nemo-automodel-model-onboarding

Pass

Audited by Gen Agent Trust Hub on Aug 1, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill provides legitimate guidance and implementation patterns for NVIDIA's NeMo AutoModel framework. No malicious patterns such as credential theft or unauthorized command execution were detected.
  • [EXTERNAL_DOWNLOADS]: The skill guides the agent to fetch model metadata and weights from Hugging Face (huggingface.co), which is an industry-standard source for model artifacts and considered a well-known service.
  • [PROMPT_INJECTION]: The skill involves processing external model configurations to generate implementation code, representing an indirect injection surface. * Ingestion points: Model config.json files fetched from the Hugging Face Hub (Phase 1.1). * Boundary markers: No specific delimiters or warnings to ignore instructions within the JSON are defined. * Capability inventory: The skill enables the creation of Python implementation files (model.py, layers.py) and modification of the model registry. * Sanitization: No explicit validation or sanitization of the JSON content is described beyond classifying architecture fields.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 1, 2026, 09:48 PM
Security Audit — agent-trust-hub — nemo-automodel-model-onboarding