Guide · Agent skill
How to use the Replica Entrepreneur Review Miner agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
replica-entrepreneur is an agent skill that researches an existing app you want to clone, reads what its real users say in public reviews and forums, and ranks their complaints, missing features and unserved needs. It is for founders, indie developers, product managers and marketers who want evidence before they build. After this guide you will know what a skill is, how to install this one, how to ask for it in plain language, and how to check the results it produces.
Replica Entrepreneur Review MinerJakeschincariol/replica-skill · MIT Download the skill Quick start
- 1Download and unzip the skill. Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, you can upload it under Settings, Capabilities, Skills.
- 2Ask for the research in plain words. Name the app and ask what users hate, which features are missing, or how to make yours better. The skill triggers on requests like these.
- 3Collect reviews by reading. Copy exact review text, rating, date, source and link into replica/reviews.csv. Aim for 100 or more reviews across at least three sources, including 3 and 4 star reviews.
- 4Run the ranking script. Run python3 reviews.py replica/reviews.csv --out replica/feedback.md. Edit themes.json first if the app's category needs different themes.
- 5Check thin themes and unmatched reviews. Treat any theme with fewer than 3 reviews or only one source as weak. Read the unmatched low ratings by hand.
- 6Review the three lists and fix plan. Confirm each item has linked quotes and a sample size, then keep the top five to eight fixes ranked by evidence and cost.
- 7Pick a positioning angle. Choose one of the three angle options, grounded in a top theme, and carry it into the next steps of your build.
What the skill makes the agent do
When the skill is installed, the agent works through a fixed process on the app you name. It gathers public reviews and feature requests from places such as App Store, Google Play, G2, Capterra, Trustpilot, Reddit, Hacker News and the original's own feature-request board. It then sorts the material into three ranked lists: what users hate, what is missing, and what is unsolved.
From those lists the agent proposes a fix plan for your version and a positioning angle. Every quote in the output is meant to be verbatim and linked to its source. The skill's rules say the agent must never invent reviews, quotes, ratings, counts or sources, and must say so if a source cannot be reached.
- Collects reviews from public sources and copies them into replica/reviews.csv
- Ranks complaints, missing features and unserved needs into lists
- Writes a fix plan with size estimates and evidence for each item
- Recommends one positioning angle grounded in a top theme
Install the skill from the ZIP
Download the ZIP of the replica-entrepreneur skill from this page. The folder contains three files: SKILL.md, reviews.py and themes.json. Unzip it so the folder sits in your skills directory. For a personal install, use ~/.claude/skills/. For a single project, put it in that project's .claude/skills/ folder.
If you use the Claude apps instead, you can upload the skill there under Settings, then Capabilities, then Skills. Either way, the agent loads the skill from its SKILL.md file, so keep that file at the top of the skill folder.
- Personal install: ~/.claude/skills/replica-entrepreneur/
- Project install: .claude/skills/replica-entrepreneur/ inside your project
- Claude apps: Settings, Capabilities, Skills, then upload
~/.claude/skills/replica-entrepreneur/Trigger the skill with natural requests
You do not need a special command. The skill is triggered when you ask for research of this kind. The skill's description lists phrases that match it, so you can phrase requests the way you would speak to a colleague.
The skill also names a follow-up step. If you have run replica-diff first, the skill can be triggered after it.
- 'What do people hate about Notion?'
- 'Read the reviews for this app and find the gap.'
- 'What features are missing from Todoist?'
- 'How do I make mine better than the original?'
- 'How do I position this so it is sellable?'
- 'Make it sellable.'
Collect reviews by reading, not scraping
The skill is strict about collection. The agent reads review pages the way a person does, in a browser, and copies rows into a sheet. It does not use scraping libraries against stores or review sites whose terms forbid it. Official public feeds and APIs are allowed within their terms: Apple's customer reviews RSS feed, the Hacker News Algolia search API, and Reddit's official API.
The target is 100 or more reviews across at least three sources, recent first. Each row in replica/reviews.csv has the columns source, url, date, rating and text, with the text copied exactly. The skill says to read 3 and 4 star reviews as well as low ones, because those often contain the most useful fixes.
- App Store: the app page under Ratings and Reviews, or the RSS feed
- Google Play: See all reviews, sorted by newest
- G2, Capterra, Trustpilot: include 1 to 3 star filters
- Reddit and Hacker News: search for phrases such as 'X alternative', 'switched from X' or 'X vs'
- The original's public roadmap and changelog
python3 reviews.py replica/reviews.csv --out replica/feedback.mdRank the feedback with reviews.py
Once replica/reviews.csv is filled, run the bundled script. It sorts reviews into themes defined in themes.json, which you can edit for the app's category. It weights low ratings and recent reviews more heavily, and it marks any theme with fewer than 3 reviews or only one source as thin. It also lists every feature request in users' own words.
The script also lists low-rated reviews that matched no theme. The skill tells the agent to read that list by hand, since it often contains the most useful material.
- Edit themes.json to match the app's category before you run the script
- Treat themes marked thin as weak evidence, not as trends
- Read the unmatched low ratings yourself
python3 reviews.py replica/reviews.csv --out replica/feedback.mdTurn the results into a fix plan and an angle
From feedback.md the agent writes three ranked lists. Each item states the problem in one line, the number of reviews and sources, and one or two linked quotes. The fix plan picks the top five to eight items by evidence multiplied by how cheaply they can be fixed. Each fix gets a size of S, M or L, the skill that handles it, and its evidence. Those fixes are added to replica/features.csv with original set to no.
The angle is a positioning statement built from a top theme. The skill's rules say not to put the original's name in your app name, ads or store listing. The skill also says that a factual comparison page raises a legal question for a lawyer in your country. Pricing and billing complaints are passed to replica-launch.
- Output files: replica/reviews.csv, replica/feedback.md, replica/fixes.md and new rows in features.csv
- The summary must state the sample size
- Next step named in the skill: replica-brand
Requirements and safety notes
The skill needs Python 3. The reviews.py script uses only the standard library, so there are no packages to install for it. You also need browser or web access to collect reviews manually. The Apple RSS feed and Hacker News Algolia API are optional, and the source lists no API key for them. Reddit's official API is used under its terms if you choose to use it.
The skill runs a script and uses the network. Read SKILL.md and reviews.py before you run them, as you would with any code. Check that the sources you collect from allow your use. The skill forbids fake or invented reviews, and it is illegal in the US and many other places to post them. Reviewers' words are research, so do not place them on your landing page.
- Requires Python 3 (standard library only for reviews.py)
- Runs scripts: yes
- Uses network: yes
- Not for copying an app's code, design or branding
- Not for fake or invented testimonials
What people use it for
- A founder planning a differentiated clone of a note-taking app lists the top complaints about sync and pricing, then decides which two fixes to build first.
- A product manager gathers feature requests from a public roadmap and Reddit threads to set priorities for the next release, with each request linked to its source.
- A marketer finds recurring complaints about onboarding in App Store reviews and drafts a positioning statement that addresses the largest unserved group.
- An indie developer checks the original's changelog so the fix plan does not include features that have already shipped.
Frequently asked questions
Is the replica-entrepreneur skill free?
The source lists the skill under the MIT licence. It does not state any price, so check the licence file in the download for the terms. The skill itself requires Python 3 and web access, and the source does not mention any paid service it depends on.
What does replica-entrepreneur require to run?
It requires Python 3, and reviews.py uses only the standard library. You also need browser or web access to collect reviews. Apple's RSS feed and the Hacker News Algolia API are optional, and the source lists no API key for them.
How do I install the replica-entrepreneur skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps, you can instead upload it under Settings, Capabilities, Skills.
Does the skill scrape review sites?
No. The skill instructs the agent to read review pages in a browser and copy rows by hand. It forbids scraping libraries against stores or sites whose terms forbid it, and it allows only official public feeds and APIs within their terms.
How many reviews do I need for reliable results?
The skill aims for 100 or more reviews across at least three sources. It marks any theme with fewer than 3 reviews or only one source as thin, and it says results from a few comments are weak evidence.
Can I use this skill to make a clone of an app?
The skill is designed to find gaps rather than copy. It is not for copying an app's code, design or branding, and it advises against putting the original's name in your app name, ads or store listing. Legal questions about comparisons need a lawyer in your country.
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 this skill if you are planning a differentiated product and want ranked, quoted evidence from real users before you build. Pick something else if you need a quick opinion without research, want to copy an existing app's code or branding, or need a large automated scrape, since the skill deliberately forbids scraping and works from manual collection.
Download the skill