test-mode-tool - Defensive security tool that prevents Claude Code

项目说明

Test Mode Tool for Claude Code

A defensi system that prevents Claude Code from making destructi during test analysis, ensuring test failures remain as diagnostic information.

![License: MIT]()

🎯 Problem Solved

The "Never Give Up" Problem: When Claude Code encounters test failures, it tends to:

Our Solution: Test Mode Tool uses automation-first commands that create hard boundaries, channeling AI persistence into producti rather than destructi.

🛡️ What Test Mode Does

When Active:

What's Blocked:

🚀 Quick Start

1. Install

2. Use Test Mode

3. Deploy to Other Projects

🏗️ Architecture

Three-Layer Protection System

1. Custom Slash Commands - /test_mode:on|off|status|clean
- Automation-First Design: All operations execute test_mode_setup.sh scripts
- Zero Manual Steps: Commands perform atomic operations automation
- User-friendly interface with automatic project isolation
- Comprehensi clean command

2. Security Hooks System - Comprehensive file modification blocking
- Automated Setup: test_mode_setup.sh configures all hooks
- PreToolUse hooks block destructi
- PostToolUse hooks log and usage
- Project context prevents interference

3. Specialized Agents - Read-only test analysis
- test-mode-observer: Analyzes tests without modifications
- test-reporter: Generates comprehensi

How Commands Work

Each command executes automation scripts directly for consistent behavior:

This automation-first approach ensures:

Project Isolation

📊 Expected Benefits

🔧 Installation Options

Project-Level (Team Deployment)

User-Level (Personal Workflow)

Other Projects

🛡️ Security Features

Multi-Layer Security Model

Threat Protection

📚 Documentation

🔍 Usage Examples

Basic Workflow

Ad Usage

🤝 Contributing

We welcome contributions to improve the Test Mode Tool! Please:

1. Security First: All contributions must maintain security standards
2. Defensi: This tool prevents destructi, not enables it
3. Project Isolation: Ensure no cross-project interference in changes
4. Comprehensi: security and functionality

📄 License

MIT License - see LICENSE file for details.

🆘 Support

🏆 Recognition

This tool addresses the fundamental "never give up" problem identified by the experiments in AI-assisted de, where AI persistence leads to destructi instead of proper issue analysis.

Inspiration & Discovery

This solution was inspired by the groundbreaking analysis from that discovered Claude Code's problematic "never give up" behavior during testing:

The experiments re that Claude Code consistently:

Our Solution

By creating hard boundaries through this Test Mode Tool, we channel AI capabilities into producti rather than harmful modification. The tool enforces the principle that test failures are diagnostic information, not problems to hide.

Credit: Special thanks to the team for identifying this critical issue in AI-assisted de and inspiring defensi like this Test Mode Tool.

项目信息

开发语言：Shell

Star 数：4

技术要点

该项目属于防护思路方向的开源资源，代码结构清晰，适合安全研究人员学习参考。项目围绕Defensive security tool that prevents Claude Code from making destructive modifications during test analysis展开，提供了可运行的实现，对理解相关安全机制、工具原理与攻防技术具有较高的研究价值。

从技术实现角度看，此类项目的核心在于对底层原理的深入理解与工程化落地。开发者在研究过程中可以结合官方文档与社区讨论，梳理项目的初始化流程、核心模块职责与数据流转，逐步掌握其中的技术细节与设计思路。

使用与学习建议

1. 下载源码压缩包，解压到本地环境。

2. 根据项目类型准备对应的运行环境与依赖。

3. 按照 README 中的步骤完成配置与部署。

4. 在隔离的测试环境中运行验证，避免影响生产系统。

免责声明

本资源来源于开源社区，仅供安全研究、技术学习与合法授权范围内的攻防演练使用，请勿用于任何非法或侵害他人权益的用途。使用本项目产生的一切后果由使用者自行承担。