The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill utilizes standard shell commands to execute local Python scripts for data preparation and model training. These operations are transparent and consistent with the stated educational purpose.
[EXTERNAL_DOWNLOADS] (SAFE): Dependencies like torch, transformers, and wandb are standard, reputable machine learning libraries installed through official package managers. External links point to established community resources such as HuggingFace and GitHub.
[DATA_EXFILTRATION] (SAFE): No sensitive file access or unauthorized network exfiltration patterns were detected. The usage of Weights & Biases (wandb) for experiment tracking is a standard industry practice and is appropriately disclosed.
[PROMPT_INJECTION] (SAFE): The content does not contain instructions designed to bypass safety filters, extract system prompts, or override agent constraints.

nanogpt