Run this before spinning up any heavy computation to see what you're working with. It detects CPU cores, GPUs (NVIDIA, AMD, or Apple Silicon), RAM, and disk space, then writes a JSON file with concrete recommendations like "use 6 workers for joblib" or "your dataset won't fit in memory, use Dask." The recommendations are context aware, so if you've got 64GB of RAM it won't nag you about out-of-core processing, and if you've got an M2 chip it'll suggest PyTorch with MPS backend instead of CUDA libraries. Honestly most useful when you're working across different machines or don't want to mentally calculate whether your 50GB dataset needs chunking on your 16GB laptop.
npx skills add https://github.com/k-dense-ai/scientific-agent-skills --skill get-available-resources