A Claude/Qwen/Gemini skill for institutional-grade investment research. Connects to live market data via MCP servers to deliver company deep dives, due diligence reports, competitive landscape analysis, market sentiment, and DCF-based intrinsic value estimates — all from natural language prompts.
- Overview
- Repository Structure
- Workflows
- DCF Valuation Subskill
- Example Prompts
- Output Format
- MCP Server Setup
- Installation
- Error Handling
- Limitations
- Disclaimer
This skill gives Claude access to real-time financial data and neural web search, enabling research workflows that typically require Bloomberg terminals, financial databases, and hours of analyst time. It is structured as a parent skill with one subskill:
| Skill | File | Purpose |
|---|---|---|
deep-financial-research |
SKILL.md |
Core research orchestration |
dcf-valuation |
subskills/dcf-valuation/SKILL.md |
Intrinsic value estimation via DCF |
The two skills are designed to work independently or together. A prompt like "Do a deep dive on AAPL with a DCF valuation" will invoke both automatically.
deep-financial-research/
├── README.md
├── LICENSE.txt
├── SKILL.md # Main skill definition
└── subskills/
└── dcf-valuation/
└── SKILL.md # DCF valuation subskill
The main skill supports five research workflows, each triggered by natural language.
Trigger phrases: "deep dive on [ticker]", "research [company]", "investment overview of..."
Pulls together a full investment-ready snapshot:
- Real-time price and key market stats via Financial Datasets MCP
- Recent news, press releases, and market narrative via Exa MCP
- Core fundamentals — revenue, margins, P/E, EV/EBITDA, debt levels
- Insider trading activity (net buys vs. sells, recent filings)
- Analyst estimates, consensus ratings, and price targets
- Synthesized bull case, bear case, and key risk summary
Trigger phrases: "due diligence on [company]", "DD on [ticker]", "red flags for..."
Everything in the Company Deep Dive, plus:
- Litigation and regulatory search: SEC investigations, lawsuits, consent orders
- Accounting integrity checks: restatements, auditor resignations, material weaknesses
- Management background research: prior roles, track record, controversies
- Short interest and institutional ownership trends
- Structured risk and catalyst identification
Trigger phrases: "competitive landscape of [industry]", "compare [company] to peers", "who are the winners in [sector]"
- Industry mapping and market share analysis via Exa
- Side-by-side fundamentals table across all major competitors (P/E, revenue growth, margins, ROE, net debt)
- Recent M&A activity, new entrants, and disruptive developments
- Positioning analysis — identifies leaders, laggards, and companies at inflection points
Trigger phrases: "DCF on [ticker]", "what is [stock] intrinsically worth", "is [company] undervalued", "fair value of..."
Runs the integrated dcf-valuation subskill. See the DCF Valuation Subskill section for full detail.
Trigger phrases: "market sentiment on [ticker]", "what is the street saying about [company]", "analyst consensus for..."
- Mainstream and financial media search via Exa (filtered to bloomberg.com, reuters.com, wsj.com, ft.com, etc.)
- Analyst estimate revisions and rating changes via Financial Datasets
- Insider sentiment (aggregate net buying/selling pressure)
- Structured bullish vs. bearish synthesis
The dcf-valuation subskill performs a full discounted cash flow analysis to estimate intrinsic value per share. It runs as a standalone workflow or is automatically invoked as part of the Company Deep Dive when valuation language is detected.
| Step | Description |
|---|---|
| 1. Data Gathering | 5-year FCF history, balance sheet, market cap, shares outstanding, sector |
| 2. Growth Rate | FCF CAGR calculated, capped at 15%, cross-validated against revenue growth and analyst estimates |
| 3. WACC Estimation | Built from risk-free rate (4.0%), equity risk premium (5.5%), cost of debt (5.5%), with sector adjustments |
| 4. FCF Projection | 5-year forward projection with annual growth decay (5% step-down per year) |
| 5. Terminal Value | Gordon Growth Model, terminal growth rate of 2.5% (GDP proxy) |
| 6. Present Value | All cash flows discounted to today; equity value derived by subtracting net debt |
| 7. Sensitivity Analysis | 3×3 matrix varying WACC (±1%) and terminal growth (2.0–3.0%) |
| 8. Validation | Sanity-checked against reported EV, terminal value ratio (target: 50–80%), and implied P/FCF multiple |
| Sector | Adjustment |
|---|---|
| Technology | +0.5% to +1.0% |
| Energy | +1.0% to +1.5% |
| Healthcare | +0.0% to +0.5% |
| Industrials | +0.0% to +0.5% |
| Consumer Staples | −0.5% |
| Utilities | −0.5% to −1.0% |
| Financials | Cost of equity only (no WACC) |
### DCF Valuation: Apple Inc. (AAPL)
#### Valuation Summary
| Metric | Value |
|---------------------|------------|
| Current Price | $189.42 |
| Fair Value | $214.70 |
| Upside/(Downside) | +13.3% |
| Verdict | Moderately Undervalued |
#### Sensitivity Analysis (Fair Value per Share)
| WACC \ Terminal | 2.0% | 2.5% | 3.0% |
|------------------|--------|--------|--------|
| 8.0% | $231 | $244 | $259 |
| 9.0% | $203 | $215 | $228 |
| 10.0% | $180 | $190 | $201 |
Full output spec is defined in subskills/dcf-valuation/SKILL.md.
| Prompt | Workflow Triggered |
|---|---|
"Do a deep dive on NVDA" |
Company Deep Dive |
"Research Tesla for a long position" |
Due Diligence |
"What's MSFT worth based on DCF?" |
DCF Valuation |
"Analyze the AI chip competitive landscape" |
Competitive Landscape |
"What's the street sentiment on META?" |
Market Sentiment |
"Deep dive AAPL with DCF valuation" |
Deep Dive + DCF (combined) |
"Red flags on NKLA" |
Due Diligence (red flag focus) |
"Is AMZN undervalued right now?" |
DCF Valuation |
Every Company Deep Dive and Due Diligence concludes with a structured summary:
### [Company] ([TICKER]) — Deep Research Summary
Current Price: $X.XX | Market Cap: $XB | As of: [timestamp]
Business Overview:
[1–2 sentences]
Key Metrics:
Revenue $XB (+X% YoY)
Gross Margin X%
P/E Ratio X.x
EV/EBITDA X.x
Net Debt $XB
Recent Developments:
• [Item] — Source (link)
• [Item] — Source (link)
Bull Case: [2–3 arguments]
Bear Case: [2–3 arguments]
Key Risks: [top risks]
Confidence: High / Medium / Low
This skill requires two MCP servers to be connected in your Claude environment.
| Server | Auth Method | Key Capabilities |
|---|---|---|
financial-datasets |
OAuth | Real-time prices, fundamentals, insider trades, analyst estimates, cash flow statements |
exa |
API Key | Neural web search, domain-filtered news, citation-rich results |
| Server | Auth Method | Key Capabilities |
|---|---|---|
lightpanda |
Separate setup | Browser automation for paywalled or JS-rendered sources |
financial-datasetsprovides real-time data during market hours; brief delays possible depending on exchangeexaresults include timestamps — prefer results from the last 30–90 days for rapidly evolving situations- Both servers are auto-discovered if registered in your MCP settings
-
Clone the repository
git clone https://github.com/your-username/deep-financial-research.git
-
Add to your Claude skill configuration
- Point Claude to
SKILL.mdas the root skill file - The
dcf-valuationsubskill atsubskills/dcf-valuation/SKILL.mdis referenced automatically
- Point Claude to
-
Connect MCP servers
- Configure
exawith your API key in MCP settings - Connect
financial-datasetsvia OAuth - Optionally configure
lightpandafor browser automation
- Configure
-
Test the setup
"Do a quick deep dive on AAPL"A successful response will include real-time price data and recent news — confirming both MCP servers are live.
| Scenario | Behavior |
|---|---|
| MCP server unavailable | Notifies user, falls back to Claude's training knowledge with explicit caveats |
| Tool timeout | Retries once; proceeds with available data and flags what's missing |
| Data discrepancy between sources | Notes the discrepancy and cites both values |
| Negative or volatile FCF (DCF) | Switches to analyst estimates or industry averages; documents the substitution |
| Calculated EV off by >30% from reported | Flags the divergence and revisits WACC/growth assumptions before presenting output |
- Not investment advice. This skill produces research analysis, not recommendations. All outputs should be treated as a starting point for further investigation.
- DCF sensitivity. Intrinsic value estimates are highly sensitive to WACC and growth rate assumptions. Small changes in inputs produce meaningfully different outputs — always review the sensitivity table.
- Data freshness. Financial Datasets provides real-time prices during market hours; fundamental data (balance sheets, cash flows) is updated on a reporting lag and reflects the most recent filing.
- High-growth and financial companies. DCF models are less reliable for pre-profitability companies, high-growth compounders, and financial institutions (which require DDM or residual income approaches instead).
- Exa coverage. Exa's neural search excels at English-language sources. Coverage of non-English or regional financial media may be limited.
This repository and the skills it contains are provided for informational and research purposes only. Nothing here constitutes investment advice, a solicitation, or a recommendation to buy or sell any security. Always conduct your own due diligence and consult a licensed financial advisor before making investment decisions.
See LICENSE.txt for full terms.