The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill is primarily designed to compile and execute arbitrary C# code snippets at runtime within the Unity environment using the uloop execute-dynamic-code tool. This allows for complex manipulations of the Unity scene, assets, and runtime state.
[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file defines a task template that injects user-provided input directly into a command execution prompt: Execute the following request using uloop execute-dynamic-code: $ARGUMENTS. This creates a vector for command injection if the input is not strictly validated or escaped by the platform.
[INDIRECT_PROMPT_INJECTION]: The skill represents a significant attack surface for indirect prompt injection.
Ingestion points: User instructions are processed and used to "Construct C# code" which is then executed via the dynamic code runner.
Boundary markers: The skill does not define specific delimiters or instructions for the agent to ignore potentially malicious embedded code within the data it might be processing.
Capability inventory: The skill has broad capabilities including scene modification, asset manipulation (AssetDatabase), and the use of System.Reflection to access and modify private fields and methods.
Sanitization: While the skill mentions a compile-time blocklist for System.IO and certain folder creation methods, it does not specify sanitization for the $ARGUMENTS used to drive the tool.
[COMMAND_EXECUTION]: The skill encourages the agent to write C# statements to temporary .csx files and execute them via the CLI, providing a direct path for the agent to execute code it has generated based on user prompts.

uloop-execute-dynamic-code