build-minecraft-model-from-image
Warn
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The
build_model.pyscript usesimportlibto dynamically load and execute Python scripts (object.py) generated by the AI agent during the workflow. This allows for the execution of arbitrary logic defined at runtime by the agent. - [COMMAND_EXECUTION]: The system executes several external command-line tools via
subprocess.run, includingBlenderfor rendering 3D views,ffmpegfor video processing, andnvidia-smifor GPU resource monitoring. - [EXTERNAL_DOWNLOADS]: The skill's scripts utilize the
uvtool to manage and download Python dependencies (numpy,pillow,opencv-python-headless) declared inline within the script files. These are well-known libraries from standard registries. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied images to inform the generation of executable Python code, creating a surface for indirect injection if a malicious image is designed to mislead the agent's code generation. \n
- Ingestion points:
scripts/read_image.pyprocesses thesource.jpginput file. \n - Boundary markers: Absent; there are no explicit delimiters or instructions to ignore embedded content in the image data. \n
- Capability inventory: Includes arbitrary Python code execution (
importlib), system command execution (subprocess.run), and file system write access. \n - Sanitization: Absent; the image data is processed into blueprints and palettes without sanitizing the resulting instructions passed to the code generation phase.
Audit Metadata