Finance Skills: a plugin collection that turns an agent into a research desk

A collection of skills, not an application

The repository is described as a collection of agent skills for financial analysis and trading, following the Agent Skills open standard, and it is worth being precise about what that means. There is no application here. There is no server, no database, no UI. The repository ships markdown skills and plugin manifests that an agent loads at request time, and the value of the whole thing is in the quality of the instructions inside each skill rather than in any code path.

That framing explains the shape of the repository tree. `plugins/` holds the skill groups, `.claude-plugin/` holds a Claude Code plugin manifest, `opencli-plugin.json` and `opencli-plugins/` suggest a second distribution channel, `CLAUDE.md` sits at the root as agent instructions, and `package.json` is almost empty: it is private, it declares pnpm as the package manager at version 11.22.0, and its only script is a version bump tool called ccbump. There is no build step because there is nothing to build.

It also explains the numbers. At 3,332 stars, 379 forks and zero open issues, with MIT licensing and JavaScript as the primary language, the repository is far more starred than forked, which fits a package people install and read rather than one they build on.

Installing the whole set or one skill

The README gives four install paths and the difference between them matters. The broad one installs plugin groups and prompts you to pick:

npx plugins add himself65/finance-skills

If you would rather not be asked, the same command takes a flag:

npx plugins add himself65/finance-skills --yes

The second path installs a single named skill instead of a group, and the example uses one from the market analysis set:

npx skills add himself65/finance-skills --skill yfinance-data

The third path targets a different agent by name, which is what makes the collection portable rather than tied to one assistant:

npx skills add himself65/finance-skills -a <agent-name>

The README heads this section with plugin groups for Claude Code, Codex, Cursor and more. A financial research kit that can be pointed at several agents is a different proposition from a Claude extension, and the per-skill install path means you can adopt one skill without taking the rest.

Market analysis is the deepest group

The `finance-market-analysis` group holds eleven skills and is where the actual modelling lives. The README describes the group as stock analysis, earnings, estimates, correlations, liquidity, ETFs, options payoff and trading strategies, all via yfinance as the data layer.

The individual entries are more specific than the group description. Company valuation is described as DCF plus relative plus sum-of-the-parts triangulation, producing an implied share price with WACC times growth sensitivity and bull, base and bear scenarios. There are separate pre-earnings and post-earnings skills, an estimate analysis skill covering revision trends and historical analyst accuracy, an ETF premium skill comparing market price against NAV, and an options payoff skill with interactive charts.

The rest of the group is closer to trading mechanics than valuation: a liquidity skill covering spreads, volume profiles, market impact and the Amihud ratio, a correlation skill for sector peers and pair-trading candidates, a SaaS valuation compression skill built around ARR multiples, and a SEPA strategy skill built on Minervini's trend template and VCP patterns with entry points and position sizing. Reading these eleven descriptions is a faster way to understand the project's ambition than any summary would be.

Readers and data providers are mostly thin adapters

Two groups are about getting information in rather than analysing it, and both lean on external projects rather than reimplementing anything. The `finance-social-readers` group covers Twitter and X, Discord, LinkedIn, Telegram, Y Combinator and a generic fallback for more than a hundred other sources. The individual readers are named after their mechanism: Discord, LinkedIn and Twitter read through opencli, Telegram reads through tdl, and Y Combinator data comes from a community API.

The fallback reader is the interesting design choice. Rather than shipping a hundred separate integrations, the generic opencli reader inherits a hundred adapters from another project, with Yahoo Finance, Bloomberg, Reuters, Eastmoney, Xueqiu, Reddit, HackerNews, Substack and arXiv named as examples. The same dependency shows up in the data providers group, where TradingView and Hyperliquid data are also read through opencli.

The `finance-data-providers` group is where credentials and rate limits become your problem. It carries a sentiment skill pulling from Reddit, X, news and Polymarket through an API, a Fintel skill for short interest, borrow rates, fails-to-deliver, 13F ownership and insider transactions, a Hormuz Strait monitoring skill, a TradingView skill backed by a bundled headless MCP server, and a Hyperliquid reader for perpetual and spot markets including funding and order book data. Every one of these needs a key or a paid tier.

The two groups that are not about finance

A generative UI group and a skill creator group sit alongside the financial ones, and both are meta. `finance-ui-tools` contains a single skill, generative-ui, described as a design system for rendering interactive HTML and SVG widgets in Claude conversations through a show_widget hook. A skill whose purpose is to make an agent draw something means the same interface is used for research results as for analysis.

`finance-skill-creator` contains one skill as well, and it is the one to look at if you intend to contribute. It is described as a tool for creating, evaluating and iterating on agent skills, with structured guidance, quality scoring and best-practice enforcement, and it evaluates existing skills against a ten-dimension rubric. A project that ships a rubric for judging its own artifacts is making a claim about how it wants to be extended.

The last group, `finance-startup-tools`, is smaller in scope and broader in audience: a single startup-analysis skill written from three viewpoints at once, a venture capital investor, a job applicant and a chief executive. That framing is unusual for a finance tool, and it suggests the project is aimed at people making decisions about companies rather than only people trading them.

Releases track upstream integrations

Three recent releases show what actually changes. v9.1.0 on 2026-07-21 added a single skill, fintel-data, described as Fintel through its REST API and its own MCP server. That is the whole release, and it is the clearest evidence of how this project grows: one data source at a time, each with its own documented integration path.

v10.0.0 on 2026-08-15 was a larger housekeeping release. It updated dependencies and CI action versions, refreshed upstream integrations and skill content, and fixed cross-shell dynamic probe fallbacks. That last change came from a first-time contributor, which is worth noticing in a repository whose other two releases were authored by the maintainer alone.

v11.0.0 followed on 2026-08-27 with no description beyond a full changelog link, so the repository pushed on 2026-08-27 and has not moved since as of the recorded 2026-10-06 date. The pattern across all three is that changes are about staying current with other people's APIs. The README also carries a warning at the very top that the project is for educational and informational purposes only, that nothing in it is financial advice, and that you should do your own research and consult a qualified advisor.

Editorial conclusion

Finance Skills is interesting less as a finance product than as a packaging decision. It takes skills that would each be a small standalone project and ships them as one installable plugin group set, which means the hard part for a user is choosing which groups to trust with API keys rather than assembling tooling by hand. The market analysis group is the deepest, with eleven skills including DCF triangulation, SEPA setups and options payoff charts, and the social readers group is the widest, reaching a hundred sources through a single fallback reader. The documentation lives on a separate site rather than in the repository, and a personal finance disclaimer sits at the top of the README. Start with the plugin installer and select only the groups you actually need, read the market analysis group before installing anything, and remember that a zero open issue count on a repository this size is a sign of low traffic rather than perfection.

Frequently asked questions

Official sources

1. himself65/finance-skills on GitHub
2. License: MIT
3. Project website
4. README
5. Releases