Guide · Agent skill
How to use the Replica Recon agent skill (replica-recon)
Updated · written with AI from the project's own documentation and checked against our catalogue data
replica-recon is an agent skill, a folder of instructions that an AI agent such as Claude loads when a request matches its description. It reverse-engineers an app into a structured recon map: screen inventory, user flows, reusable components, an inferred data model and a feature matrix. This guide explains what the skill does, how to install it, how to trigger it and what to keep in mind, so you can produce a clear plan before you build a similar product.
Replica ReconJakeschincariol/replica-skill · MIT Download the skill Quick start
- 1Download and unzip. Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into your project's .claude/skills/ folder.
- 2Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 3Make a natural request. Ask something like "map out X" or paste an app's URL and say you want to rebuild it.
- 4Scope the slice. Answer the three scoping questions: platform, the specific slice, and who the clone is for.
- 5Review the sources and screens. Check the sources table and screen inventory, including empty, error and permission states.
- 6Check the output files. Open replica/recon.md and replica/features.csv in your project and review flows, data model and skip rows.
What the replica-recon skill makes the agent do
When you ask the agent to clone or reverse-engineer an app, the skill tells it to work from public pages, help docs, app store listings, screenshots and your own account. The output is a map of the app rather than working code.
The skill writes files into your project under a replica/ folder: recon.md (the recon map) and features.csv (the feature matrix). It is the first step of a larger pack; the source describes later steps that build from its output.
- Screen inventory with stable IDs such as S01, S02, including states seen (empty, loading, filled, error, permission denied, mobile)
- User flows with the screens they pass through and edge cases
- Component list of repeated UI parts
- Inferred data model with evidence and a confidence level for each entity
- Feature matrix with priority (must, should, could) and a skip list of what cannot be cloned
Install the replica-recon skill
Download the ZIP from this page and unzip it. You can install the skill in one of two places: your personal skills folder, or a single project's skills folder.
For a personal install, unzip into ~/.claude/skills/. For a project-only install, unzip into the project's .claude/skills/ folder. If you use the Claude apps instead, upload the skill under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/
- Project only: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload
Trigger the skill with natural requests
You do not need to name the skill. The skill's description lists the kinds of requests it responds to, so the agent will usually load it when you say things like the phrases below, or paste an app's URL or App Store link and ask to rebuild it.
Give the agent the app and the slice you care about. A narrow slice gives a more useful map than an entire product.
- "Clone this app"
- "Reverse engineer X"
- "How does X work?"
- "Map out X"
- "What screens does X have?"
- "I want to build my own version of X"
- "Copy this app"
Example: scoping a request
The skill begins by scoping the work. It asks three things, or proposes answers for you to confirm: which app and platform (web, iOS, Android or desktop), which slice, and who the clone is for.
The source gives this contrast: "All of Notion" is not a project, while "Notion's pages, blocks and sharing" is. By default the skill focuses on the core loop, the one flow users pay for.
What the skill needs and what it will not do
The listing states that the skill requires nothing beyond the agent itself, does not run scripts, and uses the network. The skill works from functionality and UX patterns, not from another company's code or content.
It will not log into an account that is not yours, ask for a password, or get past a paywall or login. If a browser tool is connected to your own browser, it reads pages at human speed with you present, and it does not run crawlers or bulk downloads.
It also will not read or save an app's JavaScript bundles, decompile its binary or log its network calls to copy endpoints. Public API docs are fine to read. Screenshots are kept as reference in replica/screens/ and are never shipped in the clone.
- Public sources and your own account only
- Reading, not scraping: no crawlers, bulk downloads or loops
- No source code and no private APIs
- Check the terms: if your account's terms forbid building a competing product, the skill works from public sources only
- Licensed content, the company's data and partner deals are listed as skip rows
Tips for a useful recon map
A bad recon map leads to a bad clone, so the skill asks you to spend time here. Record every empty and error state you see, because a state you did not record is one you will not build.
Count the clicks on the happy path of each flow; the source calls this the number to beat. Mark guesses in the data model as guesses, and treat the size estimate (S, M, L or XL) as a planning aid rather than a promise.
- Start with the help center and pricing page, which show the feature list and what is gated
- Use the user's own account to see real states
- Expect the final output to end with a five-line summary and a suggested next step, /replica-architect
What people use it for
- Product builders listing the screens and flows of a feature set before deciding what to clone
- Designers and students studying how a known app organises its screens and flows
- Founders scoping a niche version of an existing app and sizing the hard parts
- Teams deciding which features are must-haves versus could-haves before any build starts
Frequently asked questions
How do I install the replica-recon skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for a personal install, or into a project's .claude/skills/ folder for a single project. In the Claude apps, upload it under Settings → Capabilities → Skills.
Do I need anything else to run replica-recon?
The listing says it requires nothing beyond the listed items, which is none, does not run scripts, and uses the network. It works from public sources and your own account.
Is replica-recon free?
The skill is listed under the MIT licence. The source does not mention pricing, so check the licence file included with the download for the terms.
Will replica-recon copy an app's code?
No. The skill is written to rebuild functionality and UX patterns only. It refuses to read source code, JavaScript bundles or private APIs, and it lists licensed content and company data as things that cannot be cloned.
Can it clone a paywalled or login-only app?
No. The skill will not log into accounts that are not yours, ask for passwords, or get past a paywall or login. It can only work from your own account and public sources.
Will the output be a working clone?
Not from one run. The output is a plan: a recon map, a feature matrix and a size estimate. The skill states it makes no promise of a perfect clone.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use replica-recon if you want a structured, evidence-linked plan of an app's screens, flows and features before you build a similar product, and you are working from public sources or your own account. Pick something else if you need copied code or proprietary content, access to someone else's account, or a finished working clone from a single run.
Download the skill