---

name: remote-geospatial-data description: "This sub-skill routes authenticated Google Earth Engine catalog, land-cover, building-footprint, and bounded satellite-timelapse workflows for geospatial applications." disable-model-invocation: true metadata: disco-role: operating license: MIT

Remote geospatial data

Use this sub-skill when the task needs geemap's Google Earth Engine integration for catalog search, NLCD, Dynamic World/ESA/ESRI land cover, Microsoft building footprints, or a Landsat, Sentinel-2, GOES, MODIS, or NAIP timelapse. Route to earth-engine-api.md for catalog, asset, authentication, and ROI decisions. Route to timelapse-workflows.md for a timelapse family and bounded execution plan. Use scripts/validate_ee_config.py before a remote request and troubleshooting.md when a request or media conversion fails.

Router

1. Decide whether the operation is catalog/NLCD, land-cover comparison, building footprints, or a timelapse. Do not use this route for generic Leafmap layer composition or local dashboard data.
2. Normalize the input ROI. Prefer a small sample ROI. For an uploaded GeoJSON/KML-derived GeoDataFrame, confirm that its CRS is WGS84 (EPSG:4326) or reproject it before calling gdf_to_ee(gdf, geodesic=False).
3. Validate dates, JSON visualization/palette values, and GeoJSON structure without contacting Earth Engine. The bundled validator checks token-variable presence only; it never authenticates and never exposes a token value.
4. Set USE_FOLIUM=1 before import geemap.foliumap as geemap. This makes the Streamlit process select the folium map backend used by this app instead of leaving backend selection implicit.
5. Authenticate only at the explicit Earth Engine boundary, using the exact initialization contract in earth-engine-api.md. Search or instantiate the intended asset, then make one bounded request.
6. Select one timelapse family, keep ROI/date/frame/output bounds explicit, and verify the requested GIF/MP4 exists before presenting it. Treat remote asset availability and imagery presence as runtime facts.

Hard boundaries

• Never put a token or token value in a prompt, source file, log, or result.
• Never report that an asset exists, imagery was returned, or a timelapse completed unless an authenticated Earth Engine call and local output check actually observed it. This skill does not perform live remote verification.
• Do not submit malformed JSON, an invalid/empty ROI, a reversed date range, or an unbounded global/time-span timelapse.
• Keep review notes and evidence outside this runtime subtree; the references here are operational instructions, not links to source application files.