nemo-mbridge-perf-cpu-offloading
Pass
Audited by Gen Agent Trust Hub on Jun 18, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill's content is focused on legitimate machine learning optimization techniques. No prompt injection, data exfiltration, obfuscation, or persistence mechanisms were found.\n- [COMMAND_EXECUTION]: The skill contains example shell commands for running training recipes and unit tests using the
uvtool. These commands are standard for development workflows in the Megatron Bridge environment and do not pose a security risk.\n- [EXTERNAL_DOWNLOADS]: No remote downloads or external script executions were detected. The skill uses local project paths and standard tooling.
Audit Metadata