image
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill reads from a local file,
.agents/marketing-context.md, to determine marketing context, which introduces a potential attack surface if that file contains untrusted or malicious instructions. - Ingestion points:
SKILL.md(reads.agents/marketing-context.md) - Boundary markers: Absent
- Capability inventory: Local execution of image processing tools
- Sanitization: Absent
- [COMMAND_EXECUTION]: The skill documentation explicitly mentions the use of local command-line tools for image optimization tasks.
- Evidence: Mentions of
ImageMagick,cwebp, andavifencinreferences/image-system.mdas common tools for photographic delivery and vector work.
Audit Metadata