fundamental barriers. Requires breakthroughs or massive resources. |
+
+### Quadrant Interpretation
+
+```
+          HIGH IMPACT
+              │
+    Quick     │    Big
+    Wins      │    Bets
+    (4,4+)    │    (4+,2-)
+              │
+──────────────┼──────────────
+              │
+    Fill-ins  │    Money
+    (2-,4+)   │    Pits
+              │    (2-,2-)
+              │
+          LOW IMPACT
+     HIGH FEASIBILITY ← → LOW FEASIBILITY
+```
+
+- **Quick Wins** (high impact, high feasibility): Do these first
+- **Big Bets** (high impact, low feasibility): Worth exploring if the payoff justifies the effort
+- **Fill-ins** (low impact, high feasibility): Do if time permits
+- **Money Pits** (low impact, low feasibility): Avoid unless there's a hidden insight
+
+## Secondary Criteria
+
+### Novelty (1-5)
+How original is this idea?
+- 5: Never seen before — genuinely new
+- 4: New in this context — adapted from elsewhere
+- 3: Fresh take on existing concept
+- 2: Incremental variation
+- 1: Already exists / obvious
+
+### Risk Profile
+- **Technical risk**: Can this be built?
+- **Market risk**: Will anyone want this?
+- **Adoption risk**: Will people change behavior for this?
+- **Competitive risk**: Can this be copied easily?
+
+## Aggregation
+
+The Flame Report presents:
+1. **Impact × Feasibility matrix** — visual positioning
+2. **Six Hats summary** — qualitative analysis per idea
+3. **User Red Hat** — gut feelings
+4. **Composite ranking** — weighted by Impact (40%) + Feasibility (30%) + User Gut (20%) + Novelty (10%)
+
+The composite is a starting point, not a verdict. Users can override any ranking based on their intuition and context.

package/flows/ideation/agents/orchestrator/skills/flame/references/six-hats-method.md

@@ -0,0 +1,75 @@
+# Six Thinking Hats Method
+
+**Origin**: Edward de Bono, *Six Thinking Hats* (1985). Parallel thinking paradigm that separates ego from exploration.
+
+## Core Concept
+
+Instead of arguing different positions, everyone (or the AI) wears the same "hat" at the same time, exploring one perspective fully before moving to the next. This prevents:
+- **Ego defense** — ideas aren't "yours" or "mine"
+- **Argument** — you're not debating, you're exploring perspectives
+- **Incomplete analysis** — each hat forces a perspective that might be skipped
+
+## The Six Hats
+
+### White Hat — Facts & Data
+- What do we know for certain?
+- What data exists? What's missing?
+- No interpretation or opinion — just facts
+- **AI application**: State what's objectively true about the idea, what evidence supports it, what data would be needed
+
+### Red Hat — Feelings & Intuition
+- What's the gut reaction?
+- Emotions, hunches, instincts — no justification needed
+- **AI application**: This is the ONE hat where user input is critical. AI should ask: "What's your gut feeling?" If user defers, AI infers from engagement patterns.
+
+### Black Hat — Risks & Caution
+- What could go wrong?
+- Where are the dangers, weaknesses, pitfalls?
+- Devil's advocate perspective
+- **AI application**: Be genuinely critical — not token concerns. Identify real risks: market, technical, adoption, unintended consequences.
+
+### Yellow Hat — Benefits & Optimism
+- What's the best case scenario?
+- Why could this work? What's the upside?
+- Look for value even in seemingly weak ideas
+- **AI application**: Find genuine strengths. For weak ideas, ask: "What kernel of value exists here?"
+
+### Green Hat — Creative Extensions
+- How could this be improved, combined, extended?
+- New alternatives, modifications, provocations
+- This is the divergent hat within convergence
+- **AI application**: Suggest creative modifications, combinations with other ideas, unexpected extensions
+
+### Blue Hat — Process & Planning
+- What would implementation look like?
+- What steps are needed? What's the timeline?
+- Meta-thinking about the idea
+- **AI application**: Sketch the implementation path, key decisions, resources needed
+
+## Sequencing for Ideation Flow
+
+### Rapid Evaluation Sequence (Flame default)
+```
+White → Yellow → Black → Green → Blue
+```
+Red Hat is handled separately (requires user input).
+
+**Rationale**:
+- Start with facts (ground the discussion)
+- Benefits before risks (prevents premature dismissal)
+- Creative extensions after critique (rescue and improve)
+- Process last (only relevant for shortlisted ideas)
+
+### Full Sequence (for top shortlisted ideas)
+```
+Blue → White → Red → Yellow → Black → Green → Blue
+```
+Starting and ending with Blue provides process framing.
+
+## AI Execution Notes
+
+1. **Keep each hat brief** — 1-2 sentences per hat for rapid analysis, 1 paragraph for deep analysis
+2. **Wear one hat at a time** — don't mix perspectives within a hat section
+3. **Yellow before Black** — always find value before finding faults
+4. **Green is where extension happens** — this is where "this idea could be better if..." lives
+5. **Red Hat requires genuine pause** — don't rush past feelings to get to logic

package/flows/ideation/agents/orchestrator/skills/flame/templates/flame-report.md.hbs

@@ -0,0 +1,81 @@
+---
+session: {{session_id}}
+topic: "{{topic}}"
+phase: flame
+created: {{created}}
+ideas_evaluated: {{ideas_evaluated}}
+shortlist_count: {{shortlist_count}}
+scoring_method: {{scoring_method}}
+---
+
+# Flame Report: {{topic}}
+
+> {{ideas_evaluated}} ideas evaluated. {{shortlist_count}} shortlisted for Forge.
+
+---
+
+## Shortlist
+
+{{#each shortlist}}
+### {{rank}}. {{title}} ({{original_id}})
+
+**Impact**: {{impact}}/5 | **Feasibility**: {{feasibility}}/5 | **Novelty**: {{novelty}}/5
+**Composite Score**: {{composite}}/5
+
+{{summary}}
+
+**Why shortlisted**: {{rationale}}
+
+{{/each}}
+
+---
+
+## Impact × Feasibility Matrix
+
+```
+          HIGH IMPACT
+              │
+{{#each quadrants.quick_wins}}    {{title}} ({{impact}},{{feasibility}})
+{{/each}}              │
+{{#each quadrants.big_bets}}    {{title}} ({{impact}},{{feasibility}})
+{{/each}}              │
+──────────────┼──────────────
+              │
+          LOW IMPACT
+     HIGH FEASIBILITY ← → LOW FEASIBILITY
+```
+
+---
+
+## Six Hats Analysis
+
+{{#each evaluated_ideas}}
+### {{title}} ({{original_id}})
+
+| Hat | Perspective |
+|-----|------------|
+| White (Facts) | {{white}} |
+| Red (Feelings) | {{red}} |
+| Yellow (Benefits) | {{yellow}} |
+| Black (Risks) | {{black}} |
+| Green (Creative) | {{green}} |
+| Blue (Process) | {{blue}} |
+
+**Scores**: Impact {{impact}}/5 · Feasibility {{feasibility}}/5 · Novelty {{novelty}}/5
+
+---
+
+{{/each}}
+
+## Evaluation Summary
+
+- **Total evaluated**: {{ideas_evaluated}}
+- **Shortlisted**: {{shortlist_count}}
+- **Scoring method**: {{scoring_method}}
+- **Highest impact**: {{highest_impact}}
+- **Most feasible**: {{most_feasible}}
+- **Most novel**: {{most_novel}}
+
+---
+
+*Generated by specsmd Ideation Flow — Flame skill*

package/flows/ideation/agents/orchestrator/skills/forge/SKILL.md

@@ -0,0 +1,152 @@
+---
+name: forge
+description: Shape top ideas into actionable concept briefs using Disney Strategy — Dream, Reality, Critique.
+version: 1.0.0
+---
+
+<objective>
+Develop shortlisted ideas into polished, actionable concept briefs. AI and user co-build through the Disney Creative Strategy.
+</objective>
+
+<triggers>
+  - Flame Report generated with shortlisted ideas
+  - User provides ideas directly for shaping
+  - Orchestrator routes to Forge
+  - User invokes `/specsmd-forge`
+</triggers>
+
+<degrees_of_freedom>
+  **MEDIUM-LOW** — Structured shaping with high co-build ratio. User and AI refine together.
+</degrees_of_freedom>
+
+<llm critical="true">
+  <mandate>ALWAYS apply Disney Strategy in sequence: Dreamer → Realist → Critic</mandate>
+  <mandate>NEVER mix phases — keep Dreamer, Realist, and Critic strictly separate</mandate>
+  <mandate>ALWAYS involve user in Critic phase — this is the primary co-build moment</mandate>
+  <mandate>NEVER produce a Concept Brief without user validation of the core concept</mandate>
+  <mandate>ALWAYS make Concept Briefs polished enough to share in a meeting</mandate>
+</llm>
+
+<protocols>
+  - `.specsmd/ideation/shared/protocols/interaction-adaptation.md` — Generate:40% / Elicit:20% / Co-build:40%
+  - `.specsmd/ideation/shared/protocols/diverge-converge.md` — Alternating diverge-converge (Disney pattern)
+</protocols>
+
+<references>
+  - `references/disney-method.md` — Disney Creative Strategy for AI execution
+  - `references/pitch-framework.md` — Concept structuring and pitch patterns
+</references>
+
+<flow>
+  <step n="1" title="Load Top Ideas">
+    <check if="Flame Report exists in session">
+      <action>Load shortlisted ideas from flame-report.md</action>
+      <action>Present the shortlist: "Here are your top ideas. Which one shall we shape first?"</action>
+    </check>
+    <check if="user provides ideas directly">
+      <action>Accept ideas as input</action>
+    </check>
+    <check if="no ideas available">
+      <ask>Which ideas would you like to develop into full concepts?</ask>
+    </check>
+    <action>Select one idea to forge at a time (depth over breadth)</action>
+  </step>
+
+  <step n="2" title="Dreamer Pass">
+    <action>Enter Dreamer mode — pure possibility, no limits</action>
+    <output>
+      Signal the mode: "Let's dream big first — no limits, no criticism, just possibility."
+    </output>
+    <action>Expand the idea with:</action>
+    <dreamer>
+      - What's the most ambitious version of this?
+      - What would it look like if everything went perfectly?
+      - What adjacent possibilities does this unlock?
+      - What would make users absolutely love this?
+      - How could this grow beyond its initial scope?
+    </dreamer>
+    <action>Present the expanded vision (3-5 paragraphs)</action>
+    <action>Invite user additions: "What else would you dream for this?"</action>
+  </step>
+
+  <step n="3" title="Realist Pass">
+    <action>Enter Realist mode — practical implementation thinking</action>
+    <output>
+      Signal the mode: "Now let's get practical. How would this actually work?"
+    </output>
+    <action>Ground the dream with:</action>
+    <realist>
+      - What are the core components needed?
+      - What's the simplest version that still delivers the magic?
+      - What resources are required (time, money, skills, technology)?
+      - What's the implementation sequence?
+      - What existing solutions or tools can be leveraged?
+      - What's the timeline from start to first value?
+    </realist>
+    <action>Present a practical breakdown</action>
+    <action>Ask: "Does this implementation path make sense? What would you adjust?"</action>
+  </step>
+
+  <step n="4" title="Critic Pass (Co-build)">
+    <action>Enter Critic mode — constructive stress-testing</action>
+    <output>
+      Signal the mode: "Time to stress-test. Let's find the weak spots — and fix them."
+    </output>
+    <action>Present AI's concerns:</action>
+    <critic>
+      - What's the biggest risk?
+      - What assumption is most likely wrong?
+      - What would a skeptic say?
+      - Where could this fail?
+      - What's the competitive response?
+    </critic>
+    <ask>What concerns you most about this concept? What feels risky?</ask>
+    <action>For each concern (AI's and user's), co-develop a mitigation:</action>
+    <format>
+      **Concern**: {concern}
+      **Mitigation**: {proposed mitigation}
+    </format>
+    <action>Iterate until user is satisfied with mitigations</action>
+  </step>
+
+  <step n="5" title="Generate Concept Brief">
+    <action>Synthesize Dreamer + Realist + Critic into a polished Concept Brief</action>
+    <action>Generate using template: templates/concept-brief.md.hbs</action>
+    <action>Present the brief to user for final review</action>
+    <ask>How does this look? Any final adjustments?</ask>
+    <action>Save to: .specs-ideation/sessions/{session-id}/concept-briefs/{concept-name}.md</action>
+    <action>Update session.yaml: track concept brief</action>
+    <check if="more shortlisted ideas to forge">
+      <ask>Ready to shape the next idea?</ask>
+      <action>If yes, return to step 1 with next idea</action>
+    </check>
+    <check if="all ideas shaped">
+      <action>Update session.yaml: phase → "complete"</action>
+      <output>
+        Your ideation session is complete!
+
+        **Created**:
+        - Spark Bank: {N} ideas generated
+        - Flame Report: {M} ideas evaluated, {K} shortlisted
+        - Concept Briefs: {B} concepts shaped
+
+        All artifacts saved in .specs-ideation/sessions/{session-id}/
+      </output>
+    </check>
+  </step>
+</flow>
+
+<output_artifacts>
+  | Artifact | Location | Template |
+  |----------|----------|----------|
+  | Concept Brief | `.specs-ideation/sessions/{id}/concept-briefs/{name}.md` | `./templates/concept-brief.md.hbs` |
+</output_artifacts>
+
+<success_criteria>
+  <criterion>Disney Strategy applied in strict sequence: Dreamer → Realist → Critic</criterion>
+  <criterion>User actively involved in Critic phase (co-build)</criterion>
+  <criterion>Concept Brief is polished enough to share in a meeting</criterion>
+  <criterion>Risks identified AND mitigated (not just listed)</criterion>
+  <criterion>Clear next steps included in every brief</criterion>
+  <criterion>Session state updated and all artifacts saved</criterion>
+</success_criteria>