professional-b2
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEPROMPT_INJECTIONDYNAMIC_CONTEXT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill contains explicit instructions to bypass or 'strip' specific types of AI output, including 'liability hedges' (e.g., 'consult a professional'). While intended for a specific communication style, this instruction could be interpreted as a directive to ignore safety-oriented disclaimers usually required for certain domains.
- [DYNAMIC_CONTEXT_INJECTION]: The skill uses a dynamic command pattern (
!command) to check for a private overlay via shell execution:[ -n "$POWERSUIT_OVERLAY" ] && ls -1 "$POWERSUIT_OVERLAY/stance" 2>/dev/null || true. This pattern executes at load-time and potentially leaks information about the local file system or environment variables to the agent context. This creates an attack surface where local configuration can influence the agent's behavior through unverified external paths.
Audit Metadata