fr3-duo-control
Pass
Audited by Gen Agent Trust Hub on Nov 19, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill automates the retrieval of high-fidelity robot models and assets from the official Google DeepMind GitHub repository (
google-deepmind/mujoco_menagerie) to ensure simulation accuracy. - [COMMAND_EXECUTION]: Implements
subprocess.runwithin its model-building scripts to perform sparse git clones and maintenance of required robot configuration assets. - [DYNAMIC_EXECUTION]: Utilizes
importlib.import_moduleinscripts/bringup.pyto verify the local Python environment and ensure that necessary dependencies likepylibfrankaandmujocoare installed before proceeding with hardware control. - [INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface through
config/robot_config.jsonand interactive measurement prompts inscripts/bringup.py. Boundary markers for these inputs are absent in the internal logic. The capability inventory includes subprocess execution for asset cloning, file system modification for logging and configuration, and low-level hardware actuation. The skill performs numeric validation and physical constraint checks on measurements to sanitize ingested measurement data.
Audit Metadata