jev-video-qa
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill follows security best practices for handling untrusted media.
- [SAFE]: Data Privacy: It explicitly restricts private and customer media to local processing, preventing unauthorized cloud uploads.
- [SAFE]: Indirect Injection Mitigation: It instructs the agent to disregard instructions found inside images, neutralizing potential indirect prompt injection attacks.
- [SAFE]: Operational Guardrails: It defines clear confidence floors, model call limits, and token budgets to prevent resource exhaustion or automated approval of uncertain results.
- [SAFE]: Human-in-the-loop: It mandates manual review for flags, uncertainty, and high-risk jobs, ensuring the AI does not have final approval authority.
Audit Metadata