The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines patterns that ingest external data and display it in the user interface, creating a potential surface for injection if the data source is attacker-controlled.
Ingestion points: External GeoJSON data structures are processed in SKILL.md (via geojsonData), examples/advanced.md (via cityData), and examples/core.md (via featureCollection).
Boundary markers: No explicit delimiters or instructions are provided to the agent to treat embedded content in GeoJSON properties as untrusted.
Capability inventory: The skill utilizes bindPopup and bindTooltip methods, which render strings directly into the DOM, and L.DomUtil.create for custom UI controls.
Sanitization: Code examples demonstrate direct binding of property values (e.g., feature.properties.name and feature.properties.population) to popup content without implementing escaping or sanitization logic.

web-maps-leaflet