The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process a wide variety of external geospatial data formats (e.g., GeoJSON, Raster imagery, NetCDF) which can serve as an ingestion point for indirect prompt injection attacks.
Ingestion points: Data is loaded using standard libraries across multiple files: gpd.read_file (GeoPandas), rasterio.open (Rasterio), xr.open_dataset (Xarray), and fiona.open (Fiona).
Boundary markers: The code examples do not include specific delimiters or instructions for the agent to ignore potentially malicious text embedded within these data formats.
Capability inventory: The skill includes capabilities to write files to the local system (rasterio.open(..., 'w'), gdf.to_file) and execute scientific command-line tools via subprocess.run.
Sanitization: There is no evidence of text sanitization or filtering for content extracted from geospatial data files before it is processed or used by the agent.
[COMMAND_EXECUTION]: The skill provides examples for integrating with external scientific GIS software like SAGA GIS and GRASS GIS using command-line executions. These examples utilize subprocess.run with argument lists, which is a secure practice to prevent shell injection, and the commands are strictly limited to the intended scientific functionality.
[EXTERNAL_DOWNLOADS]: The skill references and fetches data from established and trusted scientific organizations such as Microsoft (Planetary Computer), Google (Earth Engine), the European Space Agency (ESA), and the U.S. Geological Survey (USGS). These downloads are fundamental to the skill's purpose and originate from well-known, safe sources.

geomaster