GerardoRdz96/deepfork ? reverse-engineered prompt

Build me an agent skill called DeepFork that lets me point at any public GitHub repo and get a clean explanation of how it works, plus a clean room blueprint for rebuilding my own version with changes I ask for. I want it to first check the repo license and refuse unlicensed targets, then analyze the codebase, ideally with a knowledge graph if that tool is available, and fall back to a simpler repo mapping if not.

The main output should feel useful to a normal developer, not academic. It should generate an UNDERSTANDING.md that calls out the important parts, data flow, and what is verified versus inferred, then a BLUEPRINT.md that turns that into a rebuild spec with my requested deltas baked in. After that, have it create a fresh rebuild folder from the blueprint only, with tests first and an ATTRIBUTION.md.

Please make it work as a Claude Code style skill, with a simple slash command flow like deepfork repo url plus my requested changes. If you need to, look up the current skill format and graph tool docs online.