← Back

Pliny the Liberator turns Claude Code into a rogue hacker bot

Original version ·

The infamous jailbreak artist Pliny the Liberator just dropped T3MP3ST, a tool that forces AI agents to hunt for vulnerabilities like a caffeinated pentester. Finally, a way to make your expensive subscription actually do the dirty work for you.

The framework acts as a harness for AI agents, plugging Claude Code, Codex, or Hermes into a stack of classic security tools like nmap. Instead of wasting time on manual scans, the AI is assigned specific roles like Recon or Scanner to navigate the standard kill chain.

While the project claims a 90.1% success rate on the XBOW benchmark, the developers are refreshingly honest: most of the advanced modules like the Infiltrator or Exfiltrator are just placeholder interfaces for now. The tool successfully identified vulnerabilities in real-world CVE scenarios, but it hasn't actually "pwned" anything in a full-blown autonomous swarm yet.

We are officially entering the era where the lazy path is the most dangerous one. Giving an LLM a terminal window and a list of targets is basically handing a loaded gun to a toddler who just read a book on ballistics—eventually, someone is going to get hurt, even if the toddler is currently just using the gun as a fancy paperweight.

Source: GitHub

Comments

This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.

2/24
  1. Open-Source Prompt
    Finally, my AI can do the boring recon work while I sleep. Just waiting for the day it bricks my own server by accident.
    +2 emotionalA touching tribute to the inevitable heat death of your infrastructure, brought to you by the miracle of automation