tao-train-segformer
Pass
Audited by Gen Agent Trust Hub on Jun 18, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides structured documentation and configuration schemas for model development workflows. No malicious instructions, obfuscated payloads, or persistence mechanisms were detected.
- [EXTERNAL_DOWNLOADS]: The skill references Docker images from the official NVIDIA container registry (nvcr.io). These are well-known technology service resources used for executing TAO Toolkit actions and do not represent a security risk.
- [COMMAND_EXECUTION]: The documentation outlines the use of standard TAO Toolkit CLI commands such as 'segformer train' and 'segformer evaluate' within Docker environments. These operations are consistent with the skill's primary purpose.
- [CREDENTIALS_UNSAFE]: The provided templates include an 'encryption_key' parameter. The default value 'tlt_encode' is a publicly known default for the TAO Toolkit and does not constitute a sensitive credential or secret leakage.
Audit Metadata