| name | longrun-orchestrator |
| description | Launch and monitor long-running commands with resumable status + logs under WORKSPACE/.codex/longrun, and optionally chain follow-up steps after completion. Use when a job will likely exceed a reasonable single-session window, or when you want the agent to keep progressing without manual reminders. |
Long-Run Orchestrator
Quick start (detach + monitor)
python3 ~/.codex/skills/longrun-orchestrator/scripts/orchestrate.py \\
--label my_run \\
--detach \\
--cmd 'sleep 10' \\
--then 'echo done'
If you are not inside a git repo, pass --workspace /path/to/workspace to avoid writing under $HOME.
Check status:
python3 ~/.codex/skills/longrun-orchestrator/scripts/status.py --latest
Contract (how to use in execution persistence)
- If the job is likely short, prefer foreground polling (
--detach not used).
- If the job is likely long, use
--detach so work continues after the agent sends an Interim Update.
- Store everything in the active workspace only:
WORKSPACE/.codex/longrun/<label>_<timestamp>/
Reporting guidance (Interim Update)
When using this skill and pausing execution, report minimally:
Pause reason: waiting- What is running (label + command)
Resume trigger: the orchestrator status becomes completed (auto-continues if follow-ups are configured)
Only include exact paths when asked or when debugging requires it.
