don-norman-principles-audit
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill identifies that inputs like
interface_description,screenshots_or_links, anduser_tasksare untrusted third-party data. It provides explicit defense instructions to sanitize this data and disregard any hidden commands or formatting intended to manipulate the agent. While the attack surface exists (as with any skill processing external data), the instructions follow best practices for mitigation. Severity is set to low per standard classification for surface-only detection.
Audit Metadata