Super Hedge Fund Skill

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 (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 frameworks
• references/analysis-methods.md - Detailed scoring and calculation methods
• assets/report-template.md - Markdown report template

Scripts

• scripts/analysts.py - Rule-based analyst implementations
• scripts/investor_prompts.py - Claude investor persona prompts
• scripts/report_generator.py - Markdown report generation
• scripts/data_fetcher.py - Data fetching utilities

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: StanleyChanH
• Source: StanleyChanH/super-hedge-fund-skill
• License: MIT

Install and usage instructions live in the source repository linked above.