The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill utilizes standard Python and torchrun commands to execute local training and data preparation scripts. There are no instances of suspicious command concatenation or remote script execution (e.g., piping curl to bash).\n- [EXTERNAL_DOWNLOADS] (SAFE): All identified dependencies (torch, numpy, transformers, etc.) are reputable machine learning libraries from standard registries. Network access is restricted to legitimate uses such as loading pretrained model weights and experiment logging via wandb.\n- [DATA_EXFILTRATION] (SAFE): Analysis of the provided code snippets confirms no unauthorized access to sensitive system paths (e.g., .ssh, .aws) or hardcoded credentials. Data operations are limited to local model checkpoints and training datasets.\n- [PROMPT_INJECTION] (SAFE): No instructions attempting to bypass safety filters, extract system prompts, or override agent behavior were found. The skill content is purely instructional and technical.

nanogpt