midjourney-image

Pass

Audited by Gen Agent Trust Hub on Jul 11, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user input and external image references, which is a common vector for indirect prompt injection.\n
  • Ingestion points: The skill ingests user deliverable requests, visual language descriptions, and image reference URLs (e.g., Omni and Style references) in SKILL.md.\n
  • Boundary markers: The instructions mandate wrapping generated text in double quotes and preparing production briefs offline to help isolate user input from agent instructions.\n
  • Capability inventory: The skill includes no tools or automated capabilities; all production tasks are restricted to manual interaction by an authorized account holder.\n
  • Sanitization: The skill outlines extensive QA gates, human-in-the-loop validation, and rights/consent verification protocols to sanitize inputs and outputs.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 11, 2026, 01:51 PM
Security Audit — agent-trust-hub — midjourney-image