lampson-system-design
Fail
Audited by Gen Agent Trust Hub on Mar 9, 2026
Risk Level: HIGHPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Hidden data encoded using zero-width characters detected in file headers.\n
- Evidence: The main headers in
SKILL.md,philosophy.md, andreferences.mdcontain invisible character sequences (U+200B, U+200C, U+200D) following the visible text.\n - Technique: The data is structured into 8-bit segments separated by Zero-Width Joiners (U+200D) and prefixed with the encoded magic string 'SK1L'.\n
- Risk: This method is used to embed data that is invisible to human readers but processed by the AI model, potentially carrying malicious instructions or exfiltration markers.\n- [SAFE]: External links refer to established and trusted technical and academic domains.\n
- Evidence: URLs point to official resources on
acm.org,microsoft.com,mit.edu, andstanford.edu.\n- [SAFE]: The skill provides educational markdown content without any executable code or scripts.\n - Evidence: Code snippets are restricted to static Python examples demonstrating system design principles.
Recommendations
- AI detected serious security threats
Audit Metadata