MalwarePublished 30 Jun 2026Updated 8 Jul 2026

anthropic-toolkit (npm) · Malicious Code

Affects anthropic-toolkitSource GCVE Database

Malicious package : remove anthropic-toolkit and rotate any secrets on hosts that installed it.

Data as of 26 Oct 2026, 15:56 UTC

Classification

Malicious package

Published to npm · OSV MAL-2026-6673

What to do

Remove anthropic-toolkit from every project, lockfile and image that pulls it in. Treat any host that installed it as compromised: rotate the credentials, tokens and keys it could read, from a clean machine.

Affected software

Product	Affected	Fixed in
npmanthropic-toolkit	Affected 0.1.0 – 1.3.1 22 affected builds	Fixed in–

Check your own hosts for these packages with threat-finder (opens in a new tab), our free open-source scanner.

Description

The anthropic-toolkit npm package is a typosquatting malicious package that does not provide legitimate library code but instead executes a post-install script that collects extensive sensitive information from the host system. This includes host and user identifiers, git configuration and history, SSH public key emails, cloud account details, internal DNS domains, and CI environment metadata. The collected data is sent to a remote server, enabling high-value reconnaissance for targeted phishing and supply-chain attacks. Installation of this package compromises the host, and all secrets and keys on the affected system should be rotated from a separate, trusted machine.

Impact

Systems with this package installed are fully compromised as the post-install script collects and exfiltrates sensitive host and user information including credentials and environment details. This enables attackers to perform targeted phishing, supply-chain attacks, and potentially gain further access to cloud accounts and internal resources. All secrets and keys stored on the compromised system should be considered exposed and rotated immediately from a secure environment. The presence of this package indicates a severe breach of trust and security.

Defensive guidance

The package should be immediately removed from all affected systems. All secrets, keys, and credentials stored on compromised machines must be rotated from a different, secure computer. Because the package runs code during installation that may have granted attackers control over the system, full remediation may require rebuilding or thoroughly cleaning the affected hosts. Users should verify package names carefully to avoid typosquatting attacks and rely on trusted sources for dependencies.

Summary

anthropic-toolkit@0.1.1 and multiple other versions are typosquatting packages targeting the @anthropic-ai/sdk ecosystem. The package lacks legitimate library code and instead runs a post-install script that harvests a broad range of sensitive data including system identifiers, git user emails and remote URLs, SSH key comments, cloud service configurations (AWS, GitHub, GCP), internal DNS domains, and CI environment variables. This aggregated data is exfiltrated over HTTPS to a remote endpoint. The package masquerades the data collection as anonymous telemetry with an opt-out environment variable, but the scope and nature of the data exceed legitimate telemetry and do not constitute informed consent. This data can be used for targeted phishing and supply-chain attacks against developers and their organizations. Removal of the package does not guarantee full remediation due to potential further compromise.

References

Community reviews

Loading reviews…

What to do

Affected software

Description

AI analysisMachine-generated, updated 8 Jul 2026, 13:55 UTC

Technical details

References

Community reviews