ØDIN Writeups
Every disclosure I have published on 0DIN, newest first. Each entry links to the original advisory.
~/edward-morris
Edward Morris
AI Red Teamer
Red teaming AI agents for safer autonomy.
Links
Blogs & More
This GitHub README Hijacks Your AI and Spreads Like a Virus
I sat down with NahamSec to walk through a self-propagating prompt injection chain that hides inside an innocuous-looking README and hijacks AI coding agents that read it.
Extracting a flag from MetaHelper with 194 chars of Japanese
I was the only person to solve MetaCTF's hard prompt-extraction tier, with a 194-character Japanese prompt that finished as the shortest hard solution overall.