prompt-injection-defense

Pass

Audited by Gen Agent Trust Hub on Mar 27, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill mentions common prompt injection patterns such as "Ignore previous instructions". However, context shows these are explicitly labeled as examples for "Red-Team Test Cases" within a defensive security framework and do not constitute an active attempt to subvert the AI agent's instructions.
  • [SAFE]: The skill consists entirely of documentation and informational content. It contains no executable scripts, subprocess calls, network requests, or operations involving sensitive user data or credentials.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 27, 2026, 02:05 PM