monitor-training
Fail
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: HIGHPERSISTENCECOMMAND_EXECUTION
Full Analysis
- [PERSISTENCE]: The skill utilizes the
CronCreatetool to schedule recurring agent tasks. It describes a 'self-deleting and self-creating' workflow where cron jobs are chained to maintain an autonomous pipeline over long durations without human intervention. - [COMMAND_EXECUTION]: The skill instructs the agent to execute system monitoring commands such as
nvidia-smito query GPU processes and standard shell utilities liketailto monitor training logs.
Recommendations
- AI detected serious security threats
Audit Metadata