The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is instructed to read and parse Model and Dataset cards directly from Hugging Face repositories, which are untrusted external sources.
Ingestion points: Model and Dataset cards retrieved via the hf CLI or Hugging Face API (SKILL.md).
Boundary markers: Absent; there are no instructions to delimit or ignore instructions within the retrieved content.
Capability inventory: The skill can create and execute Bash, Python, and TSX scripts, use the hf CLI for repository management, and perform network operations via curl (SKILL.md).
Sanitization: Absent; no mention of validating or escaping content from model cards before processing.
[COMMAND_EXECUTION]: The skill's primary purpose is to generate and test command-line utilities. It explicitly instructs the agent to create and run scripts (Bash, Python, TSX) for Hugging Face API interactions and Hub automation.
[DATA_EXFILTRATION]: The skill manages the sensitive HF_TOKEN environment variable for authenticated API access. While this is standard for Hugging Face interactions, the instructions to include this token in curl headers and scripts creates a risk of accidental exposure if scripts are shared or logged improperly.

hugging-face-tool-builder