Pairs with: autonomous-loops, agentic-engineering, deployment-patterns

Safety Guard

When to Activate

• Agent running in autonomous/full-auto mode
• Working on production systems with irreversible operations
• Restricting edits to a specific directory during focused work
• Sensitive operations: migrations, deploys, permission changes, data transformations

Three Protection Modes

Mode 1: Careful Mode

Intercepts destructive commands before execution and warns, then asks for confirmation: