scenario-godot-3d-world
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The Python runner script (
scripts/gd_world.py) usessubprocess.runto execute the systemcpcommand for cloning files during asset ingestion. This is used with a fixed argument list (no shell injection risk) to perform APFS clones of 3D models. - [INDIRECT_PROMPT_INJECTION]: The skill processes externally sourced 3D assets (GLB/glTF files). While these files are parsed using standard libraries (
json,struct), they represent an ingestion point for untrusted data. However, the skill only extracts geometric metadata and applies pre-defined manifests, posing minimal risk. - [SAFE]: The skill relies on a sibling skill,
scenario-godot-expert, which is a standard pattern for modular AI agent extensions. All file operations are restricted to the local project directory and standard Godot resource paths.
Audit Metadata