The Agent Skills Directory

[SAFE]: No malicious patterns or security risks were detected in the skill instructions, schemas, or reference documents.- [COMMAND_EXECUTION]: The skill documents standard TAO Toolkit CLI commands (e.g., ocrnet train, ocrnet evaluate) and docker run configurations. These are the expected methods for interacting with the TAO environment.- [REMOTE_CODE_EXECUTION]: External container images are referenced from nvcr.io (NVIDIA Container Registry). These images are official vendor resources required for the skill's primary purpose.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external datasets containing scene images and text labels. While this creates a data ingestion surface, the risk is negligible given the specialized nature of the OCR task and the absence of instructions to interpret data content as agent directives.

tao-train-ocrnet