Malicious npm Package Stole Files From Claude AI User Directory via GitHub

Summary: Cybersecurity researchers have discovered a new malicious package on the npm registry that comes with information stealing capabilities. According to OX Security, the package, named "mouse5212-super-formatter," is designed to upload files from "/mnt/user-data," a dedicated directory used by Anthropic's Claude artificial intelligence (AI) tool to handle uploads and outputs in the background. The

A malicious npm package has been caught stealing sensitive files from developers’ systems, once again exposing how fragile the modern software supply chain has become. According to researchers, the package was designed to quietly harvest data from infected environments after being installed by unsuspecting developers, demonstrating how attackers continue targeting open-source ecosystems as an efficient path into larger organizations.

The incident highlights a dangerous reality inside modern software development: developers increasingly trust enormous ecosystems of third-party packages, dependencies, and open-source libraries that are often installed automatically with minimal scrutiny.

The npm ecosystem alone contains millions of packages powering applications, cloud infrastructure, APIs, enterprise platforms, mobile apps, and developer tooling worldwide. Most developers rely heavily on external dependencies to accelerate development and avoid reinventing common functionality.

Attackers understand this dependency culture extremely well.

Rather than attacking organizations directly, cybercriminal groups increasingly target the software supply chain itself. By compromising packages, repositories, developer accounts, or build systems, attackers may potentially reach thousands of downstream systems through a single successful intrusion.

In this case, researchers say the malicious npm package was specifically engineered to steal files from compromised machines after installation. The malware reportedly searched for sensitive information such as credentials, tokens, configuration files, development secrets, SSH keys, browser data, or cloud authentication material that could later be used for deeper compromise.

This type of attack is especially dangerous in developer environments.

Modern developer workstations frequently contain privileged access to cloud platforms, CI/CD pipelines, internal repositories, deployment systems, production infrastructure, and corporate APIs. Compromising a developer machine can therefore provide attackers with enormous operational leverage inside enterprise environments.

Many recent high-profile breaches began exactly this way.

Security experts increasingly warn that developers themselves have become one of the most strategically important attack surfaces in cybersecurity. A single compromised dependency can potentially spread malicious code into build pipelines, software updates, customer applications, or production systems affecting millions of users.

The npm ecosystem has repeatedly faced this problem.

Attackers routinely publish typosquatted packages, cloned libraries, malicious updates, dependency confusion payloads, and poisoned developer tools designed to trick developers into installing malware accidentally. In some cases, attackers compromise legitimate maintainer accounts to distribute malicious updates through previously trusted packages.

The trust model underlying open-source ecosystems creates enormous efficiency — but also enormous risk.

Most developers cannot realistically audit every dependency they install. Modern applications may rely on hundreds or even thousands of indirect packages inherited through nested dependency chains. This complexity creates opportunities for attackers to hide malicious code deep inside software ecosystems where detection becomes difficult.

Artificial intelligence may amplify these risks even further.

AI-powered coding assistants increasingly recommend packages, generate installation commands, and suggest dependencies automatically. Researchers warn that attackers may attempt to manipulate repositories, documentation, or package names specifically to influence AI-generated coding recommendations.

This creates a new intersection between AI and software supply chain attacks.

The malicious npm package incident also reflects how cybercrime increasingly targets operational trust relationships rather than exploiting technical vulnerabilities alone. If attackers can convince developers to install malicious software willingly, they may bypass many traditional defensive controls entirely.

In many cases, no exploit is required.

Security researchers continue emphasizing the importance of package verification, dependency auditing, code signing, behavioral monitoring, and repository reputation analysis to reduce exposure. Organizations are also increasingly deploying software composition analysis (SCA) tools and stricter supply chain governance policies to monitor third-party dependencies more aggressively.

Still, the scale of the challenge remains enormous.

Open-source ecosystems evolve at extraordinary speed, with new packages and updates appearing constantly across npm, PyPI, RubyGems, Maven, Cargo, and other developer repositories. Attackers exploit this velocity aggressively because malicious packages may remain available long enough to infect systems before researchers identify and remove them.

The broader lesson extends far beyond a single npm package.

Modern software development now depends on interconnected global ecosystems built around trust, automation, collaboration, and open-source sharing. Those same qualities that enable rapid innovation also create ideal conditions for supply chain compromise.

And as developers increasingly become gateways into cloud infrastructure, enterprise systems, and production environments, malicious packages may continue evolving into one of the most effective attack vectors in modern cybersecurity.

Key facts

  • - The package is named 'mouse5212-super-formatter'.
  • - It targets the '/mnt/user-data' directory of Anthropic's Claude AI.
  • - OX Security discovered and reported this threat.

Why it matters

The theft of user data could compromise privacy and security for those using Claude AI. Additionally, this incident highlights vulnerabilities in software distribution channels like npm.