The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs project migration and inspection by executing C# snippets within a live Unity Editor using the unity command eval tool. This is the intended primary mechanism for interacting with the Unity project environment.
[DYNAMIC_EXECUTION]: The skill generates C# code at runtime, including statement blocks for immediate evaluation and full class files (e.g., MaterialSnapshot, CustomFullscreenEffectFeature) to be saved in the project's Assets/Editor/ directory for compilation and execution. This allows for complex state management and material restoration during the migration process.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from the Unity project, such as asset names, shader source code, and Unity Console logs.
Ingestion points: Unity project files, scene metadata, and logs retrieved via Unity.GetConsoleLogs.
Boundary markers: The skill does not use specific delimiters but mitigates risk through strict verification requirements, ensuring that tool logs are checked against the actual saved state of the project.
Capability inventory: The skill has high capabilities, including arbitrary C# execution in the Editor, file system writes for visual captures, and package management.
Sanitization: The skill relies on phased execution gates and human-in-the-loop rollback confirmations rather than automated sanitization of project data.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill captures Editor visuals (Scene and Game views) and saves them as PNG files in the system's temporary directory (System.IO.Path.GetTempPath()) to allow the agent to verify migration results. These operations are local and no external network exfiltration of project data was observed.

migrate-birp-to-urp