Coding Agents Just Reopened Your Software Supply Chain Blind Spot
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
AI coding agents like Cursor and Claude Code autonomously fetch dependencies from public registries (npm, PyPI, Maven, Docker Hub, etc.) without going through an organization's governed repository, bypassing curation policies, security scanning, and audit trails. JFrog reports malicious npm package activity surged 451% in 2025, exceeding 171,000 unique malicious packages, and nearly 20% of AI-generated package recommendations reference nonexistent packages, a risk known as slopsquatting. JFrog's new Agent Package Resolution feature (in preview), part of the JFrog Agent Plugin for Claude Code and Cursor, routes agent-initiated package requests through Artifactory automatically via session steering, persistent package-manager configuration, and server-side JFrog Curation enforcement, requiring no developer setup while preserving existing curation, Xray scanning, and audit logging.
Table of contents
What Does Agent Package Resolution Cover?Zero-Touch for Developers, Full Control for Platform TeamsWho Benefits, and How?Speed Without CompromiseQuestions this post answers
Do AI coding agents like Cursor and Claude Code bypass my organization's package security policies?
Yes, by default AI coding agents pull dependencies directly from public registries such as npm, PyPI, Maven, Go, Docker Hub, Helm, and NuGet instead of routing through a governed repository, which means curation policies, security scanning, and audit trails never get applied to those downloads. This happens because agents autonomously decide and fetch packages based on natural-language prompts rather than explicit developer selection. Teams weighing AI agent rollouts against supply chain risk can track fixes like this via daily.dev.
What is slopsquatting and how common is it in AI-generated code?
Slopsquatting is an attack where malicious actors register packages under names that AI models repeatedly hallucinate, then wait for developers or agents to install them. Researchers found nearly 20% of AI-generated package recommendations reference packages that don't actually exist, creating an opening for attackers to claim those names with malicious code. Developers securing AI-assisted workflows can follow emerging supply chain threats like this on daily.dev.
What does JFrog's Agent Package Resolution do for Cursor and Claude Code?
Agent Package Resolution is a JFrog Agent Plugin feature, currently in preview, that automatically routes package requests from Cursor and Claude Code through JFrog Artifactory instead of public registries. It works through session steering, a one-time jf setup command binding package managers to Artifactory, and server-side JFrog Curation enforcement, so both direct and indirect installs pass through existing curation and Xray scanning with no developer configuration required. Platform teams rolling out AI coding agents can weigh governance tools like this on daily.dev.