QGIS Geospatial

QGIS

Use this skill for deterministic, file-based GIS processing through qgis_process.

Core Rules

• Confirm source files, destination files, and target CRS before running commands.
• Prefer writing to a new output path; do not overwrite source data unless user explicitly asks.
• Discover algorithm parameters from qgis_process help before execution.
• If output format is not specified, default to GPKG for vector and GeoTIFF for raster.
• For batch operations, run one representative file first, then scale to the full set.

Standard Workflow

1. Validate tool and inspect capabilities:

qgis_process --version
qgis_process list

1. Find and inspect a specific algorithm:

qgis_process list | rg -i "clip|buffer|reproject|merge|warp|dissolve"
qgis_process help native:clip

1. Execute with explicit parameters:

qgis_process run  -- \
  INPUT= \
  OUTPUT=

1. Verify output exists and report summary (path, layer count/bands, CRS, extent).

Common Patterns

# Reproject vector layer
qgis_process run native:reprojectlayer -- \
  INPUT=./data/input.gpkg \
  TARGET_CRS=EPSG:4326 \
  OUTPUT=./out/reprojected.gpkg

# Clip vector layer by overlay
qgis_process run native:clip -- \
  INPUT=./data/roads.gpkg \
  OVERLAY=./data/boundary.gpkg \
  OUTPUT=./out/roads_clipped.gpkg

# Buffer vector features
qgis_process run native:buffer -- \
  INPUT=./data/points.gpkg \
  DISTANCE=100 \
  SEGMENTS=8 \
  DISSOLVE=false \
  OUTPUT=./out/points_buffer_100m.gpkg

Safety Boundaries

• Do not guess units; confirm whether distance/area is in degrees or projected units.
• If CRS is missing/invalid, stop and ask for the intended CRS before processing.
• Avoid lossy conversion unless requested (for example dropping attributes or precision).
• For long batch runs, provide dry-run plan first: algorithm, inputs, outputs, and overwrite policy.

OpenClaw + ClawHub Notes

• Keep commands shell-safe and reproducible in OpenClaw terminal sessions.
• Keep this skill portable: no hardcoded machine paths, credentials, or private endpoints.
• For clawhub.ai publication, keep versioning/changelog semver-driven and update examples only with generic sample data.