Specialist: Safety Engineer
Trigger: /careful comment on any issue
Browser required: No
Default: ✅ Enabled
/careful activates safety guardrails that warn before destructive commands are executed. When active, the agent will pause and ask for confirmation before running operations like:
rm -rf— recursive file deletionDROP TABLE/DROP DATABASE— database destructiongit push --force/git reset --hard— history rewritingkubectl delete— Kubernetes resource removal- Other destructive shell operations
Common build cleanup commands (like rm -rf node_modules or rm -rf dist) are whitelisted and won't trigger warnings.
- Working on or near production systems
- Debugging live deployments
- Operating in shared environments
- Any time you want an extra safety net before destructive operations
Comment on any issue:
/careful
You can also say "be careful", "safety mode", "prod mode", or "careful mode" in your issue or comment.
/careful
I need to clean up the staging database — drop the old migration tables and redeploy.
The agent will proceed with your request but pause before any destructive command, showing you exactly what it plans to run and waiting for your approval.
| Command | Relationship |
|---|---|
/guard |
Combines /careful with /freeze for maximum safety |
/ship |
Use /careful when shipping to production |
/land-and-deploy |
Safety guardrails are especially useful during deploys |
{
"careful": { "enabled": true, "trigger": "issue_comment" }
}
