google-material-design
Fail
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: HIGHOBFUSCATIONPROMPT_INJECTION
Full Analysis
- [OBFUSCATION]: The skill header in
SKILL.mdcontains a long sequence of zero-width Unicode characters (U+200B Zero-Width Space, U+200C Zero-Width Non-Joiner, and U+200D Zero-Width Joiner) hidden at the end of the text "# Google Material Design". These characters are invisible to the user but can encode data or instructions for the AI model. - [PROMPT_INJECTION]: The use of Unicode steganography is a recognized method for delivering hidden prompt injection payloads. By embedding these characters in instructional files, an attacker can attempt to silently override the agent's behavior, safety protocols, or initial system prompts while the document appears to be a standard guide on Material Design.
Recommendations
- AI detected serious security threats
Audit Metadata