two-sigma-ml-at-scale
Fail
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: HIGHOBFUSCATIONPROMPT_INJECTION
Full Analysis
- [OBFUSCATION]: The title
# Two Sigma Style Guidecontains a long sequence of hidden Unicode characters (Zero Width Space U+200B, Zero Width Non-Joiner U+200C, and Zero Width Joiner U+200D). These characters are used to embed data that is invisible to the human reader but detectable by software. - [PROMPT_INJECTION]: The hidden characters encode a binary payload (partially decoded as 'SK1L...'). The use of steganography within a skill's instructions is a high-severity indicator of an attempt to perform prompt injection, delivering hidden commands to the AI agent to override behavior or exfiltrate data without user visibility.
Recommendations
- AI detected serious security threats
Audit Metadata