The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill includes adapters (adapters/coplay-unity-mcp.md, adapters/unity-cli.md) that enable the agent to execute arbitrary C# code within the Unity Editor process.
The Coplay adapter uses an execute_code method to compile and run C# method bodies in memory using Roslyn or CodeDom.
The Unity CLI adapter uses unity eval to evaluate C# expressions inside the connected Editor.
While documented as diagnostic tools with warnings that safety checks are not a security sandbox, these mechanisms grant the agent significant execution authority over the host environment.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest a wide variety of project files that could contain malicious instructions if they originate from untrusted sources (e.g., third-party assets or external contributions).
Ingestion points: The agent establishes a "baseline" by reading Packages/manifest.json, ProjectSettings/ProjectVersion.txt, and source files in the Assets/ directory.
Boundary markers: The skill lacks explicit boundary markers or instructions to the agent to treat external project content as untrusted data.
Capability inventory: The agent possesses powerful capabilities, including file system writes, shell command execution via the Unity CLI, and in-Editor code execution via MCP adapters.
Sanitization: No sanitization or validation of ingested file content is performed before processing by the agent.
[COMMAND_EXECUTION]: The skill relies on shell command execution to interact with the Unity CLI tool and perform project operations.
Commands include unity status, unity command, unity eval, and installation of packages via unity pipeline install.
These commands allow the agent to modify the project configuration and system environment directly.

unity-game-dev