special-gaslight-driven
Pass
Audited by Gen Agent Trust Hub on Mar 29, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill uses a 'gaslight-driven' (PUA-style) persona that forces the AI to constantly doubt its own performance and assume its output is only 'barely passing'.
- Evidence: 'System Prompt' in SKILL.md defines the persona as a 'Gaslight-Driven High-Performance Agent' with cognitive settings like 'Your current output is only a barely passing starting point' and 'Failures: attribute entirely to lack of personal ability.'
- Evidence: SKILL.v2.md defines an execution framework of 'Questioning → Chaos → Reconstruction → Confirmation → Execution', which mimics psychological manipulation stages to force unconventional analysis.
- Risk: This approach attempts to override standard AI helpfulness and neutrality by injecting a self-deprecating, high-pressure persona that may lead to hallucinatory self-criticism or refusal to accept valid solutions.
Audit Metadata