The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends installing several third-party packages and cloning repositories from independent GitHub accounts such as chongdashu/unreal-mcp, kvick-games/UnrealMCP, ChiR24/Unreal_mcp, and others. These sources are not recognized as trusted vendors or established organizations, posing a supply chain risk.
[COMMAND_EXECUTION]: The skill includes instructions and tools for executing system-level commands, specifically unreal_compile for hot reloading, unreal_build for platform builds, and the execution of automation tests via UnrealEditor-Cmd.exe. These operations run with the user's local privileges.
[REMOTE_CODE_EXECUTION]: The unreal_python_exec tool allows for the execution of arbitrary Python commands within the Unreal Engine environment. This capability represents a significant execution surface that could be exploited if malicious input is supplied.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8). It ingests untrusted data from Unreal Engine project files (DefaultEngine.ini, level structures) and project metadata. It lacks explicit boundary markers or sanitization logic to separate this data from agent instructions. Given its powerful capabilities—including file modification (unreal_generate_cpp), compilation, and Python execution—this ingestion point creates an exploitable attack surface where malicious data could influence agent actions.

unreal-development