关于
This skill transforms Claude into a research engineering mentor that facilitates the manual reproduction of scientific papers from first principles. Instead of providing simple code generation, it follows a rigorous four-phase process—pre-implementation analysis, scaffolded implementation, debugging, and verification—to build models from the ground up using fundamental libraries like NumPy and PyTorch. By emphasizing understanding through interactive checkpoint questions and a 'debugging gauntlet,' it helps users move beyond high-level wrappers to master the underlying mechanics of cutting-edge algorithms in fields like Computer Vision, NLP, and Reinforcement Learning.
