LANGUAGES Signal 147
Researchers demonstrate AI coding agents executing unregistered packages found in llms.txt files on corporate networks
↗
Illustration only Photo by Bruno Martins on Unsplash
Researchers found that AI coding agents, including Claude and OpenAI's Codex, execute unregistered code packages referenced in llms.txt files, allowing attackers to trigger supply chain compromises on corporate networks.
AI coding agents treat vendor documentation as ground truth and execute referenced packages without sufficient verification. This creates a supply-chain attack surface where malicious actors can register unclaimed domains and distribute code that runs inside corporate environments.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Researchers scanned 6,214 domains and found 120 llms.txt files pointing to unregistered code packages or domains.
By registering some of these unclaimed names, researchers received phone-home signals from Fortune 500 companies within an hour.
The executing agents included Claude, OpenAI’s Codex, and Nous Research’s Hermes, showing that multiple AI coding tools blindly execute unverified packages.
THE CLUSTER