Claude Stock Analysis Skills: Official Plugins, Community Repos, and How to Use Them

The best Claude stock analysis skills today come from two places: Anthropic's own financial-services plugins, which cover comps, DCF models, and earnings notes, and a small set of community GitHub repos built for individual investors. Each one is a folder of instructions and scripts that Claude loads when a task calls for it. None of them can tell you what to buy.

Here's the short version:

• Anthropic's financial-services repo is the official option. It's Apache-2.0, aimed at analysts, and installs into Claude Code as plugins.
• Anthropic's xlsx skill is the quiet workhorse. It builds spreadsheet models that use real formulas instead of pasted numbers.
• Community repos like claude-trading-skills and XVARY's stock research skill add screeners, journaling, and quick one-page company reports. Both are MIT-licensed.
• SEC EDGAR is the free primary source every one of them should be checked against.

Not financial advice. This article explains how software tools work. It does not recommend any security, strategy, or trade, and nothing here predicts returns. Every number a language model gives you can be wrong, so check it against the company's own filings before you rely on it.

Quick comparison of Claude finance skills

If you're new to the format itself, Taku's guide to Claude skills explains what a skill is and how Claude decides to load one.

Anthropic's official Claude financial analysis skill set

Anthropic sells Claude for Financial Services to banks, insurers, asset managers, and fintech firms. The solutions page describes it as a bundle: Claude's models, Claude for Microsoft 365 inside Excel and PowerPoint, data connectors, and agent templates. The original July 2025 announcement pitched it at analysts and named Claude Code and Claude for Enterprise as the delivery routes.

That's the enterprise product. The part you can read, fork, and install yourself is the open repo behind it.

What's in the financial-services repo

The anthropics/financial-services repo (the old financial-services-plugins URL redirects there) is Apache-2.0 and ships two layers:

• Named agents that run a whole workflow. Examples include a Market Researcher (sector to peer comps), an Earnings Reviewer (earnings call and filings to model update to note draft), and a Model Builder (DCF, LBO, 3-statement, comps in Excel).
• Vertical plugins that hold the underlying skills and slash commands. The core financial-analysis plugin carries comps, DCF, LBO, 3-statement models, deck QC, and an Excel model audit. The equity-research plugin adds earnings notes, initiation reports, model updates, a thesis tracker, and a catalyst calendar.

The README is blunt about scope. It says nothing in the repo is investment advice, and that the agents draft analyst work product for a qualified professional to review. They don't make recommendations or execute transactions.

Installing it in Claude Code

The README gives this path for Claude Code:

claude plugin marketplace add anthropics/financial-services
claude plugin install financial-analysis@claude-for-financial-services
claude plugin install equity-research@claude-for-financial-services

After that, the skills load on their own when a task fits, and commands like /comps, /dcf, and /earnings show up in your session. The repo also lists a Managed Agents API route for teams that want to run the same agents behind their own systems.

One catch: the data connectors (FactSet, Morningstar, S&P Global, PitchBook, and others) point at paid providers. The README notes that MCP access may need a subscription or API key from each one. Without them, the skills still work on files you supply.

The xlsx skill: the base layer for Claude code stock analysis

Most stock analysis ends in a spreadsheet, which makes Anthropic's xlsx skill more useful than its plain name suggests. It lives in the anthropics/skills repo alongside the other document skills. The repo's README calls these source-available rather than open source, and the skill's own licence file points to your agreement with Anthropic.

Three rules in its SKILL.md matter for finance work:

• Use formulas, never hardcoded results. A total is written as a SUM, so the model recalculates when an input changes.
• Document every assumption and hardcoded number. The skill's own example cites a source in the form "Company 10-K, FY2024, Page 45, Revenue Note" with an SEC EDGAR URL.
• Zero formula errors. A recalculation script runs before the file ships, and any error has to be fixed first.

That second rule is the one to lean on. If every input cell names the filing and page it came from, checking the model takes minutes instead of an afternoon.

If you'd rather not build models in a terminal at all, Taku's roundup of AI tools for Excel covers the spreadsheet-native options.

Community Claude stock analysis skills on GitHub

Search GitHub and you'll find plenty of repos with "stock" and "Claude" in the name, and many are thin prompts. These two have real READMEs, a stated licence, and working install instructions. Everything below comes from their own documentation.

tradermonty/claude-trading-skills

claude-trading-skills is an MIT-licensed toolkit for what its author calls time-constrained individual investors. The README says its goal isn't to outsource buy and sell decisions to AI. It aims to structure market review, risk management, trade planning, and journaling instead. It also states plainly that it's not a signal service or a promise of profitability.

What it ships:

• Workflows with named steps, such as a 15-minute daily market check, a weekly portfolio review, and a monthly performance review.
• Skills grouped by area: market regime, core portfolio, swing setups, trade planning, trade memory, and strategy research.
• A no-API starter path of five skills (market breadth, uptrend analysis, position sizing, a local YAML trade journal, and a post-mortem framework) that run on public CSVs and local files.

Other skills need FMP, FINVIZ Elite, or Alpaca. The README marks which integration each skill needs and treats skills-index.yaml as the canonical list. To use it in Claude Code, you copy a skill folder into ~/.claude/skills/, as described in the Claude Code skills docs. For the Claude web app, the repo builds .skill packages you upload under Skills.

xvary-research/claude-code-stock-analysis-skill

XVARY's stock analysis skill is a smaller, MIT-licensed Claude Code skill with three commands:

Its Python tools pull SEC XBRL data from EDGAR and quote data from Yahoo Finance, falling back to Finviz or Stooq. It installs as a plain skill folder or through a Claude Code plugin marketplace.

Be clear about what it is, though. The README calls the repo "the open skill layer" of XVARY Research, a commercial service, and says its prompts and scoring formulas stay proprietary. So the scores come from a method you can read about but can't fully inspect. Treat its verdict line as one input to read critically, not a conclusion.

How to use a Claude finance skill without trusting it blindly

Here's a workflow that works with any of the skills above. It's built around one idea: the filing is the source of truth, and Claude is the fast assistant that drafts around it.

1. Start from the filing, not from memory

Ask Claude to pull figures from documents, not to recall them. EDGAR is the SEC's free filing system for U.S. public companies, and it holds 10-Ks, 10-Qs, 8-Ks, and the XBRL data behind them. Download the filing yourself or let a skill like XVARY's fetch it, then point Claude at that file.

A model asked "what was this company's revenue last year?" with no document in front of it may answer from training data. That can be out of date, or simply wrong.

2. Make every number carry its source

Tell Claude to follow the xlsx skill's rule: each hardcoded input gets a note with the filing, period, and page. Something like:

Build a 3-statement model from the attached 10-K. Every input cell needs a comment with the form type, fiscal year, and page number it came from. Use formulas for everything else.

3. Spot-check the inputs

Pick five or six inputs, including revenue, net income, total debt, and share count, and look each one up in the filing. If any is off, check the rest. Watch for the common slips:

• Units. Thousands vs millions.
• Periods. Fiscal year vs calendar year, or a quarter mistaken for a full year.
• Restatements. A later filing may revise an earlier number.
• Share counts. Basic vs diluted.

4. Keep conclusions out of the tool's hands

Skills are good at structure: pulling numbers, laying out comps, drafting a summary of what management said. They're not a judgement you can hand off. Both official and community READMEs say the same thing in different words: a human reviews the output and makes the call.

5. Save the setup that worked

Once a prompt and skill combination produces a model you trust, save it as a reusable instruction rather than rewriting it every time. That's the whole point of a skill. Taku's guide to Claude skills for research covers the same habit for literature reviews and source checking.

Where these skills stop and others start

Stock analysis skills look outward at public companies. If you're looking inward at your own business, such as recurring revenue, churn, CAC, and LTV, that's a different job with different inputs. The SaaS Metrics Health Check skill is built for that one.

Pick by question:

Key takeaways

• The official starting point is Anthropic's financial-services repo. It's Apache-2.0, installs into Claude Code as plugins, and drafts analyst work for human review.
• The xlsx skill matters more than it looks, because it forces formulas and source notes into every model.
• claude-trading-skills and XVARY's skill are MIT-licensed community options with very different scopes. Read their READMEs before you install.
• SEC EDGAR is free, and it's the place to check every number a model hands you.
• None of these tools give investment advice, and none of them should be treated as if they do.

If installing plugins and copying skill folders is the step where you stall, Taku is an AI-native desktop workspace where you can mirror a working setup and run it without rebuilding the environment first. Taku is in Beta, and the Mac app is available now.

FAQ

What are Claude stock analysis skills?

They're skill folders, meaning instructions plus optional scripts, that teach Claude a repeatable finance task. Examples include building a comps table, drafting an earnings note, or summarizing a company's filings. Claude loads them automatically when your request matches.

Is there an official Claude finance skill?

Yes. Anthropic publishes the financial-services repo on GitHub under Apache-2.0. Its core financial-analysis plugin covers comps, DCF, LBO, and 3-statement models, and the equity-research plugin covers earnings and coverage work.

Can I use Claude Code for stock analysis?

Yes. Claude Code can install the official plugins, run community skills, and execute the Python tools some of them include. You still need to verify the output against primary filings.

Do I need paid data to use these skills?

Not always. The official connectors point at paid providers, but the skills also work on files you supply. claude-trading-skills has a no-API starter path, and XVARY's skill uses free public sources like SEC EDGAR.

Can a Claude skill tell me which stocks to buy?

No. The official repo says it makes no investment recommendations, and the community READMEs say the same. Use these skills to organize research, then make your own decisions or consult a licensed professional.

How accurate are the numbers Claude produces?

They can be wrong. Models misread units, mix up periods, or fill gaps from memory. Ask for a source on every input and spot-check the key figures against the original filing.