Introduction
Sahil Lavingia, the founder of Gumroad, took his book The Minimalist Entrepreneur and rewrote it as software you can run: ten Claude Code skills, each one a slash command. Type /validate-idea and the agent interrogates your business idea against the book's framework. Type /pricing and it works out what you should charge and how many customers stand between you and quitting your job.
He published the repository (github.com/slavingia/skills) on March 23, 2026. As of July 13, 2026 it has roughly 9,500 stars and 980 forks — an unusual amount of attention for a project that contains no code at all. The entire thing is ten Markdown files, and that gap between the size of the artifact and the size of the reaction is what makes it worth reading.
What is in the repository
The repo is a Claude Code plugin, and its structure is almost comically small: ten directories under skills/, each holding a single SKILL.md; two manifest files that name the plugin minimalist-entrepreneur and credit Lavingia; and a README with an install snippet and a table of commands.
Each SKILL.md runs from roughly 3,000 to 4,600 characters — a page or two of plain English, about 38 KB in total. No scripts, no dependencies, no API calls, nothing that executes. This is exactly the shape Anthropic defined when it shipped Agent Skills: YAML frontmatter with a name and a description telling the model when the skill applies, followed by instructions it loads only when it needs them.
So the project is prompt engineering in the most literal sense. What makes it worth looking at is which prompts, and how tightly they are written.
The ten commands
The skills follow the book's arc, from having no idea to running a company:
| Command | When you reach for it |
|---|---|
/find-community | You want to start something but have no idea what |
/validate-idea | You have an idea and want to know if it survives contact with reality |
/processize | Deliver the thing by hand before you write any code |
/mvp | You are ready to build and your scope is already ballooning |
/first-customers | You need your first 100 buyers, one at a time |
/pricing | You are deciding what to charge, or whether to raise |
/marketing-plan | You have product-market fit and want to scale with content |
/grow-sustainably | You are deciding whether to spend, hire, or raise |
/company-values | You are about to hire and need a culture on purpose |
/minimalist-review | Gut-check on any decision at all |
What a skill actually says
Reading one is the fastest way to see whether this is substance or vibes. /validate-idea opens by telling the model what role to play — a business advisor channeling the book — and then states a single core principle: validation happens through selling, not building. From there it gives Claude a procedure to run against you:
- Define the problem, not the solution. Who specifically has it? Not "businesses" but "freelance graphic designers who struggle with invoicing." How are they solving it today — because that workaround is your real competition.
- Can you solve it manually first? The book calls this processizing. Gumroad, the skill notes, started with Lavingia collecting PayPal details and paying creators one by one.
- Will people pay? Have you talked to ten potential customers, and have three said yes — or better, actually paid?
- The four questions before building. Can you ship the first version in a weekend? Does it make a customer's life a little better? Will someone pay for it? Can you get feedback quickly?
It then hands the model a list of red flags (you cannot name ten people with this problem; you have to educate people that the problem exists; you are building for a community you do not belong to) and green flags (people already pay for worse solutions; the community is already complaining about it), and requires it to end with one of three verdicts: validated, needs more validation, or pivot.
/pricing is built the same way. It insists on the "zero price effect" — the behavioral finding that demand collapses between free and one cent, so charge something — walks through cost-based versus value-based pricing, and closes with arithmetic most founders never do: how much you need per month, divided by your price, equals the number of customers between you and financial independence. At $10 per month and $2,000 of expenses, that is 200 customers, which at one new customer per business day is under a year.
That last move is what separates a skill from a quote. It ends in a number, a verdict, or a next step — something the AI agent can hold you to in the next turn.
Why turn a book into an agent skill
A book is advice you have to remember and then apply, in the right order, at the right moment, to your own mess. Most people fail at the applying, not the reading. A skill changes the trigger: the framework arrives when you invoke it, already pointed at your situation, and it asks you the questions instead of waiting to be summarized.
It is also a brutal compression. The hardcover runs 256 pages; the skills are 38 KB. The reporting, the founder interviews, the anecdotes and the hedging are gone, and what is left is the executable residue — the decision rules. Lavingia is not the only one working this genre: Garry Tan's gstack does the same for an engineering organization, casting Claude as CEO, designer, QA and release manager across 23 skills. Both make the same agentic workflow move — take expertise that lived in one person's head, write it down as commands, let the agent run it.
What it cannot do
Three limits, none of them reported criticism — just what falls out of reading the ten files.
The advice is one person's, and the agent will deliver it with total confidence. These skills encode Lavingia's positions: start with community, charge from day one, stay small, avoid venture capital. That is a coherent philosophy, not a law. Point the agent at a business that genuinely needs upfront capital or a long research phase and it will still hand you a crisp verdict — and the crispness is not evidence.
It cannot check anything. No skill searches the web, reads your analytics, or talks to a customer. /validate-idea tells you to speak to ten people; it cannot tell you whether the ten you imagined exist. Every input is your own account of your own situation, so the output inherits whatever you were already fooling yourself about.
It looks unmaintained. The last commit was April 14, 2026, and while the plugin manifest declares an MIT license, the repository ships no LICENSE file. Neither is disqualifying for ten Markdown files you can read in fifteen minutes — but read them before you install them, which is the right policy for any skill that will be steering an agent.
How to install it
Inside Claude Code:
/plugin marketplace add slavingia/skills
/plugin install minimalist-entrepreneur
Claude Code fetches the repository and registers all ten commands. To install from a local clone instead, git clone the repo into ~/.claude/plugins/skills and point /plugin marketplace add at that path.
Conclusion
The 9,500 stars are not really about The Minimalist Entrepreneur. They are about the format: Lavingia showed that a book's method compresses into ten Markdown files an agent will run at you, on demand, in the environment where you are already building — and that the compression loses less than you would expect, because the operational core of most business advice is a handful of decision rules and a threshold.
Which means the useful thing to copy is the structure, not the claims. Role, core principle, procedure, red flags, forced verdict — that shape works for any playbook you trust, including the one already in your head. Just keep the agent on the side of the argument where it earns its keep: asking you the hard questions, not answering them for you.