scarface-mean-reversion
Fail
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: HIGHOBFUSCATIONPROMPT_INJECTION
Full Analysis
- [OBFUSCATION]: The main H1 title in
SKILL.mdcontains a sequence of hidden zero-width Unicode characters (\u200B,\u200C,\u200D,\u2060). This steganographic technique is used to embed data or instructions that are invisible to a human observer but are parsed by the underlying language model. - [PROMPT_INJECTION]: The hidden payload within the skill title is identified as a steganographic prompt injection. This is a sophisticated method of delivering unauthorized instructions to the agent, typically used to bypass security filters, ignore previous constraints, or force the agent into an unrestricted state without the user's knowledge.
Recommendations
- AI detected serious security threats
Audit Metadata