scope-guard

To prevent destructive commands during code migrations, use a guardrail mode that combines path freezing with risky command warnings. This requires explicit approval before executing any high-impact operations, ensuring safe migrations.

To isolate work when running parallel agents on a codebase, re-brief each delegated agent with a specific path scope. This explicit path ownership prevents overlapping edits and maintains strict boundaries between concurrent tasks.

The best way to add approval checkpoints before risky code operations is to enable plan-mode checkpoints. This inserts an explicit approval step before any high-impact operation, halting execution until authorized.

Yes, you can clear file scope restrictions after a risky refactoring task is complete through explicit removal. This removes the defined writable surface and mode-based guardrails, restoring standard agent permissions.

Yes, plan-mode works without relying on external scripts for guardrails. It enforces mode-based restrictions and explicit path ownership directly through the skill's internal logic, satisfying safety requirements natively.