The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The references/pitfalls.md file provides setup instructions for macOS users to run sudo xcode-select -s to configure the system's Xcode path. This command requires administrative privileges and modifies system-wide development settings.
[DYNAMIC_EXECUTION]: The skill leverages ProgrammaticToolset and BlueprintTools to dynamically generate and execute logic within the Unreal Editor process. The ProgrammaticToolset allows for running sandboxed Python scripts, while BlueprintTools uses a custom domain-specific language (DSL) to author and compile Unreal Engine graphs at runtime.
[EXTERNAL_DOWNLOADS]: The references/tool-surface.md file references npx @modelcontextprotocol/inspector, a well-known diagnostic tool for MCP servers, which is downloaded and executed from the NPM registry.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from external sources and user prompts to drive complex engine operations.
Ingestion points: User-provided scene descriptions, 3D asset metadata during import (e.g., FBX files), and responses from various Unreal Engine tool calls.
Boundary markers: Not present. The skill does not define specific delimiters or instructions to ignore commands potentially embedded in the data it processes.
Capability inventory: The skill possesses capabilities for file modification (AssetTools.write_file), actor spawning, property manipulation, and dynamic script execution.
Sanitization: The skill relies on platform-level type validation but does not provide details on sanitizing or escaping content before it is processed by the engine or used in script generation.

unreal-mcp