tao-generate-referring-expressions
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied images and KITTI labels through VLMs, which presents an attack surface for indirect prompt injection. * Ingestion points: data.image_dir, data.kitti_label_dir, and data.input_annotations_jsonl. * Boundary markers: None mentioned in the instructions to separate data from instructions. * Capability inventory: The skill uses the Bash tool to run the auto_label CLI and the Write tool to output results. * Sanitization: Documentation does not mention sanitization of input data before processing.
- [EXTERNAL_DOWNLOADS]: The skill downloads official container images from NVIDIA's registry (nvcr.io) and provides instructions for installing the vllm package from official sources.
- [COMMAND_EXECUTION]: The skill uses the Bash tool to execute the auto_label command and server setup commands.
Audit Metadata