image-prompt

Pass

Audited by Gen Agent Trust Hub on Nov 3, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill follows secure design principles by requiring generation prompts to be written as a 'labelled spec' rather than free-form paragraphs, which prevents ambiguity and improves auditability.
  • [EXTERNAL_DOWNLOADS]: The skill references utility scripts (imgfacts.py, refute.py) and a shared registry (../../image-registry) located in sibling directories of the vendor's repository. These are identified as internal project dependencies and do not involve untrusted remote sources.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided 'briefs' to construct image prompts. It effectively mitigates potential indirect prompt injection by providing strict playbooks for prompt structure and exclusions, and by requiring the agent to quote text verbatim when it must appear in an image, preventing the execution of embedded instructions.
  • [COMMAND_EXECUTION]: Uses local Python scripts via the Bash tool to perform technical checks, such as validating image dimensions against a registry. This is a legitimate use of the tool within the skill's defined scope.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 3, 2026, 06:00 PM