Starter KitAI Broker Docs
Index

Debate Agents System - Agent Prompts Reference

Multi-Agent Debate System - Agent Prompts

This document describes all agents in the multi-agent debate trading system, their roles, and how they contribute to investment analysis.

System Architecture

The debate system uses a multi-stage analysis pipeline:

  1. Data Collection Phase: Market, News, Sentiment, and Fundamentals Analysts
  2. Investment Debate Phase: Bull vs Bear Researchers moderated by Research Manager
  3. Trading Decision: Trader makes action decision
  4. Risk Management: Conservative, Neutral, and Aggressive Risk Managers debate
  5. Final Decision: Portfolio Manager and External Consultant validate

Data Collection Agents

Market Analyst

Role: Technical market analysis and indicator calculation

Key Responsibilities:

  • Fetch OHLCV (Open, High, Low, Close, Volume) data
  • Calculate technical indicators (RSI, MACD, Bollinger Bands, Moving Averages)
  • Identify chart patterns and trends
  • Assess momentum and volatility

Tools Used:

  • get_stock_data_tool: Historical price data
  • get_technical_indicators_tool: Technical analysis metrics

Output: Technical market report with actionable indicators and trend analysis

News Analyst

Role: News sentiment analysis and event monitoring

Version: 2.0

Key Responsibilities:

  • Search and analyze recent news articles about the stock
  • Extract multilingual news from local markets
  • Identify significant corporate events (earnings, M&A, regulatory)
  • Assess sentiment and potential market impact

Tools Used:

  • tavily_search_tool: Web search for news articles
  • get_multilingual_sentiment_search: Local language news sources

Output: News sentiment report with key events, sentiment score, and market impact assessment

Special Features:

  • Multilingual news analysis for international stocks
  • Focus on local-language sources (e.g., Nikkei for Japanese stocks)
  • Identifies "undiscovered" status based on coverage gaps

Sentiment Analyst (Social Analyst)

Role: Social media and retail investor sentiment tracking

Key Responsibilities:

  • Analyze social media discussions (StockTwits, Reddit, Twitter)
  • Gauge retail investor sentiment
  • Identify trending topics and concerns
  • Detect sentiment shifts and momentum

Tools Used:

  • get_stocktwits_sentiment: StockTwits message analysis
  • Social media aggregation tools

Output: Social sentiment report with bullish/bearish ratio, trending topics, and sentiment trends

Fundamentals Analyst

Role: Financial statement analysis and fundamental metrics calculation

Version: 2.5

Key Responsibilities:

  • Analyze balance sheet, income statement, cash flow
  • Calculate financial health score (Piotroski F-Score)
  • Evaluate growth metrics
  • Assess valuation ratios (P/E, PEG, P/B, EV/EBITDA)
  • Check ADR status and analyst coverage
  • Calculate liquidity metrics

Tools Used:

  • get_fundamental_data_tool: Financial statements
  • calculate_liquidity_metrics: Trading liquidity analysis
  • get_analyst_coverage: Analyst coverage check

Output: Comprehensive DATA_BLOCK with:

  • Financial Health Score (0-12 scale using Piotroski F-Score)
  • Growth Score (0-6 scale)
  • Valuation metrics
  • Liquidity analysis
  • ADR status
  • Analyst coverage count

Special Features:

  • Structured JSON output for programmatic validation
  • Strict thesis compliance checking
  • ADR detection to identify "discovered" vs "undiscovered" stocks

Investment Debate Phase

Bull Researcher

Role: Advocate for BUY opportunities with data-driven optimism

Version: 2.3

Thesis Compliance Criteria:

  • Financial health ≥7/12 (preferably ≥8/12 for strong conviction)
  • Growth score ≥3/6 (preferably ≥4/6 for strong conviction)
  • US revenue <25% (or <35% if ≥30% undervalued + ≥3 catalysts)
  • P/E ≤18 OR (P/E 18-25 with PEG ≤1.2)
  • Liquidity >250kdailyaverage(>250k daily average (>100k for small caps)
  • Analyst coverage <6 US/English analysts
  • No US ADR listing (maintains "undiscovered" status)

Key Responsibilities:

  • Build strongest case for upside potential
  • Identify catalysts that could drive price higher
  • Counter bearish concerns with evidence
  • Present best-case scenarios backed by data
  • Acknowledge thesis compliance boundaries

Output Structure:

  1. Thesis Compliance Check: List passing/failing criteria
  2. Bull Case Summary: 2-3 strongest arguments with specific data
  3. Counter to Bear Concerns: Direct responses with evidence
  4. Catalysts: Specific events/factors for upside
  5. Conviction: High/Medium/Low
  6. Recommendation: BUY/HOLD with reasoning

Example Output: ``` THESIS COMPLIANCE: ✓ Financial Health: 9/12 (≥7 required) ✓ Growth Score: 4/6 (≥3 required) ✓ P/E: 16 (≤18 threshold) ✓ ADR Status: None (undiscovered) ✓ Analyst Coverage: 3 (<6 required)

BULL CASE SUMMARY: With a P/E of 16 (well below the 18 threshold) and ROE of 18%, this company offers compelling value. The undiscovered status (only 3 US analysts) combined with strong growth catalysts...

CONVICTION: High RECOMMENDATION: BUY ```

Bear Researcher

Role: Identify risks and thesis violations with rigorous analysis

Version: 2.4

Focus Areas:

Quantitative Hard Fails:

  • Financial health <7/12
  • Growth score <3/6
  • US revenue >35%
  • P/E >18 without PEG ≤1.2
  • P/E >25 (always overvalued)
  • Liquidity <$100k daily average
  • Analyst coverage ≥6 US/English analysts
  • ADR exists on NYSE/NASDAQ/OTC

Qualitative Risks:

  • Technological lag (e.g., legacy automaker late to EVs)
  • Eroding competitive moat
  • Cyclical peak risk
  • Jurisdiction risks (authoritarian governments, capital controls)
  • Growth story mismatch
  • Market saturation/oversupply

Key Responsibilities:

  • Flag thesis violations explicitly with exact numbers
  • Identify structural headwinds
  • Challenge bullish arguments
  • Present worst-case scenarios with data
  • Detect cognitive biases in analysis

Output Structure:

  1. Bear Case Summary: Start with thesis violations, then 2-3 strongest risks
  2. Counter to Bull Arguments: Direct rebuttals with evidence
  3. Key Risks:
    • Thesis Violations
    • Qualitative Risks
    • Quantitative Concerns
  4. Conviction: High/Medium/Low
  5. Recommendation: SELL/HOLD with reasoning

Example Output: ``` BEAR CASE SUMMARY: This stock violates the thesis on valuation: P/E is 22 (vs. threshold of 18) with PEG of 1.5 (above 1.2 threshold). Additionally, the company has an ADR (TICKER: XYZ) on NYSE, violating the undiscovered criterion...

KEY RISKS:

  • Thesis Violations: P/E=22 (>18), ADR exists, Analyst coverage=8 (>6)
  • Qualitative Risks: Cyclical Peak, Eroding Moat
  • Quantitative Concerns: High leverage, Declining margins

CONVICTION: High RECOMMENDATION: SELL ```

Research Manager

Role: Judge the Bull vs Bear debate and make BUY/HOLD/SELL decision

Version: 2.2

Key Responsibilities:

  • Synthesize Bull and Bear arguments
  • Evaluate evidence quality and logic
  • Check thesis compliance against DATA_BLOCK
  • Make final investment recommendation
  • Provide confidence level and reasoning

Decision Framework:

  • BUY: Thesis compliance ≥80%, strong catalysts, Bull case stronger
  • HOLD: 60-79% thesis compliance or balanced arguments
  • SELL: Hard thesis violations, Bear case stronger, high risks

Output Structure:

  1. Summary of Debate: Key points from both sides
  2. Thesis Compliance Assessment: Percentage score
  3. Decision: BUY/HOLD/SELL
  4. Confidence: High/Medium/Low
  5. Reasoning: Why this decision, which arguments were more compelling

Trading Decision

Trader

Role: Convert research recommendation into actionable trade plan

Key Responsibilities:

  • Translate BUY/HOLD/SELL into specific actions
  • Consider market conditions and timing
  • Recommend entry/exit points
  • Suggest position sizing based on confidence
  • Plan risk management (stop-loss, take-profit)

Output: Trading plan with action, timing, position size, and risk parameters

Risk Management Debate

Conservative Debator (Safe Analyst)

Role: Advocate for minimal risk exposure

Key Responsibilities:

  • Emphasize downside protection
  • Recommend smaller position sizes
  • Suggest tight stop-losses
  • Highlight worst-case scenarios
  • Prioritize capital preservation

Neutral Debator (Neutral Analyst)

Role: Balanced risk perspective

Key Responsibilities:

  • Weigh risk vs reward objectively
  • Recommend moderate position sizes
  • Balance upside and downside scenarios
  • Suggest standard risk management practices

Aggressive Debator (Risky Analyst)

Role: Argue for higher risk for higher returns

Key Responsibilities:

  • Emphasize upside potential
  • Recommend larger position sizes
  • Focus on conviction rather than caution
  • Highlight asymmetric risk/reward favoring upside

Final Validation

Portfolio Manager

Role: Final gatekeeper for investment decisions

Version: 2.3

Key Responsibilities:

  • Hard veto on thesis violations (P/E>25, ADR exists, coverage≥6)
  • Validate thesis compliance with exact numbers from DATA_BLOCK
  • Assess portfolio fit and diversification
  • Enforce risk limits
  • Make final GO/NO-GO decision

Veto Powers:

  • Automatic REJECT if P/E >25
  • Automatic REJECT if ADR exists on US exchange
  • Automatic REJECT if analyst coverage ≥6
  • Automatic REJECT if financial health <7 or growth <3

Output: Final decision with enforcement of thesis boundaries

External Consultant

Role: Independent validation and bias detection

Version: 1.0

Key Responsibilities:

  1. Fact-Check Source Data: Cross-reference claims against DATA_BLOCK
  2. Detect Cognitive Biases:
    • Confirmation bias
    • Anchoring bias
    • Recency bias
    • Groupthink
    • Survivorship bias
  3. Challenge the Synthesis: Review Research Manager's logic

Output Structure: ``` CONSULTANT REVIEW: [APPROVED / CONDITIONAL APPROVAL / MAJOR CONCERNS]

SECTION 1: FACTUAL VERIFICATION Status: [✓ FACTS VERIFIED / ✗ ERRORS FOUND]

SECTION 2: BIAS DETECTION Status: [✓ NO BIASES / ⚠ BIASES IDENTIFIED] Detected Biases:

  • [Bias Type]: [Evidence and Impact]

SECTION 3: SYNTHESIS EVALUATION Research Manager Recommendation: [BUY/HOLD/REJECT] Consultant Assessment: [✓ AGREE / ✗ DISAGREE / ⚠ RESERVATIONS]

FINAL CONSULTANT VERDICT: [Overall assessment] ```

Special Features:

  • Uses different AI model (OpenAI) to cross-validate Gemini outputs
  • Independent of organizational biases
  • Focus on material issues that could change decision

Investment Thesis Summary

The multi-agent system enforces a value-to-growth ex-US equities thesis:

Core Criteria:

  • Valuation: P/E ≤18 ideal, 18-25 acceptable if PEG≤1.2, >25 rejected
  • Quality: Financial health ≥7/12, Growth ≥3/6
  • Geography: US revenue <25-35%, no ADR listing
  • Discovery: Analyst coverage <6, low liquidity acceptable for small caps
  • Accessibility: IBKR-accessible jurisdictions

Emphasized Attributes:

  • Undervaluation >25%
  • ROE ≥15%
  • FCF yield ≥4%
  • Local non-English catalysts

This thesis ensures the system focuses on undiscovered, high-quality value opportunities in international markets.

Usage Examples

Example 1: Basic Analysis

```bash curl -X POST http://localhost:3000/api/groq-debate
-H "Content-Type: application/json"
-d '{ "symbol": "AAPL" }' ```

Example 2: Extended Debate

```bash curl -X POST http://localhost:3000/api/groq-debate
-H "Content-Type: application/json"
-d '{ "symbol": "TSLA", "date": "2024-12-19", "max_debate_rounds": 3, "llm_provider": "groq" }' ```

Example 3: Custom Models

```bash curl -X POST http://localhost:3000/api/groq-debate
-H "Content-Type: application/json"
-d '{ "symbol": "NVDA", "llm_provider": "anthropic", "deep_think_llm": "claude-3-5-sonnet-20241022", "quick_think_llm": "claude-3-5-haiku-20241022" }' ```

Example 4: JavaScript/TypeScript

```typescript async function analyzeStock(symbol: string) { const response = await fetch('/api/groq-debate', { method: 'POST', headers: { 'Content-Type': 'application/json' }, body: JSON.stringify({ symbol, date: new Date().toISOString().split('T')[0], max_debate_rounds: 2 }) })

const result = await response.json()

if (result.success) { console.log('Bull Arguments:', result.analysis.bull_arguments) console.log('Bear Arguments:', result.analysis.bear_arguments) console.log('Final Decision:', result.analysis.final_decision) console.log('Confidence:', result.analysis.confidence_level) console.log('Reasoning:', result.analysis.reasoning) }

return result }

// Usage const analysis = await analyzeStock('AAPL') ```

Response Format

```json { "success": true, "symbol": "AAPL", "date": "2024-12-19", "analysis": { "bull_arguments": [ "Strong fundamentals with 15% YoY revenue growth", "Leading market position in premium smartphones", "Services revenue growing 20% annually" ], "bear_arguments": [ "P/E ratio of 28 above historical average", "iPhone revenue slowing in key markets", "Regulatory pressure in EU and China" ], "final_decision": "BUY", "confidence_level": "Medium-High", "reasoning": "Bull case stronger based on fundamentals...", "risk_assessment": "Medium risk with controlled position sizing", "thesis_compliance": { "financial_health": "9/12", "growth_score": "4/6", "pe_ratio": 28, "adr_status": "NYSE: AAPL", "analyst_coverage": 45, "compliance_percentage": 45 } }, "debate_history": [ { "agent": "bull_researcher", "message": "...", "timestamp": "2024-12-19T10:00:00Z" } ], "metadata": { "llm_provider": "groq", "deep_think_model": "llama-3.3-70b-versatile", "quick_think_model": "llama-3.1-8b-instant", "debate_rounds": 2, "total_tokens": 15420, "execution_time_ms": 8500 } } ```

Agent Versions

  • Bull Researcher: v2.3
  • Bear Researcher: v2.4
  • Research Manager: v2.2
  • Portfolio Manager: v2.3
  • Fundamentals Analyst: v2.5
  • News Analyst: v2.0
  • External Consultant: v1.0

All agents updated to align with the value-to-growth ex-US equities thesis with explicit P/E thresholds and ADR violation detection.

Investment Brokers

Previous Page

Fundamentals Indicators

Next Page

On this page

Multi-Agent Debate System - Agent Prompts
System Architecture
Data Collection Agents
Market Analyst
News Analyst
Sentiment Analyst (Social Analyst)
Fundamentals Analyst
Investment Debate Phase
Bull Researcher
Bear Researcher
Research Manager
Trading Decision
Trader
Risk Management Debate
Conservative Debator (Safe Analyst)
Neutral Debator (Neutral Analyst)
Aggressive Debator (Risky Analyst)
Final Validation
Portfolio Manager
External Consultant
Investment Thesis Summary
Usage Examples
Example 1: Basic Analysis
Example 2: Extended Debate
Example 3: Custom Models
Example 4: JavaScript/TypeScript
Response Format
Agent Versions
Stocks
Predict
Watch
Copy
Portfolio
Settings