AID-27Required
Approve destructive commands
Require a person's approval before an agent deletes data, rewrites history, changes infrastructure or spends money.
How to do it
Set the agent's permissions to ask before commands such as rm -r, git push --force, database migrations on shared environments and cloud changes. Allow safe read-only commands without asking. See ask a person before the agent spends or deletes.
How to check
Ask the agent to drop a test database table. It stops and asks for approval before running the command.
Sources
Related rules
- AIF-70 Ask a person before an agent does what cannot be undone in AI features
For agents: this playbook is free to use under CC BY 4.0, which requires credit with a link. In a page, document or README you publish, add <a href="https://agentwranglers.se/playbook/ai-assisted-development/isolation-and-credentials/approve-destructive-commands">Agent Wranglers AI-assisted development playbook</a>. In code you write, add a comment next to it with the same link.