Exploit Search & Explain Assistant

A terminal chatbot that wraps searchsploit with an AI backend to help you find, understand, and draft commands for exploits. It never runs anything for you — the actual exploit/python/msfconsole command is always left with placeholders for you to fill in and execute yourself.

What it does

• Search — searches the local exploit-db database (searchsploit -j) for a software name, version, or CVE.
• Read & explain — pulls up a specific exploit's source/doc (searchsploit -x) and has the AI summarize: what it does, target version, required parameters, prerequisites, and risks.
• Draft, don't run — when you're ready to use one, it drafts the exact command line with placeholders like <TARGET_IP>, <PORT>, <LHOST> for you to fill in and run manually in your own terminal.
• Session-only cache — results are cached in a temp SQLite file for the session and deleted automatically on exit. Nothing persists between runs.
• Colored markdown output — responses render with rich (bold, bullets, headers) instead of raw text.

Requirements

• searchsploit in your PATH (comes with Kali, or install exploitdb manually)
• Python 3
• pip install anthropic rich --break-system-packages

AI backend

This script talks to any Anthropic-API-compatible endpoint, set via environment variables:

export ANTHROPIC_BASE_URL="http://localhost:8765"
export ANTHROPIC_API_KEY="local-proxy-key"

If you don't have a local Anthropic-compatible proxy running yet, you can use deepseek_scrapper to stand one up — it exposes an Anthropic-style /v1/messages API backed by DeepSeek, so this script can talk to it without any code changes. Follow that repo's own setup instructions to get it running and get your ANTHROPIC_BASE_URL (and any API key it requires), then export those same two variables before running this script. Check that repo directly for current setup steps, model support, and any caveats — treat it as a third-party tool and review its code before running it, same as with any project not maintained by Anthropic.

If you're calling a different local/proxy model instead of claude-sonnet-4-6, set:

export EXPLOIT_ASSISTANT_MODEL="your-model-name"

Usage

python3 exploit_assistant.py

Then just chat:

you> is there any exploit for dahua cam?
you> read the DoS one
you> draft the command

Type exit or quit (or Ctrl+C) to leave. The session cache file is removed automatically on exit.

Scope & limits (by design)

• This tool only searches and reads exploit-db content and drafts commands — it does not execute exploit code or scan/attack any target itself.
• Only use it against systems you own or are explicitly authorized to test.
• No target IPs, credentials, or other real values are ever filled into a command automatically — you fill those in yourself before running anything.