image-deliver
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements a structured image delivery workflow using local Python utilities (
strip.py,imgfacts.py,catalog.py,refute.py) for processing and verification. These tools are referenced via relative paths within the project environment. - [SAFE]: Detailed privacy protocols are established for metadata handling. The skill provides explicit instructions and a tool (
strip.py) to remove sensitive information such as GPS coordinates, camera details, and software timestamps before images are shipped to production. - [SAFE]: The asset catalog generation tool (
catalog.py) includes security measures to prevent directory traversal attacks by refusing to process or link images and recipes located outside the designated asset root. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted image metadata through its processing tools. While this constitutes an attack surface for indirect prompt injection, the risk is mitigated by the workflow's focus on technical stripping and sanitization of metadata chunks, and the requirement for objective measurement over trusting data-driven assertions.
Audit Metadata