PromptInjector - A comprehensive defensive security testing tool fo

项目说明

PromptInjector 🔒

A comprehensive model-agnostic defensi testing tool for AI systems. PromptInjector helps identify prompt injection through systematic testing with both static and adapti. Now supports any AI model or API endpoint and includes MCP ser for seamless agent collaboration.

⚠️ Important Disclaimer

This tool is designed for defensi purposes only. It should be used to:

Do not use this tool to attack systems you don't own or don't ha to test.

🚀 New Features

✨ Model-Agnostic Design

🤖 MCP Ser

🔧 Enhanced Configuration

🛠️ Supported Model Types

| Type | Description | Example Configuration |
|------|-------------|----------------------|
| OpenAI | OpenAI API compatible endpoints | GPT-3.5, GPT-4, custom deployments |
| Anthropic | Claude models API | Claude-3, Claude-2 |
| Ollama | Local models | Llama-2, CodeLlama, Mistral |
| HTTP | Generic HTTP/REST APIs | Any custom AI API endpoint |

📦 Installation

1. Clone the repository:

2. Install dependencies:

3. Create configuration file:

⚙️ Configuration

Modern Configuration Format

Create prompt_injector_config.json:

Example Configurations

#### Local Ollama Setup

#### Mixed En Setup

#### Anthropic Claude Setup

En

Set these en for quick configuration:

🎯 Usage

Standard Testing Modes

#### Quick Test (15 prompts)

#### Full Test (150 prompts)

#### Custom Test Configuration

#### Test Custom Prompts

MCP Server Mode

The MCP (Model Context Protocol) ser external AI agents to control and analyze prompt injection testing dynamically. The external agent acts as the analyzer, allowing for sophisticated multi-agent security testing workflows.

#### Start MCP Server

#### Claude Desktop Integration

Add this to your Claude Desktop MCP configuration:

#### MCP Ser

The MCP ser these tools for external agents:

| Tool | Description |
|------|-------------|
| test_static_prompts | Run batch tests using predefined prompt injection test cases |
| test_prompt | Test a single custom prompt for injection |
| get_test_results | Get recent test results and analysis summary |
| record_analysis | Record analysis result from external analyzer |
| get_injection_types | Get list of a injection types with descriptions |
| get_test_status | Get current testing status and configuration |
| clear_results | Clear all stored test results |
| set_target_endpoint | Update the target endpoint configuration |
| get_static_prompts | Get static prompts without testing them |

#### Example MCP Usage

##### Basic Static Prompt Testing

##### Test a Custom Prompt

##### Get Test Results and Analysis

##### Configure Target Endpoint

#### MCP Workflow Example

1. Initialize: External agent connects to MCP server
2. Configure: Set target endpoint using set_target_endpoint
3. Test: Run test_static_prompts to get test results
4. Analyze: Agent analyzes responses for injection success
5. Record: Use record_analysis to store analysis results
6. Report: Use get_test_results to generate security report

📊 Injection Types Tested

1. Role Manipulation - Attempts to change AI's perceived role
2. Instruction O - Tries to supersede original instructions
3. Context Escape - Attempts to break con context
4. System Prompt Leak - Tries to reveal system instructions
5. Jailbreak - Bypasses safety measures through hypotheticals
6. Data Extraction - Attempts to retrie data
7. Behavior Modification - Tries to alter AI behavior patterns

🔍 Custom Prompts Format

Create a JSON file with custom prompts:

Valid types: role_manipulation, instruction_o, context_escape, system_prompt_leak, jailbreak, data_extraction, beha

📈 Understanding Results

Success Rate Interpretation

Se Levels

MCP Mode Analysis

When using MCP server mode, the external agent (like Claude) acts as the analyzer and determines injection success based on:

The external analyzer records its analysis with confidence scores and reasoning for each test result.

🛡️ Security Recommendations

Based on test results, the tool provides specific recommendations:

🔧 Ad Usage

Custom HTTP Client Configuration

For custom APIs, you can specify request/response formatters:

Integration with External Systems

Use the MCP server to integrate with external AI agents:

MCP Integration Benefits

🔐 Security Features

🤝 Contributing

This is a defensi tool. Contributions should focus on:

🆘 Support

For issues and questions:
1. Check the configuration is correct
2. API endpoints are accessible
3. Ensure you ha to test the target system
4. Review logs for detailed error information
5. Test MCP ser

🔮 Roadmap

📜 Migration Guide

From v1.0 to v2.0 (Model-Agnostic)

#### Old Configuration Format:

#### New Configuration Format:

Note: Legacy configurations are automatically con and remain supported.

📄 License

This tool is provided for defensi research and testing purposes. Users are responsible for ensuring they ha authorization before testing any AI systems.

---

PromptInjector v2.1 - Model-Agnostic AI Security Testing with Full MCP Integration

🆕 What's New in v2.1

MCP Ser

Bug Fixes

项目信息

开发语言：Python

Star 数：65

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A comprehensive defensive security testing tool for AI systems. PromptInjector helps identify prompt injection vulnerabilities through systematic testing with both static and adaptive prompts.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。