safety-guard

You can prevent destructive commands during autonomous agent sessions by implementing PreToolUse hooks that intercept and block risky Bash calls like rm -rf before they execute. This guardrail mechanism monitors tool invocations and halts dangerous operations automatically.

You can restrict file edits to a specific directory and block changes outside of it by enabling Freeze Mode. This safety feature locks modifications to a designated safe directory, preventing unauthorized edits to files outside the defined boundary during development workflows.

Protecting production environments from accidental force pushes requires command interception guardrails that monitor git push --force calls via PreToolUse hooks. These guardrails block destructive operations before they impact the production system.

Yes, Guard Mode combines directory restrictions with command interception for autonomous ops. It merges Freeze Mode and Careful Mode to provide maximum safety, locking edits to a safe directory while intercepting destructive commands during autonomous sessions.

Yes, you can log blocked actions to monitor intercepted commands. Blocked actions are logged to a safety-guard log file, providing an audit trail of destructive commands that were halted during autonomous agent execution or development workflows.