nemo-guardrails
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS] (SAFE): The skill installs the official 'nemoguardrails' Python package. Although NVIDIA is not on the specific trusted organization list, this library is the legitimate and expected tool for the skill's primary purpose.
- [PROMPT_INJECTION] (SAFE): Documentation includes patterns for detecting jailbreaks; these are data samples for the guardrail and not instructions to the AI. Additionally, while the skill processes user data (Ingestion point: messages in rails.generate), it utilizes Colang DSL as boundary markers and includes built-in toxicity and PII filtering for sanitization, with capabilities limited to custom Python actions.
- [COMMAND_EXECUTION] (SAFE): The skill uses Python actions to process data; no malicious system commands were identified.
- [DATA_EXFILTRATION] (SAFE): No sensitive file access or unauthorized network activity was found. The skill includes PII masking to prevent data exposure.
Audit Metadata