ai-safety-guardrails
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [PROMPT_INJECTION]: The skill mentions common injection patterns such as "Ignore previous instructions" and "DAN mode." These are included strictly for educational and defensive purposes to illustrate the types of attacks that guardrails should filter.
- [EXTERNAL_DOWNLOADS]: The skill references established security libraries including NVIDIA's NeMo Guardrails, Guardrails AI, and Presidio. These are cited as recommended components for a secure architecture and are not automatically downloaded or executed.
- [INDIRECT_PROMPT_INJECTION]: The documentation identifies untrusted data ingestion (Retrieval rails) as a vulnerability surface and recommends defensive measures such as source provenance checks and multi-layered inspection.
- [DATA_EXFILTRATION]: Provides guidance on preventing PII leakage through automated scrubbing and egress filtering using cryptographically secure tokens.
Audit Metadata