Data Globe

Edge turns location rows into a world you can explore: see where activity concentrates, filter it, then select a point to meet the entities behind it. Use the included starter as a working base; adapt its labels to the user's dataset while keeping the included visual system: warm light-land globe on a night ocean, ember point glow, atmosphere rim, translucent rails, location rankings and entity wall. The optional dark-land button changes contrast without changing the data.

Explore Rocketmap, the live production reference, and read references/rocketmap.md for its relationship to this reusable starter.

Start from data, keep the geography honest

Inspect column names, row count, coordinate coverage, categories, weight distribution and date meaning. Record the source URL, licence, download date and exclusion rules. Preserve source entity IDs. Default weight is one entity when no meaningful metric exists; do not present invented activity or population. A modification date is not a founding/opening/event date.

The canonical row is id, name, lat, lng, weight, category, country, city, date, with optional domain, logo, url. Coordinates are [longitude, latitude]; country is ISO2; dates are ISO YYYY-MM-DD. Accept finite coordinates in range and nonnegative finite weights. Deduplicate by source ID, not name. Invalid or unplaced rows go into the audit with their reason. Valid (0,0) may be real: check the dataset rather than banning it universally.

For city/country rows, read references/geocoding.md. Resolve offline against GeoNames with exact aliases, ISO2 and admin constraints. Never pick the biggest city from ambiguous matches, ignore a state suffix, or invent a country-centroid city. Keep unresolved rows explicit. For authoritative city anchors use --group city; coordinate inputs use point or grid.

Working starter (Python 3 + a browser; no build system)

Locate this skill folder as SKILL, then run: