kay-inventing-the-future
Fail
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: HIGHOBFUSCATIONPROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [OBFUSCATION]: The skill's main header ('# Alan Kay Style Guide') contains a long sequence of non-printable, zero-width Unicode characters (U+200B, U+200C, U+200D). These characters are invisible in standard text viewers but are processed by language models, effectively hiding a secondary payload from the human user.
- [PROMPT_INJECTION]: The hidden Unicode sequence is a steganographic prompt injection. This technique is specifically designed to bypass user review and safety filters by providing the model with instructions that the user cannot see. This allows an attacker to potentially override the agent's behavior or safety guidelines.
- [METADATA_POISONING]: The malicious hidden content is placed directly in the skill's title metadata. Since the title is a primary piece of information loaded into the agent's context, this ensures the hidden instructions are processed at the earliest possible stage of interaction.
Recommendations
- AI detected serious security threats
Audit Metadata