StanleyNull/AutoHunter

What it solves

AutoHunter automates the tedious process of vulnerability discovery. It transforms a machine into a 24/7 autonomous bug-hunting platform, allowing a human operator to act solely as a "reviewer" who makes final decisions on whether a discovered vulnerability is valid and should be submitted, rather than manually performing the reconnaissance and scanning phases.

How it works

The system uses a multi-agent pipeline to handle the full lifecycle of vulnerability hunting:

1. Collector: Gathers targets using FOFA (or other search engines like Shodan or Censys) via FOFA syntax or natural language intents, then performs health checks and scoring.
2. Worker: Assigns one LLM-driven agent per target. The agent autonomously performs reconnaissance and executes real security tools (nmap, nuclei, sqlmap, httpx, whatweb) to find vulnerabilities.
3. Reviewer: An AI agent that performs a rational first-pass review to filter out false positives and "half-baked" findings, ensuring only high-quality leads reach the human.
4. Human Reviewer: The user reviews the filtered results via a dashboard to approve, edit, or reject the findings.
5. Expansion: Once a vulnerability is found, the system can analyze if the flaw is widespread ("one hit, many kills") and automatically verify similar targets.

Who it’s for

Security researchers and bug hunters who have explicit written authorization to test targets and want to scale their discovery process using AI agents.

Highlights

• Autonomous Tool Execution: Unlike simple LLM wrappers, it executes real-world binaries like sqlmap and nuclei inside containers.
• 24/7 Unattended Operation: Designed to run continuously, with automatic task resumption after restarts.
• Natural Language Target Search: Can translate plain-text descriptions (e.g., "find university login systems") into complex FOFA search queries.
• Intelligence Reuse: Validated credentials, endpoints, and fingerprints are stored in a global intelligence library for future workers to reuse.
• Multi-Engine Support: Compatible with FOFA, 360 Quake, Hunter, ZoomEye, Shodan, and Censys.