Count the apps on your phone you pay for every month. Now count the ones where you use three features and ignore the other forty. For most of us, the second number is uncomfortably close to the first.
This week I came across Replica, a free pack of eleven Claude skills by Jake Schincariol that tries to address that. You point it at an app, and it helps you build your own version. Not a copy. A version. That difference turns out to be the most interesting part, but I’ll get to it.
A dev team in a folder
Replica is eleven skills that run inside Claude Code. A skill is a set of instructions Claude follows for a specific task, a bit like handing a freelancer a briefing document instead of explaining everything from scratch. It is MIT-licensed, so free to use and change. No sign-up, no API key, nothing to connect.
Each skill covers one phase of building an app, and you run them in order:
- Recon reverse-engineers the original: its screens, flows, components and data model, from public pages, screenshots, store listings and your own account.
- Architect plans the tech stack, the database and the API.
- Design rebuilds the design system (colours, type, spacing) with your own assets.
- Build rebuilds the app screen by screen.
- Backend handles login, the database, payments and integrations.
- Test clicks through every flow and logs the bugs.
- Diff compares your version with the original and gives it a parity score.
- Entrepreneur reads what users hate about the original. More on this one below.
- Brand gives your version its own name and identity.
- Launch writes the landing page, pricing and App Store listing.
- Deploy puts it live on your own domain.
The clever bit is boring, which is usually a good sign. Every phase writes its output into a replica/ folder in your project, and the next phase reads it. Recon produces a feature list, Architect builds on that list, and Diff checks your clone against the same list. It is how a decent agency works: one shared brief, and everyone reads it before they start.
The Entrepreneur is the one that matters
Most of the pack is engineering. Useful, but a clone of an app is just the same app with fewer users. Nobody switches to a copy.
The Entrepreneur skill is where it turns into a business idea. It reads public reviews of the original app and sorts out what people complain about: missing features, price hikes, updates that broke the one thing they used it for. Then it turns that into a list of fixes for your version and a positioning angle. Every quote must be verbatim, include a link, and state how many reviews you read. No invented testimonials.
After 25 years in marketing, this is the bit that made me sit up. Reading one-star reviews of a competitor is the cheapest market research there is, and almost nobody does it properly because it is tedious. Now it isn’t. You also don’t have to build anything to use it. Run this one skill on its own, and you’ll have a decent idea whether there is a gap worth filling, or whether the original is annoyingly good.
What it won’t let you copy
“Clone any app” sounds like a lawsuit waiting to happen. Actually, the pack is stricter than most people would be on their own.
It works clean-room style: it rebuilds what an app does and how people move through it, never what the app owns. So no source code, no logos, icons or illustrations, no licensed fonts, no copy, no content. It only uses public sources and your own account, and it won’t log into anyone else’s or get past a paywall. If your account terms forbid building a competitor, it tells you and sticks to public pages.
The Brand phase checks for leftovers from the original, and Deploy refuses to go live until that check comes back clean. It also never signs up for services, buys a domain or types in your card details. It writes the steps; you do them. Which is roughly how I’d want a junior developer to behave too.
One thing the skills can’t check for you: features can still be protected by patents, and some markets are crowded for a reason. Clean room keeps you away from the obvious mistakes, not all of them.
Where to start
You need Claude Code and the pack from GitHub. Each phase is its own command, from /replica-recon to /replica-deploy, and any of them can run on its own.
My suggestion: don’t start with Recon. Start with /replica-entrepreneur on an app you already pay for and slightly resent. If the reviews show a real gap, run the rest. If they don’t, you’ve saved yourself a few weekends, and you can keep paying the subscription with a clearer conscience.
Paid apps are fine. Mediocre ones should worry.
Replica won’t end app subscriptions. Building the thing was never the hardest part of running an app; support, updates and finding users are, and no skill pack does those for you. What it does change is the cost of trying. A week of evenings instead of a development budget.
That shifts the pressure onto apps that charge a premium and stop listening to their users. Their one-star reviews used to be a complaint box. Now they are a product roadmap, and anyone with Claude Code and a free weekend can use it.
