Flame

<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}** </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 </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>