image
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest marketing context from local files such as
.agents/product-marketing.md. While this is a standard feature for personalizing output, it creates a potential surface for indirect prompt injection if those files contain untrusted content. - Ingestion points:
.agents/product-marketing.md,.claude/product-marketing.md, andproduct-marketing-context.mdreferenced inSKILL.md. - Boundary markers: The instructions do not specify explicit delimiters or warnings to ignore instructions within these context files.
- Capability inventory: Includes file reading and shell command execution for image processing and network inspection.
- Sanitization: No specific sanitization or validation of the context file content is documented.
- [COMMAND_EXECUTION]: The documentation includes example shell commands for image optimization (e.g.,
cwebp,mogrify,jpegoptim) and web inspection (e.g.,curl). While these are legitimate utilities, thecurlexample involves network access based on user-provided or context-derived URLs.
Audit Metadata