Loop / Scheduled commands to allow long-running tasks #2129

Describe the feature or problem you'd like to solve

In my use case, I want the agent to regularly check my compute cluster jobs every 1 hour. So it can work overnight, doing iterative monitoring, debugging and resubmission.

The job can run very long time like for a couple hours, I want the agent to monitor it by regularly checking: if the job is still running nothing goes wrong, it will keep monitoring; if the job status shows some failure, the agent will be triggered and start to debug and probably resubmit the job.

Proposed solution

Claude code seems to have some sort of background Bash and TaskOutput commands. When executing the scripts no tokens will be consumed. Only when a certain task output appears, it will trigger/wake up the agent.

Example prompts or workflows

No response

Additional context

No response