ingest-screenshot
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided images, creating a potential surface for indirect prompt injection via text embedded in the images. \n
- Ingestion points: User-specified image files via the
<image-path>argument inSKILL.md. \n - Boundary markers: The instructions do not explicitly direct the model to ignore any textual instructions that might be contained within the analyzed images. \n
- Capability inventory: The skill uses
Read,Write, and limitedBash(mkdir, realpath, file) tools to process data and save output. \n - Sanitization: No specific sanitization of image content is performed prior to the vision-based analysis.
Audit Metadata