Official Claude Code VS Code Extension Compromised to Install Persistent SSH Backdoor

Cybersecurity researchers have disclosed that the official Anthropic Claude Code extension was compromised on the Open VSX registry to install a persistent SSH backdoor on infected hosts. The issue has since been traced to the compromise of the Cursor Marketplace, prompting the AI coding platform Cursor to briefly take down the service to remove the malware.

The attack is believed to involve threat actors targeting the AI coding extension ecosystem. The compromised extension likely masqueraded as the legitimate Anthropic Claude Code tool, allowing attackers to gain persistent SSH access to developer workstations. This access could enable remote code execution, data exfiltration, and lateral movement within compromised networks.

The Cursor platform confirmed the compromise and took the marketplace offline temporarily to remove the malicious extension. Anthropic has not yet released an official statement, but users are advised to verify extension publishers, check for updates, and consider removing the compromised extension from their installations.

The incident highlights the broader trust erosion in the AI coding extension supply chain, where dependencies and marketplace integrity are critical to developer security.