The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires executing local Python scripts using environment managers such as uv or pixi. This is the core functionality of the skill, used to run training and evaluation tasks.
[EXTERNAL_DOWNLOADS]: The skill references standard package registries for Python dependencies and explicitly mentions the official PyTorch CUDA index (download.pytorch.org) for GPU support. These are well-known, trusted sources.
[SAFE]: The skill maintains a clean separation of experiments through structured directory trees, implements rigorous logging protocols, and includes guardrails against long-running processes or unverified data plotting.

ml-experimentation