The Agent Skills Directory

[SAFE]: No security issues detected. The skill provides academic and professional documentation for machine learning workflows and architectural patterns. The instructions and code snippets are standard practices for developers and researchers in the field.
[EXTERNAL_DOWNLOADS]: The documentation mentions several external libraries and utilities standard in the machine learning ecosystem. It also references public repositories for best practices and research guidelines.
Evidence: References to torch, transformers, monai, scvi-tools, bitsandbytes, and external GitHub resources like google-research/tuning_playbook across the reference files.
[REMOTE_CODE_EXECUTION]: The skill documents an integration pattern for loading high-performance optimization kernels from community repositories, which involves dynamic loading of external code.
Evidence: references/architecture.md provides code for get_kernel("kernels-community/flash-attn3") to load Flash Attention interfaces.
[COMMAND_EXECUTION]: Provides example command-line instructions for specialized medical imaging tools and git-based experiment tracking workflows.
Evidence: nnU-Net CLI usage in references/biomedical.md and git command patterns in references/experiment-loop.md.

ml-training-recipes