nemotron-add-model

Pass

Audited by Gen Agent Trust Hub on Jun 23, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its ingestion of external technical reports. If a report contains malicious instructions designed to influence the agent's summary or generation process, these could be reflected in the resulting skill files.\n
  • Ingestion points: Reads tech report content from a user-provided URL (e.g., arXiv HTML) and existing skill files in the repository (e.g., skills/nemotron-nano3/SKILL.md).\n
  • Boundary markers: The instructions do not define explicit delimiters or 'ignore' warnings for the external report content, although it does specify that chunks should be 'question-oriented summaries' rather than raw text.\n
  • Capability inventory: Performs file system operations including creating directories and writing multiple files (SKILL.md, INDEX.md, model-card.md, index.toml, marketplace.json).\n
  • Sanitization: No specific sanitization or validation of the external report content is mentioned beyond the summarization process.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 23, 2026, 05:56 PM
Security Audit — agent-trust-hub — nemotron-add-model