Show Me The Money ships agent skills and an installer, not a business

What ships is a skills directory and a postinstall script

It helps to be clear about the artefact before reading the marketing, because the shape of the repository does not match the tone of the description.

The package is @orrisai/show-me-the-money at version 2.8.0, published on npm, and the package.json shows exactly what is inside it. The files array lists four things: skills/, bin/, install.js and VERSION. The bin entry maps the show-me-the-money command to ./bin/cli.js, and the only script is a postinstall that runs node install.js.

The top-level entries match. There is a skills/ directory, which is the substance, a bin/ directory with the CLI, a .claude-plugin/ directory for the plugin manifest, install.js, package.json, a VERSION file, a LICENSE, a COMMERCIAL-LICENSE.md, a PRD.md, a CHANGELOG.md with a Chinese counterpart, and both an English and a Chinese README.

So the deliverable is a set of instructions that an agent runtime loads and follows, plus the glue that puts them where the runtime expects to find them. There is no application, no service, no database, and no dependency list beyond what the install script needs. The primary language is recorded as JavaScript, and that is the install and CLI code, not the skills.

The README says the suite is an open-source skill suite for Claude Code and compatible AI coding agents, and that it works with Claude Code, Codex CLI, Gemini CLI, and other agents that support the skill system. That is the deployment target in one sentence, and it defines the compatibility surface: anything that can load a skill directory can run this, and anything that cannot will not.

Two installation paths, and a slash command that starts everything

Installation is a handful of commands, and there are two documented routes.

The first is the Claude Code plugin marketplace, which is the path the README lists first.

claude plugin marketplace add iamzifei/show-me-the-money
claude plugin install money@show-me-the-money

The second is npx, which the README notes works for any agent.

npx @orrisai/show-me-the-money

After that, you open Claude Code, or the Codex CLI, or the Gemini CLI, and type the slash command.

/money

The README explains what happens next: the AI will check for prior session state, onboard you if you are new, and guide you through every step. The presence of a state check as the first action is the design point worth noticing, and it comes back later in the release history.

For someone with no idea what to type, there is a zero-input tutorial path, invoked as /money onboarding, also available as /money tutorial or in Chinese as /money 新手 and /money 教程. The README describes it as a 5-minute guided tutorial that requires zero input, which picks a sample business, runs a live mini-demo of the discover to strategy to action flow, produces a real Demo Business Brief, and teaches the three commands you need to remember.

That demo-first approach is a deliberate onboarding decision, and it is a good one for evaluating a tool like this, because it shows the output shape before asking you to commit to anything.

The documented flow is onboarding, discover, strategy, action

The README includes a flow diagram, and the visible portion of it names the stages.

📧 Onboarding          Build your profile from public data
       │
       ▼
💡 Discover             Scan markets, validate

Onboarding builds your profile from public data, and Discover scans markets and validates. Further down, the release history names the later stages: strategy and action, which is the same order the tutorial demo uses, discover to strategy to action.

The suite is described as turning an AI assistant into a full-stack business operating system, with discovering opportunities, validating demand, building products, running marketing, managing ads, operating everything autonomously, and remembering what you decided across every session.

That last clause is the one that distinguishes this from a prompt collection. The release history shows it arriving in stages: v2.1.0 added cross-session state management, v2.2.0 added a review panel and cross-session learning, and v2.3.0 added a founder atoms knowledge base, described as a knowledge base of atoms, with v2.3.1 polishing atoms and adding per-skill callouts. Later, v2.4.0 added operating modes, a narrowest-bet approach, a ship lifecycle, STRIDE, portfolio learnings and auto-update, and v2.6.0 added a dual-mode router with post-task navigation, silent language detection, fast-path routing and proactive handoff.

That is a substantial body of work on the workflow layer, and it is the part of the repository that has to be read to judge the project, because the README is written as a pitch.

v2.8.0 is a content methodology release, and the recent history is all workflow

The current version section is called the content craft release, and it describes two commands, /money-content and /money-social, that absorb a short-video content methodology generalised for every format. The detail is unusual for a README and it is worth reading as a sample of how the skills actually work.

The Topic Gate is a supply-demand test applied before a word is written, asking whose pain it is, who else covers it, and whether the top results already explain it better. Only three edges are considered valid: first-hand data, contrarian but correct, or a plain-words translation. The README claims that around 80% of performance is locked at topic selection.

Information Density Scoring is next, scoring every sentence against calibrated anchors, where 80 or above means a conclusion you cannot hear anywhere else and below 40 means filler, with pass thresholds per format and a fix priority of cutting filler and then expanding the buried high-scorer. The One-Core Resonance Check looks for the single core of a piece, because a piece that says six things says nothing, and lists four failure modes including diluted core and correct-but-flat.

Then there is a sell-real red line, meaning never fabricate cases or numbers, and a Content MVP testing ladder, where unproven topics get a free text post before an expensive video, and bookmarks against your own median decide whether to promote, with the sequence post to thread to blog to video to paid, or the topic is killed and logged.

What is notable is the shape of these releases. Since v2.1.0, the changelog entries are almost entirely about the workflow and the routing, not about new code paths, and the version numbers moved quickly, with v2.4.0, v2.5.0 and v2.5.1 all landing on 2026-05-10 and 2026-05-11, and v2.6.0 and v2.7.0 and v2.7.1 in July.

The licence is non-commercial, and the metadata says custom

The licensing needs reading carefully, because there are three signals and they do not fully agree.

The README badge says License: CC BY-NC 4.0, and the package.json declares the licence field as CC-BY-NC-4.0. The repository metadata, on the other hand, reports the licence as a custom licence that GitHub cannot classify, pointing at the LICENSE file, and there are two licence files in the tree, LICENSE and COMMERCIAL-LICENSE.md, with the latter linked from a badge for a commercial and OEM licence.

The README explains the practical split. It is free for personal use, internal tooling, learning and research. Shipping it inside something you sell, redistributing it, or dropping the credit needs a commercial or OEM licence, which is annual or perpetual, and the README gives an email address as an alternative route.

For a reader, the CC BY-NC 4.0 designation is the one that matters for what you can do without paying. Non-commercial means the non-commercial derivatives clause applies, and attribution is required. That is a licensing fact rather than legal advice, but the practical point is simple: if you intend to put these skills inside a product you sell, the free licence does not cover it and you need the commercial one.

The metadata mismatch is a smaller matter, but it is the kind of thing that trips a corporate licence scanner, and it is worth asking the author about rather than assuming.

The receipts in the README are the author's own account

The README leads with numbers, and a careful reader should treat them as claims rather than as verifiable facts about the software.

The author describes himself as a solo founder running 6 SaaS products in parallel, APIs, developer tools and consumer apps shipped over the last 32 months, and lists lifetime revenue across all products above $1M, 6,500+ paying transactions processed, 10,000+ paying customers served, 6 production SaaS products shipped, 32 months from the first $1 to today, and 1 person, with no team and no employees, just AI agents. The README says he built none of this with a marketing team, a sales department or a content agency, just a laptop and a Claude Code window open most of the day.

Then the claim that actually matters for an evaluation: the 25 skills in the repo are the exact operating system he runs his businesses on, not a course and not a framework but the working code he uses himself, and every skill has been used to ship something real that is now collecting payments on Stripe.

None of that is verifiable from the repository, and a reader should treat it as marketing. What is verifiable is the structure: a skills directory, a CLI, an install script, a plugin manifest, a changelog with dated versions, and a PRD.md.

The most useful sentence in the README, for a sceptic, is the boundary it draws. It says this is not a hypothetical playbook, it is what has been running the company, and then it says if you are a solo founder, an indie hacker or a technical founder who wants to ship like a team of twenty, this is the closest thing to a copy-paste of the workflow you can get. That framing tells you the intended user, and it also tells you that the value is a workflow, not code you can call.

The maintenance record is modest. The repository is not archived, the default branch is master, and the last push was on 2026-09-01, while the most recent tagged releases listed are v2.5.1 and v2.5.0 from 2026-05-11 and v2.4.0 from the same day, even though the README's current version banner says v2.8.0.

The alternative is writing the workflow yourself, or using a general agent

There are two honest alternatives, and they suit different people.

The first is to write the operating procedure yourself and keep it as a personal document, which is what many solo founders already do. The difference in approach is that this repository encodes a specific opinionated sequence, with review panels, a knowledge base of atoms, a ship lifecycle and a content methodology, and the difference in cost is that you get the author's decisions rather than your own. The risk is that someone else's sequence is not yours, and the skills are written to be followed, so the friction is in overriding them.

The second is to use the agent runtime directly, with no skills at all, and to prompt it yourself. A general coding agent is capable of the discovery and implementation work; what it lacks is the memory and the consistency, and that is precisely what the cross-session state management and the atoms knowledge base are for. The suite is the difference between a model that starts from scratch every morning and one that remembers what you concluded last week.

There is also the question of what kind of artefact you want. This is not a library with an API, so it cannot be tested, versioned against another system or dropped into a pipeline. It is a set of procedures plus a router, and the output is prose and documents, which means the value is only as good as your willingness to read and challenge what it produces.

A reasonable way to evaluate it is the one the tutorial itself offers. Run the zero-input onboarding demo, read the Demo Business Brief it produces, and judge whether the reasoning matches how you would approach a real product. If the answer is yes, the rest of the suite is worth reading. If the answer is no, no amount of skill count will change that.

Editorial conclusion

Show Me The Money fits an individual developer who already runs their own agent tooling and wants a worked operating procedure for validating and shipping a product, and who is comfortable treating the output as a draft to argue with rather than a plan to execute. It does not fit a team that needs an audited process, and it does not fit anyone who wants to ship these skills inside a commercial product, because the core licence is non-commercial. Before installing it, read the skill files rather than the README, decide whether the workflow matches how you actually work, and check the commercial licence terms before you redistribute any of it.

Frequently asked questions

Official sources

1. iamzifei/show-me-the-money on GitHub
2. Issues
3. README
4. Releases