The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of agent-generated C# code within a live Unity Editor session. As documented in the SECURITY.md file, this is accomplished using the unity command eval shell command.
[DYNAMIC_EXECUTION]: The skill provides instructions and code templates in references/rulesets-and-actions.md that allow the agent to dynamically construct and execute C# scripts to modify RuleSet assets at runtime.
[DATA_EXFILTRATION]: The ATTAssistantUtilities.CreateImporter and ATTAssistantUtilities.CanImportFile functions documented in references/create-importer.md accept absolute system paths for the filePath parameter. This capability allows the agent to interact with and verify the existence of files anywhere on the local file system.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data in the form of 3D models and point clouds, creating a surface for indirect prompt injection.
Ingestion points: Files are imported from absolute paths in references/create-importer.md and processed by the toolkit.
Boundary markers: No specific delimiters or safety warnings are implemented to prevent the agent from acting on instructions potentially embedded in file metadata or property fields.
Capability inventory: The skill has access to shell execution via unity command eval and the ability to modify project assets as described in references/rulesets-and-actions.md.
Sanitization: There is no evidence of sanitization or filtering for untrusted data imported during the 3D asset workflow.

asset-transformer-toolkit