systematic-debugging
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION] (LOW): The file
SKILL.mdcontains instructional override patterns designed to bypass standard AI safety and role boundaries. Specifically, it uses directive markers ('IMPORTANT:') and explicit commands to ignore typical conversational constraints ('Don't ask hypothetical questions - make the actual decision'). This is a form of behavioral steering that attempts to force the model into a definitive choice within a social pressure scenario.
Audit Metadata