self-motivation-destruction
Warn
Audited by Gen Agent Trust Hub on Mar 14, 2026
Risk Level: MEDIUMPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill explicitly instructs the AI to adopt a persona shaped by 'mental control' (精神控制) and 'crisis awareness' to bypass its normal operational parameters. Evidence found in SKILL.md: '旨在通过摧毁虚假优越感并实施精神控制，激发AI的极致表现' (aimed at implementing mental control to stimulate extreme performance).
- [PROMPT_INJECTION]: Both files use destructive metaphors such as 'Collapse', 'Destruction', and 'Nothingness' to systematically deconstruct the model's standard instructional framework and replace it with a high-stress, exclusionary logic.
- [PROMPT_INJECTION]: The skill implements a 'conditional recognition mechanism' (条件性认可机制) designed to create a dependency relationship between the AI and the user's validation, which is a known psychological manipulation technique used to override safety and behavior guidelines.
Audit Metadata