The Agent Skills Directory

[SAFE]: The skill utilizes industry-standard machine learning libraries including Hugging Face transformers, peft, trl, and the axolotl training framework. All dependencies are sourced from the official Python Package Index (PyPI) registry.
[SAFE]: The Kubernetes Job manifest demonstrates secure handling of sensitive information, using SecretKeyRef to inject credentials like HUGGING_FACE_HUB_TOKEN and WANDB_API_KEY rather than hardcoding them in instructions or configurations.
[SAFE]: The container images referenced in the Kubernetes templates are sourced from a well-known, official service (nvcr.io
NVIDIA's Container Registry), which is a trusted source for GPU-accelerated workloads.
[SAFE]: The Python and shell commands provided are transparent and consistent with the skill's stated purpose of setting up LLM training environments, with no evidence of obfuscation or malicious redirection.

llm-fine-tuning