The Agent Skills Directory

[COMMAND_EXECUTION]: The pipeline.py script in the Tartu example optionally utilizes Docker to run a pinned QGIS container (qgis/qgis:3.44.3) for compiling GIS projects. This execution is performed with local user permissions and targets a specific, stated objective of providing a first-class QGIS view for the analysis.
[DYNAMIC_EXECUTION]: The skill employs duckdb for spatial query execution and dynamically generates a Python-based QGIS builder script to automate project archival. These mechanisms are integral to the skill's primary function of enabling reproducible GIS analysis and do not exhibit malicious characteristics.
[EXTERNAL_DOWNLOADS]: The provided examples download spatial data from authoritative Estonian government infrastructure, including the Estonian Environment Agency (WFS), Tartu City Government (ArcGIS REST), and Maa- ja Ruumiamet (S3/Bulk). These downloads are documented, serve the stated analytical purpose, and target trusted regional services.
[INDIRECT_PROMPT_INJECTION]: The skill processes external GIS data formats (GeoPackage, GeoJSON, WFS) that could theoretically contain malicious content. However, the analysis identified that the tool implements proper sanitization and escaping logic for its generated dashboard output, and primarily targets authoritative public sector data sources.
[SAFE]: The skill demonstrates safe practices, such as delegating credential management to environment variables and providing extensive validation gates to ensure analytical integrity and provenance. No prompt injection attempts, obfuscation, or persistence mechanisms were found.

spatial-sql