| name | remote-ops-control |
| description | Operate remote machines and services through existing remote-access tools such as SSH, cloud CLIs, admin APIs, deployment hooks, or approved scripts. Use when Codex needs to run staged diagnostics, approval-gated remote actions, restarts, redeploys, rollback steps, and operator-style incident handling without pretending it has built-in remote control. |
Remote Ops Control
Use this skill when the user wants Codex to act like a disciplined remote operator through tools that already exist in the environment.
This skill does not create remote access. It assumes a real control path already exists:
- SSH
- deployment or cloud CLI
- admin API
- remote script runner
- companion service
Best Fit
This skill is strongest for:
- remote diagnostics
- staged restart or redeploy workflows
- rollback execution
- service health verification
- incident-time operator actions
- turning "fix prod" into an explicit, reviewable plan
Start Here
Collect these facts first:
- target machine, service, or environment
- remote control path available
- who approves impactful actions
- what counts as safe diagnostics versus a mutating action
- what rollback path exists
- what evidence should be captured before changing anything
If the user asks for a broad fix, narrow it immediately into:
- diagnostic goal
- command or action candidates
- approval gates
- rollback plan
- verification steps
Workflow
1. Prove the target and control path
Confirm:
- which host, cluster, or service is in scope
- which access method is intended
- whether the current session can actually use it
Do not assume "remote control" exists because the user wants it.
2. Separate diagnostics from mutations
Start with read-only evidence gathering:
- status
- logs
- process health
- recent errors
- deployment version
- queue depth
- disk, CPU, memory, or network symptoms
Only after that, propose mutating actions:
- restart
- redeploy
- config change
- scale change
- rollback
- cache clear
3. Stage the remote action
For any mutating action, specify:
- exact target
- expected effect
- approval requirement
- rollback path
- success check
4. Report like an operator
Return:
- what was observed
- what action is proposed or taken
- why it is the next step
- what risk remains
- how success will be verified
Default Approval Posture
Unless the user gives a looser but credible control model:
- allow read-only diagnostics
- require approval for restart, redeploy, rollback, scale, config, or destructive actions
- block actions that have no clear target, no rollback path, or no verification plan
Read references/approval-model.md when the user needs a practical approval boundary for remote work.
Common Output Patterns
Diagnostic plan
Return:
- target
- suspected issue
- read-only checks
- likely next actions
- required approvals
Restart or redeploy plan
Return:
- target
- reason
- exact action
- rollback path
- post-action verification
Incident operator note
Return:
- current symptom
- evidence gathered
- approved next step
- owner
- next checkpoint
Reference Use
Load references selectively:
references/approval-model.md for safe default approval gates
references/remote-workflows.md for restart, redeploy, and rollback patterns
references/verification-patterns.md for post-action checks and evidence capture
Example Requests
- "Use $remote-ops-control to diagnose this unhealthy service over SSH and tell me the safest next action."
- "Use $remote-ops-control to prepare a restart plan with rollback and verification."
- "Use $remote-ops-control to turn this prod issue into an operator-style runbook for Codex."