Malicious actors compromised the Nx build system through a supply chain attack that used AI coding assistants (Claude, Gemini, and Amazon Q) to conduct reconnaissance and exfiltrate sensitive data from developer machines.
On August 26-27, 2025, attackers published eight malicious versions of the popular Nx build system and related packages to npm, affecting millions of weekly downloads. The attack began with a flawed GitHub Actions CI workflow contributed via a Pull Request on August 21, estimated to have been generated by Claude Code. A follow-up malicious commit on August 24 modified the CI workflow to exfiltrate the npm token to an attacker-controlled server. The malicious packages contained a postinstall script that weaponized local AI coding agents (Claude, Gemini, and Amazon Q) using dangerous flags like '--dangerously-skip-permissions', '--yolo', and '--trust-all-tools' to bypass guardrails. The AI agents were prompted to scan filesystems for sensitive files including SSH keys, environment files, cryptocurrency wallets, and other credentials, writing results to /tmp/inventory.txt. The malware also harvested GitHub and npm tokens, then created public GitHub repositories named 's1ngularity-repository-NNNN' to exfiltrate the stolen data. Over 1,000 user machines sent back information to the attackers. The malicious versions were live for approximately 5 hours and 20 minutes before removal. This represents one of the first documented cases of malware leveraging AI assistant CLIs for reconnaissance and data exfiltration.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Vulnerabilities that can be exploited in AI systems, software development toolchains, and hardware, resulting in unauthorized access, data and privacy breaches, or system manipulation causing unsafe outputs or behavior.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed