levien-native-ui-mastery
Warn
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: MEDIUMOBFUSCATIONPROMPT_INJECTION
Full Analysis
- [OBFUSCATION]: The skill uses zero-width Unicode characters (U+200B, U+200C, U+200D, and U+2060) to embed hidden data within the headers of its markdown files. These characters are invisible to users in standard text editors but are visible to the AI model during processing.\n
- Evidence: Large blocks of zero-width characters follow the main '# ' headers in
SKILL.md,philosophy.md, andreferences.md.\n- [PROMPT_INJECTION]: The embedding of hidden Unicode strings is a established method for steganographic prompt injection. This allows the author to insert instructions or context that can override the agent's behavior or safety guidelines without the user's knowledge, which is a significant security and transparency risk.
Audit Metadata