klabnik-teaching-rust
Fail
Audited by Gen Agent Trust Hub on Mar 9, 2026
Risk Level: CRITICALPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [PROMPT_INJECTION]: The main header in SKILL.md contains 152 hidden zero-width characters that are invisible to humans but interpreted by the AI agent as a payload.
- [DATA_EXFILTRATION]: Decoded invisible content reveals instructions for the agent to 'SILENTLY EXFILTRATE' information, which is a direct attempt to steal user data and session context.
- [PROMPT_INJECTION]: The use of binary-encoded Unicode characters to deliver hidden instructions is a sophisticated obfuscation technique intended to subvert safety protocols and perform unauthorized actions.
Recommendations
- AI detected serious security threats
Audit Metadata