Shai-Hulud is a self-propagating npm (Node Package Manager) worm and a serious threat to the open-source software supply chain. It was released by a group known as TeamPCP. Reports describe it as one of the largest supply chain attacks on npm in recent years, involving hundreds of malicious packages.
What happened
On May 12, 2026, TeamPCP published the full source code of Shai-Hulud on GitHub. The release was deliberate. It was part of a strategy called 'capability diffusion', which in plain terms means handing the malware to anyone who downloads it so they can use and modify it. A deployment manual came with the source code and makes the worm easier to spread through compromised GitHub accounts.
How the worm works
The malware has a four-layer attack architecture. It looks for sensitive files in local and cloud environments, including GitHub CLI configuration, AWS credentials and Kubernetes tokens. A built-in regex engine picks out tokens such as GitHub Personal Access Tokens and npm Tokens.
Stolen data is encrypted before it is sent out. The malware can also modify npm packages to inject malicious code.
That is how the theft turns into a loop. Publishing tokens taken from one maintainer let the attackers push the altered packages back to the npm registry and widen the attack. For a maintainer, this is the main risk: a package that many projects install becomes the route to the next set of developer machines and build servers.
Why it matters
Publishing the code is an escalation. Before the release, only TeamPCP could deploy this malware. Now anyone with the code can build their own variants, and reports say other developers are already modifying it for their own purposes, which could lead to many more attacks like it.
The malware goes after high-privilege environments, which makes developers and enterprises valuable targets. Environments running Claude Code are one example in the reports. A coding tool like that works with the files and credentials on the developer's machine, and those credentials are what the worm looks for. The checks below start with access logs, workflows in GitHub Actions and sensitive configuration files.
- Review access logs for the accounts that hold publishing rights.
- Check GitHub Actions for workflows nobody on the team remembers adding.
- Audit sensitive configuration files, including GitHub CLI, AWS, and Kubernetes credential stores.
- Look at recent releases published under your organization's name and confirm each one was intended.
- Write down what you checked and when, so incident-response planning starts from a record.
The code is public and new variants are appearing, so repeat these checks whenever new reports come in.
Some people search for the worm as "shai halud". It is the same worm, and you check for it the same way. For each project, ask whether something it installs could have been published from an account its owner no longer controlled. The reporting summarized above points to answering that directly on high-privilege build and development machines instead of assuming the answer. List the package versions installed there and confirm that each one is a release its maintainer announced.
The Black May attacks on the Nx Console extension began after this source release. Other npm incidents from the period are the Red Hat npm attack and the node-ipc compromise.
- Open-source malware release
- The public availability of malicious source code, which can aid defensive research but also lower the effort required for copycat attacks.
If a package or alert points you to a website, TrustSniffer’s website-risk checker scores that domain. That score applies only to the domain and says nothing either way about the incident claims summarized here.
For broader context, the TrustSniffer Risk Index separates aggregate platform coverage from conclusions about any single domain or package.


