The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the mcp__ue58-mcp__execute_python_code tool to run Python scripts directly within a running Unreal Engine environment. This is the intended primary purpose of the skill to automate 3D content generation.
[DATA_EXPOSURE_AND_EXFILTRATION]: Instructions include reading local Unreal Engine log files (e.g., <project>/Saved/Logs/<project>.log) using standard Python open() and glob modules to verify execution status. This access is restricted to the local project workspace and does not involve exfiltration to external domains.
[METADATA_POISONING]: The SKILL.md file recommends using a tool named mcp__desktop-commander__read_file for debugging, although this tool is not included in the allowed-tools configuration. This appears to be a documentation inconsistency rather than a malicious attempt to bypass permissions.
[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for analyzing existing Unreal Engine assets and production graphs. While this involves processing external data (PCG assets), the execution is controlled by the agent following established developer templates, and no specific injection surfaces were identified.

unreal-pcg-automation