nemo-mbridge-perf-sequence-packing
Pass
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides legitimate technical documentation and configuration snippets for deep learning training features within the Megatron-Bridge framework. Analysis found no evidence of prompt injection, data exfiltration, or persistence mechanisms.\n- [COMMAND_EXECUTION]: The skill includes documentation for running unit tests using the
uvtool to verify the training environment. These commands are standard for the developer use case and do not involve unauthorized privilege escalation or remote script execution.
Audit Metadata