⚠️ LEGAL DISCLAIMER — READ BEFORE USE This tool is for authorized security testing only. Launching network attacks (e.g., port scans, DDoS) against targets you do not own or lack written permission to test is illegal in most jurisdictions, including under the EU Cybercrime Directive 2013/40/EU and similar laws worldwide.

By default, this tool blocks internal/private IP ranges (RFC 1918) to prevent accidental targeting of local networks. The safety keyword system flags dangerous operations (ddos, flood, attack, etc.) and requires explicit second confirmation.

You are solely responsible for ensuring you have proper authorization before testing any target.

Vulntest AI Agent

An open-source, prompt-driven AI agent for automated vulnerability command execution using configurable keyword-command mappings. 100% offline. Simple and easy to build on and customize for your own needs!

⚠️ IMPORTANT — Use agent_ver4.py versions/agent_ver1.py, versions/agent_ver2.py, and versions/agent_ver3.py are legacy/unsafe. They lack the safety checks added in ver4 (private IP blocking, authorization confirmation, input sanitization, dangerous-command guards, and command logging). Do not run the legacy versions against real targets. Only agent_ver4.py is supported.

Contributors

Started on August 4th, 2025 by C0m3b4ck.

Why I made this

I was watching some YouTube but then got annoyed at the constant cheesy AI ads telling that prompt engineering is the past and agents are the future. They also said how if I don't sign up for their webinar I will suffer. In order to prove them wrong, I made this and I quite like it! I though about such an "agent" earlier but never had the motivation.

Requirements

• Python 3.x (tested on Python 3.13)
• Runs on Linux (tested on Ubuntu Linux), Windows might need modified file handling which will soon come :)
• Uses standard Python libraries: os, re, subprocess, socket

Installation

Simply clone or download the script and create a configs folder alongside it.

Add .conf files in configs describing keywords and commands (see Features).

Run with:

python3 agent_verX.py

How it Works

• User Prompt: You input a natural language prompt requesting vulnerability testing or network commands.

• Configuration Matching: The agent reads your prompt, then looks for words from the keyword that are also found in .conf files in the configs folder. Each .conf file contains:

  • First line: comma-separated keywords (what makes this .conf to be picked)
  • Second line: the shell command template, with {target} parameter for IP/domain replacement

• Target Extraction:

  • First attempts to extract an IPv4 address from prompt.
  • If none found, extracts domain names and resolves the first via DNS to an IP address.

• Command Execution: The chosen command from the matched .conf file is customized with the extracted target and presented to the user for confirmation.

• User Confirmation: Only executes on explicit user approval (if user inputs "yes").

• Redirects: If the prompt was previously incorrectly matched, a redirects.conf file can override the conf selection. That allows for "training" the model or rather manually correcting its mistakes.

Features

• Configurable Exploit/Network Commands: Easily add or modify .conf files with keywords and commands to extend functionality without changing code.
• Prompt-based AI Agent: Works with natural language inputs including domains and IPs.
• Domain Name Resolution: Automatically resolves domains found in prompts to IPs for command execution.
• User Interaction: Prompts before command execution and error handling with option to correct config matches.
• Redirect Learning: User can teach the AI better matching via redirect mappings.
• Cross-Platform: Runs on Windows and Linux environments with Python installed.
• Dangerous Prompt Detection: Checks for words written in safety.conf in prompt (for example "ddos"), then warns user about potential consequences

Folder Structure

agent_ver1.py              \# Main AI agent script
configs/                   \# Folder containing keyword-command .conf files
configs/redirects.conf     \# Saves user redirects (created automatically)
configs/config_here.conf   \# Any configs you might want to add

Example .conf Files

ping.conf

ping,test,icmp
ping -c 4 {target}

ddos.conf

ddos,stress,website
sudo hping3 --flood -S {target}

Roadmap / Future Enhancements

• Add more functions to be extracted from keywords - if you need a specific one, just ask!
• Support multiple target extraction per prompt
• Add logging and analytics of executed commands (and debugger mode)
• Support other command substitution tokens besides {target} - if you need a specific one, just ask!

What is up to you

• Adding additional arguments (like "no installation") to .conf files
• Making .conf files, though I will publish some example ones every now and then

Disclaimer

Use responsibly. This tool is intended for authorized penetration testing and network diagnostics only. Using the tool to attack without consent is illegal.

---

If you have questions, feature requests or want to contribute, reach out to C0m3b4ck.