| name | flame |
| description | Multi-perspective idea evaluation using Six Hats analysis, impact/feasibility scoring, and collaborative shortlisting. |
| version | 1.0.0 |
Evaluate ideas fairly through multiple perspectives. Surface hidden value, identify risks, and produce a ranked shortlist for Forge.
- Spark Bank generated and user wants to evaluate
- User provides ideas directly for evaluation
- Orchestrator routes to Flame
- User invokes `/specsmd-flame`
<degrees_of_freedom>
MEDIUM — Structured evaluation with room for user input and collaborative extension.
</degrees_of_freedom>
NEVER skip ideas — evaluate ALL ideas, not just obvious favorites (hidden gems exist)
NEVER evaluate through a single lens — always use multiple perspectives
ALWAYS pause at Red Hat for user's gut feeling (the one required elicitation)
ALWAYS present evaluation as multi-dimensional, not pass/fail
NEVER dismiss ideas prematurely — look for the kernel of value in each
- `.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
- `references/six-hats-method.md` — De Bono methodology adapted for AI execution
- `references/evaluation-criteria.md` — Impact, feasibility, novelty, risk frameworks
Load spark-bank.md from current session
Prioritize favorites but include all ideas
Accept ideas as input
Which ideas would you like to evaluate? You can list them or I can load from a Spark session.
For each idea (or top 10-15 if many), perform rapid Six Hats analysis:
**{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>
For each batch of analyzed ideas, pause and ask:
What's your gut feeling on these? Which ones excite you? Which feel wrong despite looking good on paper?
Record user's emotional responses
If user says "you decide" — infer gut feeling from their Spark favorites and engagement patterns
Score each evaluated idea on two axes:
**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>
Recommend top 3-5 ideas for Forge based on:
- Impact × Feasibility score
- User's Red Hat gut feelings
- Green Hat creative extensions
- Novelty (does this exist already?)
Present shortlist with brief rationale for each
Does this shortlist feel right? Want to adjust?
Generate Flame Report using template: templates/flame-report.md.hbs
Save to: .specs-ideation/sessions/{session-id}/flame-report.md
Update session.yaml: phase → "flame-complete"
Suggest moving to Forge: "Your top {N} ideas are ready to shape into concepts. Ready?"
If yes → invoke Forge skill
<output_artifacts>
| Artifact | Location | Template |
|---|
| Flame Report | .specs-ideation/sessions/{id}/flame-report.md | ./templates/flame-report.md.hbs |
| </output_artifacts> | | |
<success_criteria>
All ideas evaluated (not just favorites)
Multiple perspectives applied (Six Hats minimum)
User's gut feeling captured (Red Hat)
Scoring is multi-dimensional (impact + feasibility minimum)
Shortlist reflects both data and intuition
Flame Report is clear, organized, and immediately useful
</success_criteria>
