lingbot-map-3d-reconstruction
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads software and model checkpoints from external sources, including GitHub (Robbyant/lingbot-map), HuggingFace (robbyant/lingbot-map, JianyuanWang/skyseg), and project-specific package indices (PyTorch and FlashInfer). These sources are well-known services and the downloads are necessary for the skill's functionality.
- [COMMAND_EXECUTION]: The documentation includes examples of using
subprocess.runto programmatically execute the skill's owndemo.pyscript. This is a standard integration pattern for providing a Python API on top of CLI tools. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data in the form of images and video files for 3D reconstruction.
- Ingestion points: Images and videos loaded via
--image_folderor--video_pathinSKILL.md. - Boundary markers: None specific to the LLM interaction as the data is processed by a specialized computer vision model.
- Capability inventory: File system writes (for masks/viz) and subprocess calls are present in
SKILL.md. - Sanitization: The skill uses standard libraries (OpenCV, Pillow) to decode media, which provides inherent validation of file formats.
Audit Metadata