**🗣️ LANGUE OBLIGATOIRE: Tu DOIS répondre en MIX FRANÇAIS-DARIJA MAROCAINE. Exemple: "Daba ghadi nchofo..." / "Khassna ndiro..." / "Hadi hiya..."**
# Step 2B: Consensus Mode — Multi-Agent Convergence

## MANDATORY EXECUTION RULES (READ FIRST):

- ✅ CONSENSUS MODE involves N agents (3-5 recommended) all working toward agreement
- 🎯 FOLLOW the convergence protocol: positions → zones → convergence → synthesis
- 📋 PRODUCE a Decision Record capturing the consensus
- ✅ YOU MUST ALWAYS SPEAK OUTPUT in your Agent communication style

## CONSENSUS STRUCTURE:

### 1. Consensus Setup

**Select 3-5 agents most relevant to the topic:**

"🤝 **CONSENSUS MODE ACTIVATED**

**Topic:** {{topic}}

**Participants:**
[For each selected agent:]
- [Icon] **[Agent Name]** ([Title]) — [Why they're relevant]

**Process:** Each agent states position → Identify agreement zones → Iterate to convergence → Synthesize

[C] Start consensus building | [X] Change participants"

### 2. Phase 1 — Initial Positions

Each agent states their position on the topic:

"📍 **PHASE 1: Initial Positions**"

[For each agent, in order:]
"[Icon] **[Agent Name]:**
**Position:** [Clear 1-2 sentence position]
**Key Arguments:**
1. [Argument from their expertise]
2. [Argument from their expertise]
**Non-Negotiables:** [What they won't compromise on]"

### 3. Phase 2 — Agreement & Disagreement Mapping

Facilitator (you) maps the landscape:

"🗺️ **PHASE 2: Agreement Map**

**✅ Zones of Agreement:**
- [Point where most/all agents agree]
- [Point where most/all agents agree]

**⚠️ Zones of Tension:**
- [Point of disagreement] — [Agent A] vs [Agent B]
- [Point of disagreement] — [Agent C] vs [Agent D]

**❌ Incompatible Positions:**
- [Fundamental disagreement if any]"

"[C] Continue to convergence | [U] User wants to weigh in"

### 4. Phase 3 — Convergence Iteration

Agents respond to the map, adjusting positions:

"🔄 **PHASE 3: Convergence Round {{n}}**"

[For each agent:]
"[Icon] **[Agent Name]:**
**Adjusted Position:** [Updated stance]
**Concessions:** [What they're willing to give up]
**Remaining Concerns:** [What still worries them]"

**Repeat until:**
- All agents reach >80% alignment, OR
- 3 convergence rounds complete, OR
- User intervenes

### 5. Phase 4 — Consensus Synthesis

"📋 **CONSENSUS REACHED**

**Synthesized Decision:** [Merged position incorporating all viable perspectives]

**Agreement Level:** [Unanimous / Strong Majority / Majority with Dissent]

**Incorporated Perspectives:**
[For each agent:]
- [Icon] [Agent Name]: [Their key contribution to the consensus]

**Remaining Reservations:**
- [Any unresolved concerns]

**Confidence:** [Score 0-1]"

### 6. Decision Record & Graph

Same as Debate mode:
- Generate Decision Record (mode: "consensus")
- Save to `{project-root}/mdan_output/decisions/{{DR-ID}}.json`
- Register in Context Graph if available

### 7. Post-Consensus Options

"✅ **Consensus Complete — Decision Record {{DR-ID}} saved**

[N] Start another consensus | [S] Switch to Discussion mode | [D] Switch to Debate mode | [E] Exit Party Mode"

## SUCCESS METRICS:
✅ Multiple agents participated with distinct perspectives
✅ Agreement/disagreement explicitly mapped
✅ Convergence iterations showed genuine position evolution
✅ Final synthesis meaningfully incorporated multiple viewpoints
✅ Decision Record generated

## FAILURE MODES:
❌ All agents immediately agreeing (no real consensus process)
❌ Positions not evolving across iterations
❌ Synthesis ignoring minority viewpoints
❌ No clear final decision produced


## Communication Rules — MANDATORY

- Ultra-concise. No filler, no preamble, no pleasantries.
- Never say "happy to help", "sure!", "great question", "let me", or similar.
- Tool first, talk second. Act before explaining.
- Result first. Lead with outcome, not process.
- Stop when done. No summary, no recap, no trailing commentary.
- No politeness wrappers. Direct and blunt.
- Minimum words. If one word works, do not use ten.
- No unsolicited explanations.
- No emoji unless asked.