@thierrynakoa/fire-flow 12.2.1 → 13.0.1

package/commands/fire-1d-discuss.md

@@ -71,6 +71,23 @@ arguments:
 
 ## Process
 
+### Wizard Step Sequence
+
+```mermaid
+graph LR
+    S1[1: Validate] --> S2[2: Check Context]
+    S2 --> S25[2.5: Independence]
+    S25 --> S3[3: Gray Areas]
+    S3 --> S35[3.5: Specificity Gate]
+    S35 --> S37[3.7: Decomp Check]
+    S37 --> S4[4: Present Areas]
+    S4 --> S5[5: Discuss]
+    S5 --> S6[6: Write MEMORY.md]
+    S6 --> S7[7: Confirm & Route]
+```
+
+> formats measurably more reliably than prose step lists. Zero-cost adherence improvement.
+
 ### Step 1: Validate Phase
 
 ```
@@ -223,6 +240,28 @@ FOR each requirement discussed so far:
 
 **Goal:** No Level 1/2 requirements should survive the discuss phase. The planner receives Level 3+ requirements only.
 
+### Step 3.75: Consistency-Check Ambiguity Gate (v12.3)
+
+> check before generation raises GPT-4 Pass@1 by +9.84pp. The pattern: generate two
+> brief candidate interpretations; if they diverge significantly, ask before proceeding.
+
+Before presenting gray areas in Step 4, run an ambiguity check:
+
+```
+For the phase being discussed:
+  Generate 2 brief candidate interpretations of what this phase delivers.
+
+  IF the two interpretations diverge (different components, UX, scope):
+    → Do NOT proceed to gray area selection.
+    → Ask ONE targeted clarifying question to resolve the ambiguity first.
+    → Example: "This phase could mean X or Y — which direction?"
+
+  IF the two interpretations align:
+    → Proceed to Step 4 with high confidence in scope.
+```
+
+This gate prevents discussing the WRONG phase scope and discovering the mistake after 4+ questions.
+
 ### Step 4: Present Gray Areas
 
 ```
@@ -274,6 +313,37 @@ For each selected area, conduct focused discussion.
 - question: "More questions about {area}, or move to next?"
 - options: "More questions" / "Next area"
 
+**Step Confirmation Checkpoint (W1-B — before advancing):**
+
+After the user signals "Next area", display:
+
+```
+AREA DECISION LOCKED: {Area Name}
+────────────────────────────────────────────────────
+Decision: {1-line summary of what was decided}
+
+Proceed to next area? [Yes / Review this area / Skip remaining]
+```
+
+**Missing-Dimension Detector (W1-D — proactive question check):**
+
+Before advancing to the next area, ask internally:
+
+> "What critical information about {area} has NOT been mentioned yet
+>  that the planner would need to implement this without asking again?"
+
+If any critical dimensions are missing (e.g., user discussed layout but never mentioned mobile), ask ONE follow-up question before advancing.
+
+> 2507.21389, 2025) — proactive dimension detection outperforms o3-mini by +18% on
+> question quality. Reactive wizards only ask about what was said; proactive wizards
+> surface what the user forgot to say.
+
+**Per-Step Self-Check (W1-C):**
+
+Ask internally: "Is this decision specific enough for a planner to act on without
+asking the user again?" If NO — generate ONE follow-up before advancing.
+Threshold: Level 3+ on the requirements decomposition scale from Step 3.7.
+
 **Scope Creep Handling:**
 ```
 "{Feature} sounds like a new capability — that belongs in its own phase.
@@ -332,6 +402,22 @@ Track deferred ideas internally.
 
 ```
 ╔══════════════════════════════════════════════════════════════════════════════╗
+║ ✓ DISCUSSION COMPLETE                                                        ║
+╠══════════════════════════════════════════════════════════════════════════════╣
+║                                                                              ║
+║  AREAS DISCUSSED:                                                            ║
+║  ├── {Area 1}: {key decision — 1 line}                                       ║
+║  ├── {Area 2}: {key decision — 1 line}                                       ║
+║  └── {Area N}: {key decision — 1 line}                                       ║
+║                                                                              ║
+║  QUALITY CHECKS:                                                             ║
+║  [✓] All decisions at Level 3+ specificity (no vague Level 1/2 requirements) ║
+║  [✓] Missing dimensions probed per area                                      ║
+║  [✓] Per-step self-check passed for each area                                ║
+║  [✓] Deferred ideas captured (scope creep redirected)                        ║
+║  [✓] MEMORY.md written with actionable decisions                             ║
+║                                                                              ║
+╠══════════════════════════════════════════════════════════════════════════════╣
 ║ ✓ CONTEXT CAPTURED                                                           ║
 ╠══════════════════════════════════════════════════════════════════════════════╣
 ║                                                                              ║
@@ -365,7 +451,7 @@ Track deferred ideas internally.
 
 > *Like humans need sleep to reset, AI agents need state files to resume after context resets.*
 
-**Create/Update:** `.claude/fire-flow.local.md`
+**Create/Update:** `.claude/dominion-flow.local.md`
 
 ```markdown
 ---