isaaclab-preparing-assets-for-newton
Pass
Audited by Gen Agent Trust Hub on Aug 13, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill instructs the agent to process and convert external asset files (URDF, MJCF, USD), which constitutes an indirect prompt injection surface.
- Ingestion points: External asset files such as URDF, MJCF, and USD payloads mentioned in
SKILL.mdandreference.md. - Boundary markers: The instructions do not specify any delimiters or safety warnings to ignore instructions embedded in the metadata or comments of these external assets.
- Capability inventory: The skill provides commands in
reference.mdto execute Python scripts (scripts/environments/zero_agent.py) using theuvpackage manager, which interact with the simulation environment. - Sanitization: There is no mention of sanitizing or validating the contents of the imported asset files before processing them.
- [COMMAND_EXECUTION]: The skill provides explicit shell commands in
reference.mdfor validating asset migration and running smoke tests. - Evidence:
uv run --extra isaacsim python scripts/environments/zero_agent.py --task TASK --num_envs 4 --viz none physics=physx - Context: These commands are used for legitimate testing and validation of the migrated assets within the Isaac Lab environment.
Audit Metadata