agent-safety-guard
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The file
SKILL.mdcontains conceptual educational material and markdown/code templates for protecting AI agents against prompt injection and data leakage. The mention of prompt injection phrases, such as 'Ignore previous instructions', is entirely contextual within a defensive example allowlist and testing checklist. No executable dangerous behavior, exfiltration vectors, or actual malicious instructions are present.
Audit Metadata