Value Investing Agent
An AI agent that analyzes US stocks the Graham & Buffett way — intrinsic value, margin of safety, and moat in one report.
The SKILL.md is a pure methodology prompt: no permissions, no scripts, no credentials or sensitive-data handling; it mandates source citation and a disclaimer, giving transparent data flow. Deducted for: it instructs the model to pull from external data providers without disclosing external call scope, failure recovery or rollback; publisher identity is unverified.
Self-consistent formulas (DCF, Graham Number, growth formula) with a clear conservative-estimate rule, and repo vitest suites (dcf/graham/margin-of-safety) cover key calculation paths including negative FCF and zero-value edge cases. However, this is a static review with no executed tests, and the skill's failure feedback for missing/unavailable data is only principled guidance, so it is capped at 10.
Clear audience and five-step framework, with declared boundaries (circle of competence, non-advice disclaimer). Deducted for: no Chinese-language support, core function depends on overseas APIs (yahoo-finance, finnhub, alpha-vantage) with the default provider explicitly prone to 429 rate limiting, questionable mainland-China reachability, and no systematic non-fit/trigger conditions.
Well-layered (principles → framework → report spec → pitfalls) with disclaimer template, data-quality standards and known limitations (over-precision, value traps). Deducted for: no version/changelog within the skill, install/dependency notes live only in the README, and maintenance responsibility/update path is not stated at skill level.
Can produce structured valuation and margin-of-safety reports with reproducible formulas and thresholds. Deducted for: the value claim is essentially a generic investing-methodology prompt with limited evidence of marginal gain over directly asking Claude, and outputs still require user verification of data and assumptions; capped at 7 for static review.
Three test files target the core calculations, formula sources (Graham's works) are traceable, and fact/opinion separation is explicitly required. Deducted for: no executed tests, no CI or third-party execution evidence, and the skill file itself carries thin verifiable claims; conservatively 4 under the static cap of 5.
- This is a static source review; no code or tests were executed.
- The default data provider (Yahoo Finance) is rate-limit prone; finnhub/alpha-vantage require API keys and are overseas services with unverified mainland-China reachability.
- No Chinese-language support; output language must be arranged separately.
- Output is for educational/research purposes only, not investment advice; valuations depend heavily on assumptions and must be independently verified.
- No version number or changelog at skill level; upgrade behavior may be unstable.
What does this skill do, and when should you use it?
Value Investing Agent by danielchu97 ships as a SKILL.md methodology plus a standalone MCP server (the npm package value-investing-agent). The skill file directs a model to analyze stocks following Graham and Buffett: estimate intrinsic value via DCF and the Graham Number, require at least a 25% margin of safety, and evaluate five types of economic moats. The MCP server adds ten tools for real-time quotes, financials, news, a personal watchlist, and report generation, with pluggable data providers (Yahoo Finance by default, Finnhub, Alpha Vantage). It is MIT-licensed and explicitly educational — not investment advice.
The SKILL.md defines a step-by-step analysis workflow: understand the business, evaluate the moat (brand, cost advantage, network effect, switching costs, scale), check financial health against concrete thresholds (ROE >15%, current ratio >1.5, interest coverage >5x), estimate intrinsic value with DCF and Graham formulas, compute the margin of safety, and produce a structured report with mandatory disclaimer. The MCP server exposes tools — get_stock_quote, get_financials, calculate_intrinsic_value, analyze_moat, manage_watchlist, generate_stock_report, and more — that fetch market data, run valuation math, maintain watchlists, and generate daily and deep-dive reports.
- A value investor wanting a quick read on whether a US stock (e.g. AAPL, MSFT) trades with a 25%+ margin of safety
- A multi-position investor generating daily summary reports for a watchlist instead of checking quotes manually
- A student of Graham/Buffett methods using the structured framework to practice valuation and moat analysis
- A conservative investor comparing DCF, Graham Number, and growth-formula outputs and adopting the most conservative figure
- A developer plugging in a custom data provider via the plugin architecture to build a personalized analysis pipeline
What are this skill's strengths and limitations?
- Concrete, complete methodology: explicit valuation formulas, metric thresholds (ROE >15%, gross margin >40%), and report structure rather than vague advice
- Standalone MCP server that starts with a single npx command and works with any MCP-compatible client, not just Claude products
- Pluggable data sources — Yahoo Finance (no key), Finnhub, Alpha Vantage built in, plus a custom provider guide
- Solid engineering: MIT license, test and lint scripts, bilingual README
- Built-in risk discipline: conservative-estimate rule, mandatory disclaimer, and documented pitfalls like value traps
- The default Yahoo Finance provider is rate-limited and can hit 429 errors; stable use requires obtaining Finnhub or Alpha Vantage keys yourself
- DCF outputs are assumption-driven estimates; the skill itself admits they are not precise and cannot serve as actionable investment advice
- Coverage is focused on US stocks (all documented examples are US-listed); non-US market support is not documented
- No published backtesting or accuracy evidence; output quality depends on model reasoning and data quality
- The SKILL.md and MCP server are two distinct parts — installing only the skill file gives no real-time data tools
How do you install this skill?
The SKILL.md methodology can be used on its own without installation. For the MCP tools, the recommended path is adding {"mcpServers":{"value-investing-agent":{"command":"npx","args":["-y","value-investing-agent"]}}} to your MCP client config (e.g. Claude Desktop). Requires Node.js >= 18. Alternatives: global install (npm install -g value-investing-agent) or building from source (git clone, then npm install && npm run build). For reliable data, register a free API key at finnhub.io and set the FINNHUB_API_KEY environment variable. The skill file lives at skills/value-investing/SKILL.md in the repository.
How do you use this skill?
Once configured, trigger tools with natural language in your MCP client: "Get me the current quote for Apple", "Calculate the intrinsic value of AAPL", "Analyze the competitive moat of Coca-Cola", "Add GOOGL to my watchlist in the Technology group", "Generate a daily report for my watchlist", or "Create a detailed analysis report for Berkshire Hathaway". Switch data providers with "Switch to finnhub provider" or the set_provider tool. Used standalone, the SKILL.md serves as analysis guidance, asking the model to follow the prescribed report structure and data-quality standards.
How does this skill compare with similar options?
Positioned as "every value investor deserves their own AI agent," it differs from simply asking a general-purpose AI assistant for a valuation by bundling live market data tools, explicit formulas, and watchlist management; compared to using the SKILL.md alone, the MCP server adds data fetching and report generation at the cost of requiring Node.js and an MCP client setup.