The Agent Skills Directory

[SAFE]: The skill facilitates CenterPose model workflows (training, evaluation, export) using the NVIDIA TAO Toolkit. All documented behaviors are consistent with its stated purpose.\n- [SAFE]: External dependencies are restricted to official NVIDIA container images hosted on nvcr.io, which is a trusted vendor registry. The skill does not attempt to download or execute code from unknown or untrusted remote sources.\n- [SAFE]: The skill utilizes the Bash tool to execute docker and centerpose commands. This is standard and required functionality for the TAO Toolkit environment.\n- [SAFE]: No hardcoded credentials, sensitive file access, or data exfiltration patterns were detected. The encryption_key parameters in the templates use non-sensitive placeholders like tlt_encode.\n- [SAFE]: No obfuscation techniques, such as Base64 encoding of commands or hidden Unicode characters, were found in the skill's instructions or metadata.

tao-train-centerpose