skills.rest
ibytechaos
ibytechaosCOMMUNITY
Explore all skills by this author →
GitHub

safety-guard

Intercepts destructive shell commands and restricts file edits during autonomous agent sessions.

1|Updated Oct 11, 2025
Search Tags:#production#hooks#autonomous agents#safety#guardrails#file protection#destructive commands
One-click install
npx skills add https://github.com/ibytechaos/claude --skill safety-guard-ibytechaos
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: safety-guard
Source: https://github.com/ibytechaos/claude/tree/main/plugins/everything-claude-code/skills/safety-guard
Command: npx skills add https://github.com/ibytechaos/claude --skill safety-guard-ibytechaos

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Autonomous agents and production work carry the risk of destructive operations like rm -rf, force pushes, or dropped tables. This Skill intercepts dangerous commands before execution and confines file edits to approved directories. ## Core Features & Use Cases - Careful Mode: Detects destructive patterns (rm -rf, git push --force, DROP TABLE, kubectl delete, chmod 777) and asks for confirmation with safer alternatives. - Freeze Mode: Locks Write/Edit operations to a specific directory tree so agents cannot touch unrelated code. - Guard Mode: Combines both protections for maximum safety during full-auto agent runs, with all blocked actions logged to ~/.claude/safety-guard.log. - Use Case: Before letting an agent run autonomously on a production repository, activate Guard Mode scoped to src/api/ so the agent can read everything but only write to that directory while destructive commands are blocked everywhere. ## Quick Start Enable safety guard in guard mode restricted to the src/api directory before running this agent autonomously.

Frequently Asked Questions about safety-guard

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

FAQPage Schema
How do I prevent an AI agent from running destructive commands?▼

Activate Careful Mode, which intercepts dangerous patterns like rm -rf, git push --force, and DROP TABLE before execution. It shows what the command does, asks for confirmation, and suggests a safer alternative.

How to restrict Claude Code edits to a specific directory?▼

Use Freeze Mode with a command like /safety-guard freeze src/components/. Any Write or Edit outside that directory tree is blocked with an explanation, keeping the agent focused on one area.

What commands does safety-guard block?▼

It watches for rm -rf on sensitive paths, git push --force, git reset --hard, git checkout ., DROP TABLE/DATABASE, docker system prune, kubectl delete, chmod 777, sudo rm, npm publish, and any command using --no-verify.

Does safety-guard work with autonomous agent sessions?▼

Yes, Guard Mode is designed for full-auto sessions, combining command interception with directory write restrictions. It is recommended for codex -a never sessions and logs all blocked actions to ~/.claude/safety-guard.log.

How do I turn off safety-guard protections?▼

Run /safety-guard off to unlock all active modes. This removes both the destructive command interception and any directory-scoped write restrictions.

Related Skills

Explore More Software Engineering Skills→
joonlabjoonlab
8

hook-creator

Create and configure Claude Code lifecycle hooks for event-driven shell commands.

Community
Advanced
llmh333llmh333

safety-guard

Block destructive shell commands and restrict file edits to a specified directory.

Community
Intermediate
AlexiosBluffMaraAlexiosBluffMara
2

requesting-code-review

Automate pre-commit verification with static security scans and quality gates.

Community
Advanced
whatswithaviwhatswithavi

hosted-agents

Build sandboxed remote infrastructure for background coding agents with warm pools and multiplayer sessions.

Community
Advanced