mobility-gen
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external simulation files which could serve as a surface for indirect instructions if the processing engine evaluates metadata from these sources.
- Ingestion points: Scripts such as
record_trajectories.pyandcustom_footprint_robot.pyprocess external environment files (scene_usd) and occupancy data (omap_yaml). - Boundary markers: The skill does not employ specific boundary markers for these binary and structured data formats.
- Capability inventory: The skill has capabilities to write to the file system (e.g.,
save_stage,np.savez) and control simulation execution throughSimulationManagerandSimulationApp. - Sanitization: No explicit sanitization or validation of internal simulation file metadata is performed during ingestion.
Audit Metadata