Hacker News
'Orchid' Repo-Confusion Campaign: 9,330+ Malicious GitHub Repos Detected, 100K+ Infected, AI Coding Agents Targeted
A solo researcher (alias 'Orchid') mapped a largely automated GitHub campaign of thousands of fake repositories carrying crypto-stealing trojans, publishing a detection script and a list of 9,330 malicious repos; Apiiro pegs the broader repo-confusion problem at 100,000+ infected repositories. Attackers clone fresh repos (not popular ones), strip the README, inject poisoned download buttons, and re-commit every few hours to evade detection — and the fakes often outrank originals in search. The acute risk for builders: AI coding agents that pull dependencies or 'find a library' can be steered straight into poisoned repos, so agent file/network access needs hard guardrails.
↳ Follow the thread