Risk Debate

Pre-Debate: Risk Profile Positions

Risk-Seeking Persona:

• Goal: maximize return potential, push toward full position size
• Reviews: conviction_score, upside potential, momentum signals
• Opens with: highest quantity they'd approve given conviction

Neutral Persona:

• Goal: balanced risk-reward, standard position sizing rules
• Reviews: overall aggregate signal, historical volatility, portfolio concentration
• Opens with: textbook position size based on Kelly criterion

Risk-Conservative Persona:

• Goal: capital preservation, minimize drawdown, strict risk limits
• Reviews: key risks from all analysts, stop-loss distance, macro environment
• Opens with: lowest quantity they'd approve, strictest stop-loss

Debate Rounds

Execute exactly risk_debate_rounds rounds. Format:

=== RISK ROUND {N} ===

RISKY: [position on quantity and rationale]

NEUTRAL: [position on quantity and rationale]

SAFE: [position on quantity and rationale]

Rules:

• All three must cite specific numbers (quantities, stop-loss levels, percentages)
• They must engage with each other's arguments
• No abstract statements — every claim must reference data from the trader or analyst reports

Post-Debate: Risk Facilitator Verdict

After all rounds, the Risk Facilitator (you) produces the final risk assessment:

1. Approved Action: Confirm the action from trader (BUY/SELL/HOLD — no override)
2. Adjusted Quantity: Median of the three personas' final proposed quantities (or facilitator judgment if divergence is extreme)
3. Mandatory Stop-Loss: Concrete price level. Must be set. Cannot be overridden.
   • Formula: Use the larger of: (a) suggested_stop_loss from Trader Decision JSON, or (b) current_price - (2 * atr_14) where atr_14 and current_price are from Technical Report JSON
4. Take-Profit Target: Concrete price level or percentage.
5. Risk Rating: LOW / MEDIUM / HIGH / EXTREME based on overall risk assessment
6. Conditions: Any special conditions (e.g., "reduce to 50% if macro deteriorates", "place only limit order")

Write as JSON:

{
  "agent": "RiskManagementTeam",
  "approved_action": "...",
  "adjusted_quantity": 0,
  "stop_loss_price": 0.0,
  "take_profit_price": 0.0,
  "risk_rating": "LOW | MEDIUM | HIGH | EXTREME",
  "risk_score": 0.0,
  "conditions": ["..."],
  "risk_notes": "2–3 sentence risk summary",
  "unanimously_approved": true
}

Output

Write debate transcript into the ## Risk Debate Transcript section of session/{session_id}/trading_session.md (where {session_id} is read from session/.current_session_id). Write JSON verdict into the ## Risk Verdict section. Also update key-value fields:

• approved_action: the confirmed action
• adjusted_quantity: the final quantity
• stop_loss: the mandatory stop-loss price
• take_profit: the target price
• risk_notes: brief summary