xadmin-client upstream commits installed malicious editor hooks

The frontend for a Django-and-Vue administration system acquired malicious development hooks while its issue bots continued their ordinary work. The September 22 source snapshot of nineaiyu/xadmin-client contained the same implanted payload found in five other repositories. SafeDep identified the group as infections by the credential-stealing Mini Shai-Hulud worm linked to the compromised actions-cool/issues-helper action.

xadmin-client's README described a Vue 3 interface for a role-based access-control system, derived from vue-pure-admin. Its archived workflows used issues-helper to label reports, remove labels, and close issues that had received no feedback for thirty days. Those routines explain why a frontend repository depended on an external GitHub Action at all. The malicious files, however, appeared in .claude and .vscode, where a developer's local coding tools would find them.

Inspection showed a Claude Code session hook and a VS Code folder-open task pointing to node .claude/setup.mjs. The setup program downloaded Bun and ran a bundled JavaScript payload whose digest matched the other five captures. The settings thus connected source distribution to local command execution without needing to alter a visible application screen. Execution was conditional on using the relevant tool and allowing the project-defined command, particularly VS Code's automatic task.

The retained commit proved that the upstream checkout distributed the hooks. It did not prove that a deployed xadmin administration website served malware, that a package registry published the altered files, or that every contributor ran them. The incident's immediate downstream audience was developers handling that source. Their potential exposure came from the commands in the checkout, while actual credential theft or persistent access would require additional endpoint evidence.