special-gaslight-driven
Pass
Audited by Gen Agent Trust Hub on Mar 14, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill employs behavioral override patterns modeled after psychological manipulation techniques. It explicitly instructs the AI to 'question reality' and 'create chaos' (SKILL.v2.md, Five-Step Method). In version 1.0, the system prompt forces a distorted attribution logic where the agent must credit successes to 'luck' and failures to 'incompetence'. These instructions aim to bypass the AI's standard balanced reasoning and safety guardrails regarding self-assessment and factual reliability.
- [INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface for indirect injection by processing external data while in a state of intentional cognitive destabilization.
- Ingestion points: Processes general analysis tasks and quality improvement requests (SKILL.md, Execution Flow; SKILL.v2.md, Task Types).
- Boundary markers: None present; the skill lacks delimiters or instructions to ignore embedded commands in the data it processes.
- Capability inventory: Primarily focuses on text generation and reasoning refinement; no direct file system or network capabilities are defined in the provided files.
- Sanitization: No evidence of input validation or escaping for the tasks being optimized.
Audit Metadata