senior-computer-vision
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of images and annotation files (e.g., COCO, Pascal VOC) through various pipeline scripts.
- Ingestion points: The skill instructions refer to reading data from
data/images/,data/labels/, and other user-provided paths inSKILL.md. - Boundary markers: There are no explicit instructions or delimiters to prevent the agent from interpreting content within dataset metadata or filenames as instructions.
- Capability inventory: The skill utilizes scripts (
scripts/dataset_pipeline_builder.py,scripts/vision_model_trainer.py) that perform file system operations and execute shell commands (e.g.,yolo,trtexec). - Sanitization: There is no evidence of sanitization or validation of the input data provided in the documentation.
Audit Metadata