schedule-host-job

name	schedule-host-job
description	Schedule one-off or recurring local-time commands from always-on-claude container coding sessions through the host-native v2 scheduler. Use when the user asks Codex to run something later, overnight, at a specific local time, after hours, or on an hourly/daily/weekly/monthly cadence from a container session without direct host scheduler access.

Schedule Host Job

Use the always-on-claude v2 scheduler to submit commands from the container to the host. The host validates requests, installs native scheduler entries, and later runs the command inside the claude-dev container with docker compose exec.

Commands

Use the repo-managed CLI:

/home/dev/dev-env/scripts/runtime/aoc-schedule.sh

Common operations:

/home/dev/dev-env/scripts/runtime/aoc-schedule.sh at "2026-04-26 09:00" -- "npm test"
/home/dev/dev-env/scripts/runtime/aoc-schedule.sh daily --time 03:00 -- ./scripts/nightly.sh
/home/dev/dev-env/scripts/runtime/aoc-schedule.sh hourly --minute 0 -- ./scripts/hourly.sh
/home/dev/dev-env/scripts/runtime/aoc-schedule.sh weekly --weekday friday --time 21:00 -- ./scripts/report.sh
/home/dev/dev-env/scripts/runtime/aoc-schedule.sh list
/home/dev/dev-env/scripts/runtime/aoc-schedule.sh status <job-id>
/home/dev/dev-env/scripts/runtime/aoc-schedule.sh logs <job-id>
/home/dev/dev-env/scripts/runtime/aoc-schedule.sh delete <job-id>

Workflow

1. Use the current working directory when it is under /home/dev/projects.
2. If the current directory is not under /home/dev/projects, pass --cwd /home/dev/projects/<repo> or ask which repo should own the job.
3. Submit one-off jobs with at "YYYY-MM-DD HH:MM" -- <command>, or recurring jobs with hourly, daily, weekly, or monthly.
4. Report the returned job id plus the exact status and logs commands.
5. Use list, status, and logs instead of inspecting host scheduler state directly.

Safety

• Do not mount the Docker socket or SSH back to the host to schedule jobs.
• Do not run host launchctl, systemctl, crontab, or Docker directly from a container coding session for this workflow.
• Confirm before scheduling destructive, externally visible, cloud-costing, or production-impacting commands.
• If the CLI says the v2 inbox is missing, tell the user to run /update in the manager session and restart the container so the schedule mounts are available.