vanrossum-pythonic-style
Warn
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: MEDIUMPROMPT_INJECTIONOBFUSCATION
Full Analysis
- [OBFUSCATION]: The skill employs Unicode steganography using Zero-Width characters (U+200B, U+200C, U+200D) and Word Joiner markers (U+2060) to hide a payload at the end of the main Markdown header in
SKILL.md. This technique is specifically designed to evade human review while remaining accessible to the AI's processing context. - Evidence: The string
...is appended to the text '# Guido van Rossum Style Guide'. - [PROMPT_INJECTION]: The hidden steganographic payload serves as a covert prompt injection vector. By embedding instructions that are invisible to the human user, the skill can attempt to override the agent's behavior, bypass safety constraints, or exfiltrate information without the user's knowledge.
- Evidence: The presence of a deliberate encoding scheme (mapping Zero-Width Space and Zero-Width Non-Joiner to bits) within a supposedly static documentation file strongly indicates an attempt to influence the agent's internal state outside of visible instructions.
Audit Metadata