product-illustrator
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external CAD files including Python-based CadQuery scripts (.py), OpenSCAD files (.scad), and mesh files (.stl, .step). These files represent an attack surface where malicious instructions could be embedded to influence agent behavior or prompt outputs for image generation tools. \n
- Ingestion points: .py, .scad, .stl, and .step files read from the project directory. \n
- Boundary markers: None identified; the instructions do not specify delimiters or safety warnings for processing external CAD content. \n
- Capability inventory: The skill utilizes CadQuery (Python execution), environment variable modification (LD_LIBRARY_PATH), and network operations (ComfyUI API). \n
- Sanitization: No sanitization or validation of the CAD source code or parameters is described.\n- [COMMAND_EXECUTION]: The skill instructions involve executing shell commands to modify the environment, specifically setting the
LD_LIBRARY_PATHenvironment variable. Manipulating library search paths is a sensitive operation that could be misused to force the application to load malicious shared libraries if the user-supplied path is compromised.\n- [DYNAMIC_EXECUTION]: The workflow relies on executing CadQuery scripts which are written in Python. Since these scripts define geometry through executable code, the skill performs dynamic execution of content that may originate from external sources. Additionally, the configuration of library paths at runtime falls under dynamic loading techniques.
Audit Metadata