color-management
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The Python scripts in the scripts directory execute shell commands via subprocess.run to interface with ImageMagick, Exiftool, ArgyllCMS, and LittleCMS. These calls are essential for the skill's primary color science functions and use list-based arguments to mitigate shell injection risks.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata from image files and ICC profiles, such as profile descriptions and copyright text, which are presented to the agent's context without boundary markers or sanitization.
- Ingestion points: Image and ICC files provided as input to analysis scripts.
- Boundary markers: Absent in script output and generated reports.
- Capability inventory: File system access and execution of multiple CLI tools via subprocess across all analysis scripts.
- Sanitization: Extracted metadata strings are not validated or escaped before being displayed.
Audit Metadata