Inside the TanStack npm Supply‑Chain Attack and the Mini Shai‑Hulud Worm

Modern software development depends heavily on open‑source packages and automated CI/CD pipelines. In May 2026, attackers exploited that trust model in a coordinated supply‑chain campaign known as Mini Shai‑Hulud, compromising more than 170 npm and PyPI packages used across major developer ecosystems.

The campaign targeted popular projects including TanStack, UiPath, Mistral AI, and OpenSearch, spreading malware that attempted to steal developer and cloud credentials. Instead of stealing maintainers’ passwords, the attackers exploited weaknesses in automated GitHub Actions release pipelines to publish malicious versions through legitimate channels.

The TanStack compromise

TanStack was one of the most visible victims of the campaign. On May 11, 2026, between 19:20 and 19:26 UTC, attackers published 84 malicious versions across 42 packages in the @tanstack/* namespace on npm.

Security analysis found that the altered packages contained credential‑stealing malware targeting developer machines and CI/CD systems.

Many TanStack packages are widely used in the React and JavaScript ecosystem, with some libraries downloaded millions of times weekly. That popularity meant compromised releases could quickly spread into developer environments and build pipelines.

The TanStack incident turned out to be only one part of a larger coordinated operation. Across ecosystems, researchers later identified:

• 172 compromised packages
• More than 400 malicious versions
• Activity across both npm and PyPI

Most of these releases appeared during a 48‑hour window on May 11–12, 2026.

How attackers exploited GitHub Actions

Instead of directly stealing npm publish credentials, the attackers targeted the automation used to publish packages.

TanStack’s post‑incident analysis found that the attack combined several weaknesses in GitHub Actions workflows:

• misuse of the pull_request_target workflow pattern
• cache poisoning across the fork‑to‑base repository boundary
• extraction of a GitHub Actions OIDC token from the runner process

By chaining these issues together, attacker‑controlled code executed inside the release pipeline and extracted credentials capable of publishing packages. The malicious versions were then released through the project’s legitimate publishing identity.

Because the pipeline itself performed the release, the packages could appear authentic and carry valid build provenance or signatures. This helped the malicious releases blend in with normal package updates.

How the Mini Shai‑Hulud worm spread

The malicious packages carried code designed to behave like a self‑propagating supply‑chain worm.

When installed, the packages executed scripts during dependency installation or runtime that downloaded additional payloads. In several cases this happened through package lifecycle hooks or code executed during import.

Once running, the malware attempted to:

• harvest credentials and secrets
• access CI/CD environments
• reuse stolen tokens to publish additional compromised packages

This approach allowed the attack to move laterally between developer machines, build systems, and open‑source projects.

Because package managers routinely execute scripts during installs, simply installing a compromised dependency could activate the malware.

Credentials and systems targeted

The malware focused primarily on developer and cloud credentials, which could enable further compromises across infrastructure and repositories.

Security reports indicate the worm searched systems for secrets such as:

• AWS IAM credentials
• GitHub personal access tokens
• npm publish tokens
• HashiCorp Vault tokens
• Kubernetes service account credentials
• SSH private keys
• Docker configuration files and environment variables

Researchers also observed the malware scanning developer machines and CI runners across many file paths for stored secrets and configuration data.

Any environment that installed a compromised package was considered potentially exposed and required credential rotation.

What OpenAI said about user data

Because some affected libraries were widely used across developer ecosystems, questions emerged about potential downstream impact on technology companies.

OpenAI stated that it found no evidence that user data was accessed or breached as a result of the supply‑chain incident involving the TanStack libraries.

The statement addressed speculation that services depending on affected packages might have exposed customer information.

Why the attack mattered

The Mini Shai‑Hulud campaign highlighted several major shifts in supply‑chain attacks.

Compromising automation instead of accounts
Attackers bypassed the need for maintainer credentials by hijacking automated release pipelines.

Signed malicious packages
Because the pipeline published the packages, malicious releases could appear legitimate and carry trusted provenance.

Cross‑ecosystem reach
The campaign simultaneously targeted both npm and PyPI, dramatically increasing its potential impact.

Worm‑like propagation
The malware attempted to steal credentials and reuse them to compromise additional packages and systems automatically.

Together, these factors made the campaign one of the most significant open‑source supply‑chain incidents reported in 2026.

Security lessons for developers and teams

The response from maintainers and security researchers reinforced several practices for defending modern software supply chains.

Treat dependency installs as code execution
Installing a package can run scripts or import code that executes immediately.

Rotate credentials after potential exposure
Any environment that installed affected packages should rotate all accessible secrets.

Harden CI/CD workflows
Security guidance emphasized reviewing GitHub Actions configurations, limiting token permissions, and avoiding unsafe workflow patterns such as pull_request_target where possible.

Audit dependency versions
Teams were urged to verify that no compromised versions were installed during the May 2026 incident window.

Monitor build and publishing pipelines
Unexpected dependency installs, unusual network activity, or unauthorized publishing actions may indicate supply‑chain compromise.

The bigger lesson for open‑source ecosystems

Modern software supply chains rely on automation and shared dependencies. The Mini Shai‑Hulud attack demonstrated how those same systems can become powerful propagation mechanisms when compromised.

By targeting build pipelines and developer infrastructure, attackers can distribute malware through trusted channels and reach thousands of downstream projects within hours.

As a result, many organizations now treat CI/CD pipelines, dependency management, and developer environments as critical security boundaries, not just convenience tooling.