gemgis
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external GIS data files (Shapefiles, GeoJSON, GeoPackage, and GeoTIFF) whose attributes are ingested into the agent's context during data preparation and summarization workflows.
- Ingestion points: Spatial data is loaded from user-specified files using
geopandas.read_file()andrasterio.open()withinscripts/prepare_gempy_data.pyand documented examples. - Boundary markers: There are no explicit delimiters or instructions to the agent to disregard potential instructions that might be embedded within the metadata or text attributes of the geological data.
- Capability inventory: The skill possesses file system access, including the ability to read GIS files and write processed data to CSV and text files (
scripts/prepare_gempy_data.pylines 218, 223, and 241). - Sanitization: While the scripts perform technical validation such as geometry repair and column verification, there is no filtering or sanitization of text attributes to prevent an agent from interpreting them as instructions.
Audit Metadata