guard

Block destructive commands and restrict edits to an approved directory.

1|Updated Mar 27, 2026

Search Tags:#guardrails #production debugging #safety mode #destructive command protection #directory freeze #filesystem write restrictions

Ready to run

npx skills add https://github.com/yibeichan/schist --skill guard-yibeichan

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: guard
Source: https://github.com/yibeichan/schist/tree/main/.claude/skills/guard
Command: npx skills add https://github.com/yibeichan/schist --skill guard-yibeichan

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Prevents accidental destructive operations and unwanted edits outside a controlled area by combining pre-execution safety checks with a directory-scoped write boundary.

Core Features & Use Cases

Destructive command warnings: Warns before running high-risk operations such as rm -rf, DROP TABLE, and force-push style actions.
Freeze boundary enforcement: Blocks edit/write actions that target files outside the user-selected directory.
Use case: Use while debugging or handling production-adjacent tasks where you want tight guardrails but still need the agent to operate normally inside the approved path.

Quick Start

Tell the agent to enable guard mode by asking it which directory should be the only allowed edit target.

Frequently Asked Questions about guard

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I prevent destructive commands during live production debugging?▼

To prevent destructive commands during live production debugging, use a safety mode that blocks high-risk operations like rm -rf or DROP TABLE before execution. This reduces operational risk by preventing unwanted edits outside a controlled directory area.

How do I restrict filesystem write permissions to a specific directory?▼

You restrict filesystem write permissions by establishing a freeze boundary that blocks edit and write actions targeting files outside an approved directory. The system requires you to resolve this boundary to an absolute, trailing-slash normalized path.

Can I still operate normally inside an approved directory with guardrails enabled?▼

Yes, you can operate normally inside an approved directory with guardrails enabled. The freeze boundary only blocks modifications outside the user-selected path, allowing the agent to function freely within the controlled area.

What destructive operations are blocked by safety mode?▼

Safety mode blocks destructive commands including rm -rf, DROP TABLE, and force-push style actions. It applies pre-execution checks via hook scripts to intercept these high-risk operations before they run.

How do I enable directory freeze protection for production-adjacent changes?▼

To enable directory freeze protection for production-adjacent changes, tell the agent to enable guard mode and specify which directory should be the only allowed edit target. This activates pre-tool checks for Bash commands and Edit actions.