geological-modelling

Pass

Audited by Gen Agent Trust Hub on Sep 13, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow that ingests external spatial data (Shapefiles, GeoTiff, Borehole data) and has the capability to write files (VTK format). This combination creates a surface for indirect prompt injection where malicious input data could potentially influence downstream automated actions.
  • Ingestion points: The skill reads geological_contacts.shp, orientations.shp, dem.tif, and boreholes.shp using geopandas and rasterio as described in SKILL.md.
  • Boundary markers: The instructions do not implement delimiters or specific warnings for the agent to ignore instructions that might be embedded within metadata or data fields of these spatial files.
  • Capability inventory: The workflow includes file system write operations (grid.save('geological_model.vtk') in SKILL.md) and interactive 3D visualization capabilities.
  • Sanitization: There is no explicit sanitization, validation, or filtering of the external data content mentioned in the orchestration steps.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 13, 2026, 10:05 AM
Security Audit — agent-trust-hub — geological-modelling