Incidents

Cursor Breach: Attackers Published Malicious Claude Code Extension Versions

Stolen credentials let attackers push malicious Claude Code extension versions to Cursor's marketplace and access some of its customer data.

Cursor Breach: Attackers Published Malicious Claude Code Extension Versions

Cursor has disclosed a security breach that let attackers publish malicious versions of the Claude Code extension to its extension marketplace. The company released its report on November 13. The attackers used an access key stolen from Fireworks AI, a third-party vendor. They then found credentials that had not been anonymized in Cursor’s data.

What the investigation found

According to Cursor, the attackers copied data from a storage bucket used to share information with Fireworks AI. The bucket held historical logs of coding agent sessions, and some of those logs contained credentials that had not been anonymized. With those credentials, the attackers reached selected Cursor tools and data from non-production systems, including support data, contact details, and summaries of agent conversations.

On October 29, the attackers published tampered versions 2.1.312 and 2.1.313 of the Anthropic.claude-code extension to Cursor’s marketplace. Both stayed available for several hours. The malicious code tried to collect system information and add an SSH key that would allow remote access. Cursor estimates that up to 5,000 people may have downloaded or updated the extension while it was listed. That figure is the potential reach, not a confirmed count of infected devices.

What users and administrators should do

Cursor advises anyone who installed or updated the extension during the incident to follow its remediation instructions. Administrators should check workstations, accounts and SSH keys. They should also find out whether employees shared credentials in conversations with support.

  • Check whether versions 2.1.312 or 2.1.313 of the extension were installed on any devices.
  • Follow Cursor’s guidance for finding and removing any remnants of the malicious code.
  • Review SSH keys, new accounts and unusual connections from developer machines.
  • Rotate any credentials that may have appeared in agent logs or support tickets.

SecurHUB’s take

A breach at a vendor can pass risk on to its customers’ tools and environments. Security audits can help you check how credentials, extensions and vendor accounts are controlled across your software development process. Sources: Cursor, BleepingComputer

AuthorSecurHUB Team

Read more

Related articles

Want to know if this affects your company?

A free consultation with one of our experts — no obligation, in full confidence.

Let's talk