The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exposes an attack surface for indirect prompt injection by processing data from the Unity scene and project assets (such as object names, component properties, and asset metadata) which could be influenced by external collaborators or untrusted project files.
Ingestion points: Data enters the agent's context through query skills like batch_query_gameobjects, batch_query_components, and batch_query_assets using parameters such as queryJson and searchFilter.
Boundary markers: The instructions do not define specific delimiters or warnings for the agent to treat data retrieved from the Unity environment as untrusted.
Capability inventory: The skill possesses significant write capabilities, including modifying properties (batch_preview_set_property), renaming objects (batch_preview_rename), and deleting temporary objects (batch_cleanup_temp_objects).
Sanitization: The skill description does not specify sanitization or escaping procedures for scene-derived data before it is processed or presented to the user.
Note: This surface is inherent to the skill's primary purpose of batch processing scene data and is mitigated by the mandatory preview-confirmation gate (confirmToken) which ensures human-in-the-loop validation for all bulk operations.

unity-batch