The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines an API surface that allows for the ingestion of untrusted data via external URLs in material definitions.
Ingestion points: The maps field in detail.material (documented in reference.md) accepts arbitrary URLs for textures, which can be provided via scene documents or user input.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat content from these external URLs as untrusted or to disregard potential instructions within the data.
Capability inventory: The skill refers to @jgengine/shell capabilities that use useTexture and useLoader to perform network operations to fetch data from these URLs.
Sanitization: The documentation does not specify any sanitization, validation, or filtering of the content retrieved from external URLs.

jgengine-world