The Agent Skills Directory

[SAFE]: The skill downloads content from HuggingFace (huggingface.co), which is a well-known and trusted repository for machine learning models and datasets.
[SAFE]: Credential management instructions follow security best practices, advising the use of environment variables and explicitly prohibiting the logging or plain-text storage of API tokens.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external data, which inherently creates a surface for indirect prompt injection.
Ingestion points: Files and folder snapshots downloaded from HuggingFace repositories (e.g., unsloth/gemma-4-12b-it-GGUF).
Boundary markers: The provided documentation does not include specific delimiters or instructions for the agent to ignore potentially malicious commands embedded in downloaded files.
Capability inventory: The skill environment allows access to Bash and Read tools.
Sanitization: There are no descriptions of sanitization or validation processes for the downloaded content prior to its use by the agent.

hf-download