the-visa-interview
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill operates entirely within the text-based interaction environment. It does not attempt to access external network resources, write to the file system, or execute shell commands.\n- [SAFE]: The instructions include robust anti-patterns that explicitly forbid the agent from assisting with fraudulent behavior, concealment of facts, or evading lawful questions, which serves to protect the integrity of the agent's behavior.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided information regarding their visa type, personal situation, and paperwork (SKILL.md). While this represents an attack surface, the risk is mitigated by a total lack of dangerous capabilities (no subprocess, file-write, or network calls identified in scripts) and the presence of boundary markers like the 'out of character' debrief sections and truthfulness banners. The simulation is structurally designed to prioritize consistency and truth, providing inherent logic-based sanitization.
Audit Metadata