pyvista
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external 3D mesh and voxel data files, creating a surface for indirect prompt injection attacks.
- Ingestion points: The
pv.read()function used inSKILL.mdandscripts/visualize_surface.pyingests data from external files such as.vtk,.stl,.obj, and.vti. - Boundary markers: There are no explicit instructions or delimiters provided to the agent to ignore or isolate natural language instructions that might be embedded within metadata or comments of these files.
- Capability inventory: The skill includes file system read/write capabilities, specifically for reading data meshes and exporting screenshots (
plotter.screenshot()) or interactive HTML (plotter.export_html()). - Sanitization: There is no evidence of sanitization or validation of the content within the 3D data files before they are processed by the PyVista library and the agent context.
Audit Metadata