mujoco-playground
Pass
Audited by Gen Agent Trust Hub on Nov 19, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions include cloning the mujoco_playground repository from the official GitHub organization to set up the workspace.
- [COMMAND_EXECUTION]: The skill uses shell commands for environment setup, including repository cloning and package installation via uv pip for standard robotics libraries such as ruckig, scipy, and mujoco.
- [INDIRECT_PROMPT_INJECTION]: The skill architecture processes scene configurations and sensor data during operation. 1. Ingestion points: The setup_scene callback function, camera RGB streams, and robot joint measurements. 2. Boundary markers: Handled programmatically via callback signatures. 3. Capability inventory: MuJoCo model compilation, physics simulation, and physical robot device communication. 4. Sanitization: Joint commands are validated and clipped against the robot's physical limits defined in the model specification.
Audit Metadata