scenario-unreal-expert
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The toolkit executes shell commands to manage Unreal Engine processes (
UnrealEditor,UnrealEditor-Cmd,RunUAT.sh) and system utilities (ps,lsof,sw_vers,sysctl,xcodebuild). These operations are documented in the skill's protocol and are necessary for engine automation and environment preflight checks. - [DYNAMIC_EXECUTION]: The toolkit utilizes the
runpymodule to execute locally staged Python scripts and uses string concatenation to build Python code for remote execution within the Unreal Editor process. These mechanisms are central to the skill's purpose of driving engine-side automation and are used with appropriate boundary markers. - [INDIRECT_PROMPT_INJECTION]: The skill presents an indirect prompt injection surface due to its data processing capabilities.
- Ingestion points: It reads and parses external data from Unreal Engine logs (
ue_run.py,ue_stat.py), performance trace files (ue_stat.py), and image files (ue_review.py). - Boundary markers: It implements a structured JSON envelope labeled
UE_RESULTto specifically isolate and verify job results from the standard console output. - Capability inventory: The agent can execute system commands, manage processes, and run Python code within the Unreal Engine environment.
- Sanitization: The toolkit employs JSON validation, regular expression filtering, and specialized parsers (CSV, JSONL, PNG) to extract facts from engine-generated data.
Audit Metadata