llamaguard
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS] (SAFE): The skill references Meta's official LlamaGuard models on HuggingFace. Meta is a trusted organization under the analysis scope.
- [COMMAND_EXECUTION] (SAFE): Shell commands are limited to standard package installation (pip) and official CLI authentication (huggingface-cli) required for accessing gated models.
- [DATA_EXFILTRATION] (SAFE): No unauthorized network requests or sensitive file access patterns (e.g., SSH keys or environment secrets) were identified.
- [PROMPT_INJECTION] (SAFE): No instructions attempt to bypass safety filters; the skill's primary intent is to provide moderation and alignment for AI responses.
Audit Metadata