geospatial-data-discovery
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references established geospatial catalogs (e.g., Microsoft Planetary Computer, Overture Maps, USGS) and GitHub repositories for open-source GIS tools. These are well-known, industry-standard resources.
- [COMMAND_EXECUTION]: Includes documentation and code snippets for common GIS command-line utilities like GDAL, OGR, Osmium, and PDAL. These are used for their intended primary purpose of data format conversion and local spatial analysis.
- [CREDENTIALS_UNSAFE]: The skill explicitly instructs against hardcoding secrets, mandating the use of credential references (e.g., environment variables like GOOGLE_APPLICATION_CREDENTIALS) and read-only warehouse connections, adhering to security best practices.
- [INDIRECT_PROMPT_INJECTION]: The skill acts as a data discovery surface by reading external metadata and catalog records.
- Ingestion points: STAC catalogs, Portolan collections, and provider metadata pages (detailed in references/data-sources.md).
- Boundary markers: Instructions require the agent to cite checked pages and dates, and to label recommendations as unverified if source metadata is unavailable.
- Capability inventory: Access to Python GIS libraries (pystac-client, geopandas), Shell utilities (gdal, ogr2ogr), and DuckDB for analytical processing.
- Sanitization: Instructions focus on verifying field definitions and code lists against authoritative provider metadata rather than relying on inferred or remembered values. (Severity: Low).
Audit Metadata