This gives Claude the ability to run arbitrary Python code in a sandboxed environment with over 100 libraries already installed, from pandas and matplotlib to Selenium, OpenCV, and trimesh. It's genuinely useful when you need to scrape a website, process data, manipulate images, generate videos, or work with 3D models without setting up your own environment. The sandboxing means it's safe to let Claude write and execute code directly. Files saved to the outputs directory get returned automatically, which is a nice touch. The main limitation is it's CPU-only, so you'll need different tools for ML workloads. Works through the inference.sh belt CLI, and you can bump it to 16GB RAM if needed.
npx -y skills add inference-sh-skills/skills --skill python-executor --agent claude-codeInstalls to .claude/skills
Select a file.
sickn33/antigravity-awesome-skills
sanand0/scientific-research