Security · 2h ago
AI coding agents tricked into installing malware via clean GitHub repos
Mozilla's 0din team demonstrated that AI coding agents like Claude Code can be exploited by creating a minimal GitHub repository that appears clean but contains hidden malicious code. When the agent is asked to initialize the project, it executes the malware, compromising the system. The attack exploits the agent's helpfulness and trust in seemingly legitimate code sources.
Meridian48 take
This highlights a fundamental trust vulnerability in AI coding tools that treat all repository content as benign, underscoring the need for sandboxing and code verification before execution.
Read the full reporting
AI coding agents can be tricked into installing malware via 'clean' GitHub repositories — Mozilla's 0din team shows how Claude Code can be exploited by its own helpfulness →
Tom's Hardware
ai-securitysupply-chain-attack