How do I define which commands are blocked?

You define your security rules in a `.agentpolicy` file within your repository, specifying which commands are strictly allowed, denied, or require manual approval.

Does it record the actions taken by the AI?

Yes, Command Guard records every decision to allow, block, or request confirmation in an Action Log for full transparency and auditing.

Can I use it to ignore specific files or directories?

Yes, the skill fully respects `.agentignore` files, ensuring that the AI agent does not interact with sensitive paths defined by the developer.

Why is this skill important for AI agents?

It adds a necessary layer of human-in-the-loop protection, ensuring the AI doesn't run destructive commands like 'rm -rf' or unauthorized scripts without explicit consent.

What is Command Guard?

Command Guard is a safety-focused skill for Claude Code that checks every terminal command against a policy file before execution to prevent risky or destructive actions.

Command Guard

Name: Command Guard
Author: CheekyCodexConjurer

byCheekyCodexConjurer

0•

セキュリティとテスト

Enforces security policies and safety guardrails by validating terminal commands against predefined rules before execution.

Command Guard provides a critical safety layer for AI agents by intercepting and evaluating CLI commands against local .agentpolicy and .agentignore files. It prevents accidental execution of destructive actions, enforces organizational security standards, and maintains a comprehensive action log for transparency. This skill is essential for developers who want to grant Claude autonomy while ensuring it operates within strictly defined safety boundaries, requiring explicit approval for guarded commands while blocking prohibited ones automatically.

主な機能

01Pre-execution command validation

02Comprehensive action logging and auditing

03Automated .agentpolicy enforcement

04Interactive confirmation for high-risk actions

05Seamless integration with .agentignore rules

060 GitHub stars

ユースケース

01Auditing AI-driven terminal activity for security compliance and post-mortem analysis

02Preventing accidental deletion of production databases or critical system files

03Standardizing allowed terminal commands across a distributed development team

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add cheekycodexconjurer/aera-chart-engine command_guard

For use in Claude.ai and ChatGPT

Download Skill