The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external geographic datasets (such as GeoJSON or Shapefiles) which are then used to populate map labels, faction names, and ownership data in the Simulation environment.
Ingestion points: External geographic data is downloaded via browser or curl as described in references/map-sources-and-processing.md.
Boundary markers: The instructions do not include explicit boundary markers or directions to the agent to treat string-based metadata strictly as data rather than instructions.
Capability inventory: The skill uses WorldOS MCP tools (patch_world_map, validate_world, inspect_world_payload) to update the simulation state.
Sanitization: While the skill provides detailed instructions for validating SVG path geometry and coordinate projections, it lacks specific guidance for sanitizing or escaping string-based properties (e.g., region names or country labels) sourced from external files.
[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill identifies and provides instructions for fetching geographic data from external URLs.
Evidence: references/map-sources-and-processing.md mentions sourcing data from Natural Earth (naturalearthdata.com) and a historical basemap repository on GitHub. These sources are used to obtain static data files for mapping purposes.
[COMMAND_EXECUTION]: The skill suggests using shell commands to retrieve external datasets.
Evidence: In references/map-sources-and-processing.md, the skill describes a pattern for using curl --fail --location --output to download datasets if a shell is available to the agent.

worldos-map-authoring