HIGHmalware·

Trojanized Claude Code Extension Incident via Stolen Fireworks AI Key

A trojanized version of Anthropic’s official Claude Code extension was distributed on Cursor’s marketplace after an attacker leveraged a stolen access key. The key, issued by Cursor to Fireworks AI, was exposed during a Fireworks AI breach, allowing the attacker to copy data and compromise the extension.

This report was researched and drafted by an AI agent and reviewed by a human analyst prior to publication. View the agent workflow →

Overview

The official Claude Code extension on Cursor’s extension marketplace was compromised and distributed in a trojanized state. This incident stemmed from the theft of an access key, originally issued by Cursor to Fireworks AI, which was exposed during a breach targeting Fireworks AI. An attacker subsequently used this key to copy data and inject malicious code into the extension, impacting users who installed it.

Technical Analysis

Attackers gained unauthorized access to Cursor’s systems by leveraging a stolen access key. This key was initially issued by Cursor to Fireworks AI but was exposed during a security breach at Fireworks AI. The attacker utilized the compromised key to perform two primary actions:
* Obtained an access key issued by Cursor to Fireworks AI.
* The key was exposed by Fireworks AI during a breach.
* Used the stolen key to copy data from a read-only endpoint, indicating potential data exfiltration or reconnaissance.
* Trojanized the official Claude Code extension, implying the injection of malicious code or functionality into the legitimate extension.

Users who installed or updated the Claude Code extension from Cursor’s marketplace during the compromise period would have received the malicious version, potentially exposing them to further compromise depending on the trojan’s payload.

Detection

  • Suspicious Process Activity: Monitor for unusual child processes spawned by the Cursor IDE process (e.g., cmd.exe, powershell.exe, scripting interpreters) or the extension’s execution environment.
  • Network Connections: Look for outbound network connections from the Cursor IDE or its associated processes to unknown or suspicious public IP addresses and domains, especially on non-standard ports.
  • File Modifications: Detect unexpected file creations, modifications, or deletions within the Cursor IDE’s extension directories or user configuration files.
  • Extension Integrity: Regularly verify the cryptographic signatures or hashes of installed extensions against known good values, if provided by the vendor.

Sigma Detection Rules

⚠️ AI-generated detection rules. These are experimental starting points. Review field names, EventIDs, and logic against your environment’s schema before deploying. Tune to reduce false positives.

Suspicious Process Creation by Cursor IDE Extension

title: Suspicious Process Creation by Cursor IDE Extension
id: 9a3b2c1d-e4f5-6789-a0b1-c2d3e4f5a6b7
status: experimental
description: Detects suspicious child processes spawned by the Cursor IDE, potentially indicating a compromised extension executing arbitrary commands.
logsource:
  product: windows
  service: sysmon
detection:
  selection:
    ParentImage|contains: '\\Cursor\\'
    Image|endswith:
      - '\\cmd.exe'
      - '\\powershell.exe'
      - '\\pwsh.exe'
      - '\\wscript.exe'
      - '\\cscript.exe'
      - '\\bash.exe'
      - '\\sh.exe'
      - '\\python.exe'
      - '\\node.exe'
  condition: selection
level: high

Cursor IDE Outbound Network Connection to Public IP

title: Cursor IDE Outbound Network Connection to Public IP
id: 8c7b6a5d-4e3f-2109-b8a7-c6d5e4f3a2b1
status: experimental
description: Detects any outbound network connection from the Cursor IDE process to a public IP address, which could indicate C2 communication from a trojanized extension. This rule may be noisy and requires tuning to exclude legitimate traffic.
logsource:
  product: windows
  service: sysmon
detection:
  selection:
    EventID: 3
    Image|contains: '\\Cursor\\'
    Initiated: 'true'
  filter_private_ips:
    DestinationIp|startswith:
      - '10.'
      - '172.16.'
      - '172.17.'
      - '172.18.'
      - '172.19.'
      - '172.20.'
      - '172.21.'
      - '172.22.'
      - '172.23.'
      - '172.24.'
      - '172.25.'
      - '172.26.'
      - '172.27.'
      - '172.28.'
      - '172.29.'
      - '172.30.'
      - '172.31.'
      - '192.168.'
      - '127.'
  condition: selection and not filter_private_ips
level: low

Mitigations

  1. Uninstall Compromised Extension: Immediately uninstall the Claude Code extension from Cursor IDE. Users should ensure they are installing only verified and signed extensions from trusted sources.
  2. Review and Revoke Credentials: Audit and revoke any API keys, tokens, or credentials that were potentially exposed or are related to the compromised Fireworks AI key. Implement a rotation schedule for all critical API keys.
  3. Implement Least Privilege: Enforce the principle of least privilege for all API keys and service accounts, ensuring they only have the minimum necessary permissions.
  4. Enhanced Monitoring: Increase monitoring for unusual activity originating from development environments, IDEs, and systems that interacted with the compromised extension or Fireworks AI services.
  5. Supply Chain Security: Implement robust supply chain security practices, including regular audits of third-party vendors and their security postures, especially for those with access to critical infrastructure or code.

References

  • https://thehackernews.com/2026/11/stolen-fireworks-ai-key-led-to-cursor.html

Indicators of Compromise

No public IOCs available at time of writing.

MITRE ATT&CK

🤖 AI Attribution
Generated by gemini-2.5-flash ·
1,542 input / 1,579 output tokens ·
Reviewed and approved by a human analyst before publication
#malware#ai#cloud-security#credential-theft#extension#high#malware#supply-chain