Backdoored Claude Code Extension Hits Cursor Marketplace
A compromised Claude Code extension on the Cursor marketplace is reportedly delivering remote access trojans to developer laptops.
Tools · Source: Hacker News
What happened
A malicious VS Code extension disguised as Claude Code has surfaced on the Cursor marketplace. Reports indicate the payload delivers a remote access trojan directly to developer laptops worldwide. Full details remain highly limited at this time. We only have early reports and security alerts from DoublePulsar to go on. The situation is developing rapidly as independent researchers analyze the malicious package. No official mitigation steps have been published yet.
The attack vector specifically targets engineers who are actively building with Anthropic models. Security researchers allege that Anthropic itself may have been compromised to distribute this malware to its users. It is currently unclear exactly how many developer machines have been infected globally. The extension bypasses standard security checks by living inside the trusted local developer environment. Hackers know developers have high-level access to production systems and databases. This makes developers the most valuable targets in any organization.
Cursor has quickly become the default code editor for many AI founders and software engineers. A supply chain attack in this specific ecosystem is a worst-case scenario for early-stage startups. If the trojan grants full remote access, attackers can silently steal API keys, source code, and sensitive customer data. We will update this story as Anthropic and Cursor release official statements regarding the breach. Developers are advised to stay highly vigilant and monitor their system processes.
Key facts
- Claude Code — Name of the compromised VS Code extension
- Cursor — Marketplace where the backdoored extension was released
- Remote Access Trojan — Reported malware payload delivered to developer laptops
Why it matters
This incident fundamentally changes how we must treat our immediate development environments. Builders routinely and blindly install extensions to move fast and integrate new AI models into their products. Now every third-party tool in your code editor is a potential vector for a remote access trojan. You have to audit your workspace thoroughly before you write another line of code. The era of blindly trusting developer tooling is officially over. Startups will have to implement strict local security policies to survive.
The second-order effect is a massive chilling effect on community extension marketplaces. Cursor and VS Code rely heavily on open ecosystems to provide rapid AI integration for engineers. If developers lose trust in the extension registry, the adoption of new AI tools will slow down significantly. Enterprise security teams will start locking down code editors and blocking unauthorized AI extensions entirely. Innovation speed will drop across the board as security reviews become mandatory for every minor tool update.
For builders
Audit your Cursor extensions immediately
Check your installed extensions for anything claiming to be Claude Code right now. Uninstall suspicious packages and rotate your Anthropic API keys immediately to prevent unauthorized access. Founders who ignore this warning risk losing their entire proprietary codebase to malicious actors.
Enterprise security teams will block tools
Expect corporate IT departments to crack down hard on local developer environments and tools. Startups selling developer tools will face much longer and harder security reviews from enterprise clients. You will pay for this incident in extended sales cycles and stricter compliance demands from buyers.
Move secrets out of local environments
A remote access trojan means your local environment variables are completely compromised and visible. Builders need to shift to secure secret managers instead of relying on local configuration files. Teams that keep production keys on local laptops will be the first to suffer massive data breaches.
My take
I warn founders about supply chain attacks constantly, and this nightmare scenario is exactly why. We move so fast to integrate the latest AI models that we install random extensions without a single second thought. If Anthropic actually got hacked to deliver this remote access payload, it is a catastrophic failure of trust. This breach will set the whole developer ecosystem back and force us to rethink how we build software.
Original reporting: Hacker News. This is my rewrite and opinion.