Super Hedge Fund Skill
Multi-agent stock analysis system combining rule-based analytics with Claude-powered investor personas.
⚠️ Educational purposes only - NOT investment advice
Workflow
digraph workflow {
rankdir=TB;
node [shape=box, style="filled,rounded", fillcolor="#f0f0f0"];
input [label="1. Parse Input\nExtract tickers"];
data [label="2. Fetch Data\nPrice/Financials/News"];
agents [label="3. Run Agents\nRule + Claude"];
risk [label="4. Risk Analysis\nVolatility/Position"];
output [label="5. Generate Report\nMarkdown"];
input -> data -> agents -> risk -> output;
}
Agents Quick Reference
| Agent | Type | Focus |
|---|---|---|
| Fundamental | Rule | ROE, margins, debt, growth |
| Technical | Rule | EMA, RSI, MACD, momentum |
| Valuation | Rule | DCF, Owner Earnings |
| Sentiment | Rule+Claude | News, insider trades |
| Buffett | Claude | Moat, ROE, intrinsic value |
| Wood | Claude | Disruptive tech, growth |
| Burry | Claude | Deep value, contrarian |
| Lynch | Claude | PEG, understandable biz |
Execution
Step 1: Parse Input
Extract: ticker symbols, date range, capital
Mode: full (default) or brief
Step 2: Fetch Data
Use WebSearch for:
- Current price & 52-week range
- Financial metrics (ROE, P/E, margins, debt)
- Recent news headlines
Step 3: Run Agents
Rule-based (deterministic):
- Fundamental analysis → signal + confidence
- Technical analysis → signal + confidence
- Valuation analysis → signal + confidence
Claude-powered (interpretive):
- For each investor persona, analyze with their philosophy
- Return: {signal, confidence, reasoning}
Step 4: Risk Analysis
Calculate: annual volatility from price data
Determine: risk level → position limit
- Low (<15%): 25% max
- Medium (15-30%): 20% max
- High (30-50%): 15% max
- Very High (>50%): 10% max
Step 5: Aggregate & Output
Count signals: bullish / bearish / neutral
Determine consensus by majority
Generate Markdown report
Signal Icons
| Signal | Icon |
|---|---|
| Bullish | 🟢 |
| Bearish | 🔴 |
| Neutral | 🟡 |
Common Mistakes
| Mistake | Fix |
|---|---|
| Giving real investment advice | Always add disclaimer |
| Missing data errors | Use fallback estimates |
| Single-agent reliance | Must aggregate 8+ signals |
| Overconfident signals | Show confidence %, acknowledge uncertainty |
References
references/investor-agents.md- Investor persona prompts and frameworksreferences/analysis-methods.md- Detailed scoring and calculation methodsassets/report-template.md- Markdown report template
Scripts
scripts/analysts.py- Rule-based analyst implementationsscripts/investor_prompts.py- Claude investor persona promptsscripts/report_generator.py- Markdown report generationscripts/data_fetcher.py- Data fetching utilities