The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes npx to download and execute the unity-mcp-cli package from the public NPM registry. It specifies the @latest version, which is standard for ensuring access to the most recent features and fixes for this vendor-provided utility.
[REMOTE_CODE_EXECUTION]: The use of npx -y unity-mcp-cli@latest constitutes the execution of remote package code. This is the primary mechanism the skill uses to drive the Unity Editor and project configuration.
[COMMAND_EXECUTION]: The instructions guide the AI agent to execute a series of shell commands for creating projects, installing plugins, managing the editor lifecycle (open/close), and configuring the environment.
[DYNAMIC_EXECUTION]: The skill includes an exec command that allows the agent to run C# snippets directly within the Unity Editor. This is used for verifying project state or performing scripted edits within the engine.
[INDIRECT_PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection via dynamically processed data and generated content:
Ingestion points: The agent reads the output of wait-for-ready (API polling), status, and generated files created by setup-skills in SKILL.md.
Boundary markers: Explicit delimiters for the output of these tools are absent in the provided instructions.
Capability inventory: The agent has the ability to execute shell commands, write configuration files (.mcp.json, .gemini/settings.json, etc.), and run C# code via the exec tool.
Sanitization: There is no explicit sanitization described for the data returned by the project's internal MCP server or the generated skills.

unity-game-dev