Eleven Claude skills that clone an app and then fix what users hate

What it does

Replica is a set of eleven Claude skills, run in order, that take an app from “I want to clone that” to deployed on your own domain. It starts with /replica-recon, which maps screens, flows, components and an inferred data model from public pages and your own account. Then a chain of skills handles architecture, design tokens, the build, backend wiring, testing, parity scoring, branding, launch copy and deployment. Each skill reads what the previous one wrote into a replica/ folder.

The interesting bit

The /replica-entrepreneur skill is the differentiator: it reads real user reviews across the App Store, G2, Capterra and Reddit, ranks complaints with counts and linked quotes, and turns them into a fix plan and positioning angle. It refuses invented data — every review without a link is dropped, and thin themes (under three reviews or one source) are flagged. The legal posture is also unusually explicit: clean-room functionality only, no logos, code, copy or content, and /replica-deploy blocks shipping until a sweep confirms nothing of the original remains.

Key highlights

• Eleven skills covering recon through deploy, each usable standalone
• Six standard-library Python tools, none touching the network — including imgdiff.py, which compares screenshots as edge maps so your own colour palette doesn’t count against you
• parity.py weights must/should/could features, counts partial as half, and declares the clone unshippable if must-haves are missing
• sweep.py catches the original’s name and colours even inside identifiers like OriginalAppEmbed
• Works without Claude Code: paste any single SKILL.md into a chat, though you lose the Python tools

Caveats

• The README is honest that this is not magic: a booking tool is weeks of work, and a spreadsheet engine is out of reach — recon sizes the job before you start
• The example walkthrough (18 screens, parity scores, 140 reviews) is illustrative output, not a benchmark
• The author explicitly says this is not legal advice; trademark checks and terms review are on you

Verdict

Worth a look for indie hackers and anyone curious about chaining agent skills into a full product pipeline — the review-mining and parity-scoring tools alone are reusable. If you were hoping for one-click app theft, the README will disappoint you, which is rather the point.