The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references the installation of the wandb library from the Python Package Index (PyPI) and provides links to official Weights & Biases documentation and GitHub repositories hosted under well-known domains (wandb.ai, github.com/wandb/wandb).
[CREDENTIALS_UNSAFE]: The documentation includes instructions for authenticating with the Weights & Biases platform, such as using wandb login and setting the WANDB_API_KEY environment variable. These examples correctly use the placeholder your_api_key_here rather than hardcoding actual credentials.
[DATA_EXFILTRATION]: While the skill involves uploading experiment metrics, model checkpoints, and datasets to the Weights & Biases platform (wandb.save, wandb.log_artifact), this behavior is the intended primary purpose of the tool for experiment tracking and MLOps, and it targets a well-known, legitimate service.

weights-and-biases