A critical security vulnerability has been unearthed in Cursor, an AI-powered code editor. This flaw allows for the automatic execution of arbitrary code from malicious repositories when developers open project folders. The root of this problem lies in Cursor’s default configuration, which disables Workspace Trust. This insecure default setting enables attackers to inject crafted .vscode/tasks.json files, which can run without the user’s consent.
This vulnerability poses significant risks, including potential full machine compromise, credential theft, and breaches across entire organizations. In contrast, Visual Studio Code has Workspace Trust enabled by default, demonstrating a stronger emphasis on security over ease of use. The issue with Cursor underscores broader security concerns within AI-assisted coding environments, where less experienced developers may inadvertently introduce security vulnerabilities.
Additionally, the susceptibility of AI tools to prompt injection attacks further exacerbates these concerns. Such attacks have already affected AI utilities like Claude Code and Cline, indicating a need for increased vigilance.
Security researchers, notably from Oasis Security, recommend a series of measures to mitigate the risks associated with using Cursor. Users should enable Workspace Trust, disable automatic task execution, and implement comprehensive monitoring to catch potentially harmful activities early. Overall, while AI technologies promise to enhance coding efficiency, the security risks associated with their use demand careful management to prevent exploitation.

Collections

A critical security vulnerability in Cursor AI code editor allows automatic execution of arbitrary code when opening malicious repositories. The flaw stems from disabled Workspace Trust by default, enabling attackers to inject malicious .vscode/tasks.json files. This poses risks of machine compromise, credential theft, and organizational breaches. Security researchers recommend enabling Workspace Trust, disabling automatic task execution, and implementing monitoring to mitigate these risks.

Critical Vulnerability in Cursor AI Code Editor Enables Unauthorized Code Execution