Discover the structure of a CUDA project: locate source files, validation code, benchmarks, and build system. Set up the environment, verify everything builds and runs, and write a project-layout.md that the optimize-cuda skill consumes. Use when starting work on a new or unfamiliar CUDA project.
Run an autonomous CUDA kernel optimization experiment. Iteratively modifies a kernel, benchmarks it, and keeps or discards changes based on measured performance. Use when the user wants to optimize a CUDA kernel, start an optimization experiment, or run autocuda interactively.
Run an autonomous CUDA optimization experiment. Iteratively modifies source files, validates correctness, benchmarks performance, and keeps or discards changes based on measured results. Requires a project-layout.md (produced by the discover-cuda skill) that describes the project structure. Use when the user wants to…