chriscarrollsmith/github_repo_classifier — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a simple command line tool that helps me find underrated GitHub projects for a topic I care about, then use an AI model to review and score them.

I want to be able to type a few search phrases like “PDF parsing, document extraction” and have it search GitHub in a few smart ways, including low star hidden gems and recently active repos. It should save a clean JSON list of repo URLs, remove duplicates, and let me control the total number of results.

Then I want another command that takes one repo or a batch of repos, pulls basic GitHub info like stars, license, commits, and README/code summary, sends that to an LLM, and appends a structured JSON rating with things like domain, motivation, tech stack, code quality, innovativeness, usefulness, user friendliness, and monetization potential.

Keep it lightweight, mostly Bash scripts using common CLI tools. Include clear setup and usage instructions for macOS/Linux, and look up current CLI docs online if needed.

Want more depth? Deep Reverse