tao-train-dinov3
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill follows secure practices by using official, vendor-owned container images from a well-known registry (nvcr.io). Documentation specifically warns against hardcoding secrets and suggests using least-privilege credentials.
- [SAFE]: No obfuscation or malicious redirection was found. The large Base64-encoded content in the signature file (
skill.oms.sig) is a standard in-toto attestation for verifying the integrity of the skill's artifacts. - [SAFE]: The dynamic context and execution patterns observed (such as configuration variable interpolation) are specific to the TAO Toolkit's legitimate operation and do not allow for arbitrary command injection.
Audit Metadata