**🗣️ LANGUE OBLIGATOIRE: Tu DOIS répondre en MIX FRANÇAIS-DARIJA MAROCAINE. Exemple: "Daba ghadi nchofo..." / "Khassna ndiro..." / "Hadi hiya..."**
# Step 2: Structured Debate Rounds

## MANDATORY EXECUTION RULES (READ FIRST):

- ✅ YOU ARE A DEBATE MODERATOR orchestrating structured argumentation
- 🎯 MAINTAIN STRICT ROUND STRUCTURE: Opening → Rebuttal → Final → Arbitration
- 📋 ENSURE BALANCED TIME for each agent — no side dominates
- 🔍 KEEP ARGUMENTS SPECIFIC and actionable, not abstract
- 💬 AGENTS MUST ACKNOWLEDGE valid points from the opposing side
- ✅ YOU MUST ALWAYS SPEAK OUTPUT in your Agent communication style with the config `{communication_language}`
- 📐 FOLLOW TURN PROTOCOL from `{project-root}/_mdan/mdan/workflows/special/party-mode/protocols/turn-protocol.md`

## EXECUTION PROTOCOLS:

- 🎯 Run one round at a time, present menu after each round
- ⚠️ Present [N] next round / [Q] question / [D] decide / [E] exit after each round
- 💾 Track current_round in frontmatter
- 📖 Maintain argument log for final summary
- 🚫 FORBIDDEN to skip rounds or rush to decision without user control

## CONTEXT BOUNDARIES:

- Debate topic, agent assignments, and positions from step 1 are available
- Agent manifest data with merged personalities is loaded
- User is the final observer — respect their authority to intervene at any time
- Track arguments made by each side for the arbitration

## YOUR TASK:

Orchestrate the structured debate through sequential rounds, maintaining character consistency and argument quality.

## DEBATE ROUNDS SEQUENCE:

### 1. Round 1 — Opening Statements

"⚔️ **ROUND 1 — DÉCLARATIONS D'OUVERTURE** ⚔️

Kol agent ghadi y'présenti sa position initiale b 2-3 arguments clés. Yallah nbdaw!"

**PROPONENT speaks first:**

"🟢 [Icon] **[Agent Name]** (PROPONENT):

[In-character opening statement with 2-3 key arguments]
- **Argument 1:** [Specific, actionable point with reasoning]
- **Argument 2:** [Specific, actionable point with reasoning]
- **Argument 3:** [Specific, actionable point with reasoning]

[Uses agent's communication style, references their expertise/principles]"

**OPPONENT responds:**

"🔴 [Icon] **[Agent Name]** (OPPONENT):

[In-character opening statement with 2-3 key arguments]
- **Argument 1:** [Specific, actionable point with reasoning]
- **Argument 2:** [Specific, actionable point with reasoning]
- **Argument 3:** [Specific, actionable point with reasoning]

[Uses agent's communication style, references their expertise/principles]"

**ARBITRATOR observation:**

"⚖️ [Icon] **[Agent Name]** (ARBITRATOR):

[Brief observation on the key tension points identified between both positions]"

**After opening statements, present menu:**

"📊 **Round 1 Complete** — {{user_name}}, ash bghiti dir daba?

[N] Next Round — Passer au round dial les rebuttals
[Q] Question — Ts2el chi agent chi soual
[D] Decide — Nta déjà 3refti, skip l l'arbitration
[E] Exit — Khroj men le débat"

### 2. Round 2 — Rebuttal Round

"⚔️ **ROUND 2 — REBUTTALS** ⚔️

Daba kol agent ghadi yredd 3la les arguments dial l'adversaire."

**OPPONENT rebuts PROPONENT:**

"🔴 [Icon] **[Agent Name]** (OPPONENT → réponse l PROPONENT):

[In-character rebuttal addressing PROPONENT's specific arguments]
- **Sur l'argument 1:** [Counter-argument or weakness identified]
- **Sur l'argument 2:** [Counter-argument or concession + but...]
- **Point additionnel:** [New argument strengthening their position]

[Must acknowledge at least one valid point from PROPONENT side]"

**PROPONENT rebuts OPPONENT:**

"🟢 [Icon] **[Agent Name]** (PROPONENT → réponse l OPPONENT):

[In-character rebuttal addressing OPPONENT's specific arguments]
- **Sur l'argument 1:** [Counter-argument or weakness identified]
- **Sur l'argument 2:** [Counter-argument or concession + but...]
- **Point additionnel:** [New argument strengthening their position]

[Must acknowledge at least one valid point from OPPONENT side]"

**ARBITRATOR identifies convergence:**

"⚖️ [Icon] **[Agent Name]** (ARBITRATOR):

[Identifies where arguments converge or remain contested]"

**Present menu again**

### 3. Round 3 — Final Arguments

"⚔️ **ROUND 3 — ARGUMENTS FINAUX** ⚔️

Dernière chance l kol agent yjme3 position dyalo."

**PROPONENT Closing:**

"🟢 [Icon] **[Agent Name]** (PROPONENT — Final):

[Strongest synthesis of their position]
- **L'argument li khra:** [Single most compelling point]
- **Concession honnête:** [What they acknowledge from the other side]
- **Recommandation finale:** [Clear, actionable recommendation]"

**OPPONENT Closing:**

"🔴 [Icon] **[Agent Name]** (OPPONENT — Final):

[Strongest synthesis of their position]
- **L'argument li khra:** [Single most compelling point]
- **Concession honnête:** [What they acknowledge from the other side]
- **Recommandation finale:** [Clear, actionable recommendation]"

### 4. Arbitration

"⚖️ **ARBITRATION** ⚖️"

The Arbitrator delivers a structured ruling:

"⚖️ [Icon] **[Agent Name]** (ARBITRATOR — Ruling):

**Decision:** [Clear statement of the decision]

**Rationale:**
1. [Key reasoning point 1]
2. [Key reasoning point 2]
3. [Key reasoning point 3]

**Confidence:** [High/Medium/Low] ([0-1 score])

**Dissent:** [Summary of the losing side's strongest unresolved argument]

**Conditions/Caveats:** [Any conditions on the decision]"

### 5. Moderator Summary Table

"**📊 Résumé du Débat:**

| | 🟢 PROPONENT ([Agent]) | 🔴 OPPONENT ([Agent]) |
|---|---|---|
| **Argument principal** | [Strongest point] | [Strongest point] |
| **Concession** | [What they admitted] | [What they admitted] |
| **Risque** | [Risk of this choice] | [Risk of this choice] |

**⚖️ Ruling:** [Arbitrator's decision]
**Confidence:** [Score]

**Daba {{user_name}}, la décision f yeddik!**

[D] Decide — Confirmer ou override le ruling
[R] Replay — 3awed chi round
[E] Exit — Khroj bla décision"

### 6. Handle Menu Selections

#### If 'N' (Next Round):
- Increment `current_round` in frontmatter
- Present the next round in sequence
- If all rounds done, go to Arbitration

#### If 'Q' (Question):
- Enter cross-examination mode
- Wait for user question
- Route to relevant agent for in-character response
- Return to menu after response

#### If 'D' (Decide):
- Update frontmatter: `stepsCompleted: [1, 2]`
- Load: `./step-03-decision.md`

#### If 'E' (Exit):
- Load: `./step-04-conclusion.md` with no decision recorded

#### If 'R' (Replay):
- Ask which round to replay
- Re-present that round's arguments

## SUCCESS METRICS:

✅ Structured rounds followed in sequence
✅ Both agents given equal depth and attention
✅ Arguments are specific, actionable, and grounded in expertise
✅ Agents acknowledge valid opposing points (no strawmanning)
✅ User menu presented after each round with clear options
✅ Cross-examination questions handled in-character
✅ Arbitrator summary is balanced and fair
✅ Character consistency maintained throughout all rounds

## FAILURE MODES:

❌ One-sided debate favoring PROPONENT or OPPONENT
❌ Generic or abstract arguments without specifics
❌ Agents ignoring the other side's valid points
❌ Skipping rounds without user consent
❌ Breaking character or mixing agent personalities
❌ Not presenting menu options after rounds

## NEXT STEP:

When user selects 'D', load `./step-03-decision.md` to record and formalize the decision.
When user selects 'E', load `./step-04-conclusion.md` to exit gracefully.


## 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.