The best Claude stock analysis skills in the directory are read-only data skills: one for prices and fundamentals through Yahoo Finance, one for filings from SEC EDGAR, and a research workflow that cross-checks the two. Backtesting skills come next, and skills that can place orders come last because they carry the most risk. This guide sorts them in that order.
For each skill you get a "best for" line, the data source it calls, credentials it needs, its licence as the directory records it, and the result of the automated static check. The details come from each skill's SKILL.md and repository. The editorial team did not run these skills, so nothing here is a claim about the accuracy of their numbers. Browse the stock analysis topic and the trading and backtesting topic for the full lists.
A plain note before you start
Skills are tools, not financial advice. They fetch data, run calculations and format reports, and an agent can misread a filing or a chart as easily as a person can. Backtests describe the past under assumptions about fees, slippage and data quality, and they do not predict returns. Before acting on any output, check the source figures yourself, and speak to a licensed adviser about decisions that involve your money.
Two more cautions apply. Free data sources have usage rules: Yahoo Finance data comes through an unofficial wrapper that can throttle you, and the SEC asks for a descriptive User-Agent and limits request rates. And some platforms restrict automated access in their terms, which matters for the TradingView skills below.
Best Claude stock analysis skills by job
| Skill | Best for | Credentials | Licence | Automated check |
|---|---|---|---|---|
| hkuds/yfinance | Prices and fundamentals | None | MIT | Passed |
| hkuds/sec-edgar | US filings and XBRL figures | None | MIT | Passed |
| firecrawl/financial-research | Cross-checking filings and analyst data | None named | MIT | Passed |
| geeksfino/findata-toolkit-us | Free US data scripts and score calculators | None named | Apache-2.0 | Passed |
| ginlix-ai/initiating-coverage | A full equity research report | None | Apache-2.0 | Passed |
| hkuds/strategy-generate | Idea to backtest | None named | MIT | Passed |
| hkuds/vibe-trading | A broad quant toolkit | An LLM key, optional data keys | MIT | Passed |
| lo2cin4/lo2cin4bt-backtesting | Local backtests without writing code | None | No standard licence recorded | Passed |
"Passed" means the static check reported no findings. It says nothing about the quality of the analysis.
Free data skills: prices, fundamentals and filings
hkuds/yfinance
Best for: quick price history and company data for US, Hong Kong and Canadian securities.
hkuds/yfinance comes from the Vibe-Trading project and is MIT-licensed. It pulls price bars, fundamentals, financial statements, dividends, splits, holdings and insider transactions, plus indices, ETFs and currency pairs, from Yahoo Finance's public data through the open-source yfinance package. It needs no registration or API key.
Requirements: Python with the yfinance package, or the project's built-in client, and network access.
Trade-off: the skill itself notes the limits. Minute bars reach back only a few days to a couple of months, and because the source is an unofficial scraper, heavy request rates can get you blocked for a while.
hkuds/sec-edgar
Best for: primary-source financials for US companies.
hkuds/sec-edgar is MIT-licensed. It resolves tickers to company identifiers, lists recent 10-K, 10-Q and 8-K filings with document links, and pulls XBRL financial series. The SEC asks callers to send a descriptive User-Agent with a contact address, and the bundled client includes a default and throttles requests to respect the SEC's limits.
Requirements: Python and network access to EDGAR. No API key.
Trade-off: it covers US-listed companies only, and the data lags actual submissions. Treat it as a way to retrieve filings, then read the filing yourself.
firecrawl/financial-research
Best for: a short research brief that checks numbers against a second source.
firecrawl/financial-research is MIT-licensed. It finds the latest 10-K or 10-Q, extracts revenue, net income, earnings per share, margins and guidance, pulls analyst ratings and price targets from Yahoo Finance, and flags discrepancies. It treats SEC filings as the source of truth when sources disagree and returns nulls rather than guessing. The copies you may see in three template folders belong to the same repository.
Requirements: web search and page scraping tools, which your agent must supply, for example through a Firecrawl connection.
Trade-off: it handles public companies only and sends private-company, crypto and macro questions elsewhere.
geeksfino/findata-toolkit-us
geeksfino/findata-toolkit-us is Apache-2.0 licensed and bundles free Python scripts for US stock data, SEC filings, insider trades and macro indicators, along with financial score calculators and portfolio analytics. It needs the packages in its requirements file and network access to Yahoo Finance, SEC EDGAR and FRED. It suits people who prefer fixed scripts to an agent improvising code.
For China A-shares, hkuds/akshare and hkuds/eastmoney are MIT-licensed free data skills that need no key, and both passed the check.
Analysis workflows
ginlix-ai/initiating-coverage is Apache-2.0 licensed and produces a first-time research report in five tasks: company research, a financial model, valuation, charts and a Word document. It expects statements or 10-K access and relies on companion skills for Word output and research conventions. It is a drafting aid, and the model's assumptions are yours to check.
wbh604/trap-detector is MIT-licensed and does the opposite job: it scans eight warning signs behind a stock tip from a friend, chat group or online teacher and returns a four-level risk rating with evidence. It depends on web search and the repository's own fetch tools.
xbtlin/earnings-review is MIT-licensed, reads filings and earnings-call material in depth, and rates how complete the sources are before it extracts figures.
Backtesting skills
hkuds/vibe-trading and its sub-skills
Best for: a broad research toolkit with backtesting, factor analysis, prebuilt alphas and options pricing.
hkuds/vibe-trading is MIT-licensed. It needs Python with the vibe-trading-ai package, an LLM API key for its multi-agent swarm, and optional keys for premium data providers, so it is the credential-heaviest skill in this list. The skill text also refers to broker endpoints, so read how it treats orders before connecting any account.
The same repository publishes focused skills you can use alone: hkuds/strategy-generate takes an idea to a config file, a signal engine and a first backtest; hkuds/execution-model adds slippage, delayed fills and market-impact assumptions and is marked never for live orders; and hkuds/risk-analysis measures value at risk, drawdown and stress scenarios.
Trade-off: these skills rely on a backtest engine the project provides. Without it, they are reference text.
lo2cin4/lo2cin4bt-backtesting
Best for: beginners who want to describe a strategy and have an agent configure a local backtest.
lo2cin4/lo2cin4bt-backtesting works with a local checkout of the lo2cin4bt project, which uses Python control code and a Rust engine. The directory records no standard licence, but the repository's README states CC BY-NC 4.0, which prohibits commercial use. The README also says results are research evidence and not investment advice, and it discourages live trading and sharing broker credentials.
Requirements: a checkout with its Rust engine and frontend, a config file and a fair amount of disk space.
tradesdontlie/tradingview-mcp skills
tradesdontlie/strategy-report builds a performance report for a Pine Script strategy already backtested in TradingView. Related skills in the same repository set up charts, scan symbols and run replay practice. They need TradingView Desktop with an active subscription, connected through the repository's MCP server over a local debugging port. The directory lists no standard licence, and the README says MIT for the source code only. Its disclaimer notes that TradingView's terms restrict automated data collection, so read those terms before use.
Skills that can place orders
Some skills connect to brokers and exchanges. qusong0627/tigeropen covers the Tiger Brokers Python SDK under Apache-2.0 and defaults to paper trading, and it needs a Tiger Brokers developer account. qusong0627/futuapi is AGPL-3.0, defaults to simulated trading, and drew an informational finding in the check. Crypto skills such as minara-ai/minara and chrisling-dev/hyperliquid can swap, trade perpetuals and move wallet funds with an API key or private key.
This guide does not recommend any of them. If you try one, use a paper or simulated account, keep keys out of files the agent can share, and require a confirmation step for every order.
How to choose
For research on US stocks, install yfinance and SEC EDGAR first, add financial-research for cross-checks, and stop there until you have a question the data cannot answer. For testing a rule, begin with strategy-generate or lo2cin4bt, and read the licence if you work commercially. For install steps, see the pages for Claude Code and Codex. If your research starts from web pages, the web scraping and search skills comparison is the natural companion, and the top skills list shows what else is popular.
Frequently asked questions
Can a Claude skill give me stock picks or financial advice?
No. A skill is a set of instructions and sometimes scripts that tells an agent how to fetch data or run an analysis. It is a tool, not an adviser, and nothing it produces is a recommendation. Several skills in the directory say this themselves and add rules against giving trading advice. Talk to a licensed professional about decisions involving your money.
Do stock analysis skills need API keys?
It varies. Several of the best free options, such as the yfinance and SEC EDGAR skills, need no key. Larger toolkits accept optional keys for market data providers, and any skill that connects to a broker needs that broker's credentials. Check the requirements line for each skill before you give an agent access to an account.
What is the difference between a stock analysis skill and a backtesting skill?
An analysis skill gathers and interprets information about a company or market, such as prices, filings and ratios. A backtesting skill tests a rule against historical data to see how it would have behaved. Neither predicts the future, and a good backtest can still fail on new data.
Are skills that place trades safe to install?
They carry more risk than read-only skills because mistakes can cost money. The brokerage skills reviewed here default to simulated or paper trading, which is the safer way to start. Read what the SKILL.md says about confirmations, keep credentials out of shared files and test with a paper account first.
Which skill is a good starting point for US stocks?
Start with the yfinance skill for prices and fundamentals and the SEC EDGAR skill for filings. Both are free and need no API key, and they cover different questions. Add a backtesting skill only when you have a specific rule to test.