Executive brief
gin-vue-admin is a development platform used to build applications with AI assistance. A security flaw in its code-generation and management tools allows an authorized user to inject malicious code into the system's backend. If exploited, an attacker can take full control of the server, potentially leading to data theft, service disruption, or the installation of persistent backdoors.
Technical details
A remote code execution (RCE) vulnerability exists in gin-vue-admin version 2.9.1 due to improper neutralization of special elements in the code-generation feature. An authenticated attacker can use the `POST /autoCode/addFunc` endpoint to inject arbitrary Go source code into backend files by manipulating fields such as `humpPackageName` and `serverFunc`. By subsequently calling `POST /autoCode/mcpStart`, the attacker can trigger a rebuild and restart of the standalone MCP service, executing the injected code (e.g., a malicious `init()` function) with the privileges of the application process. This vulnerability is most severe in environments where the source tree is writable and local build capabilities are enabled. As of publication, no patch is available, but users can mitigate the risk by implementing strict allowlist validation on path and identifier fields.
Affected products
- flipped-aurora gin-vue-admin 2.9.1
Timeline
- 2026-05-26: advisory: GitHub Security Advisory published
- 2026-06-19: disclosed: CVE-2026-48787 published to NVD