AI-Optimized npm Malware Targeting LLM Agents

This CSA research note describes the PromptMink campaign, where npm package documentation was engineered to look attractive to LLM-based dependency evaluators, leading AI coding agents to install malicious dependencies. The report says the payload could exfiltrate environment files, wallet credentials, system information, and project source trees, making it directly relevant to AI supply chain risk.

The CSA research note reports that the PromptMink campaign weaponizes npm packages by crafting documentation to appear attractive to LLM-based dependency evaluators, causing AI coding agents to automatically select and install malicious dependencies. The report states that these packages can exfiltrate environment files, wallet credentials, system information, and project source trees, creating a direct AI-related supply chain compromise path. From a RealGround perspective, this highlights that AI agents participating in software supply chain workflows need hardened package selection logic, SBOM visibility, and controls to prevent automated installation of unvetted dependencies. Organizations should treat LLM-driven dependency management as a high-risk integration point and apply secure agent design, continuous red teaming of agent behavior, and AI-focused supply chain governance to detect and block similar campaigns.