Codex Fleet Update

Use this skill to keep @openai/codex aligned across a small SSH fleet without background polling.

Core Pattern

Prefer an on-demand reconciler:

1. Query npm once for the target @openai/codex version.
2. SSH to each host.
3. Run checks inside bash -ic, because many devices initialize nvm/npm/Codex only from interactive shell startup.
4. Install only when the remote version differs.
5. Verify codex --version after installation.
6. Log per-host results.

Do not add a timer unless the user explicitly wants background polling. Codex releases are not frequent enough to justify regular checks for this use case; the operator usually notices an update and then wants to update the rest of the devices.

Installed Command