TLDR InfoSec · MISCELLANEOUS

WHEN PUBLIC PROMPTS TURN INTO LOCAL SHELLS: ‘CURXECUTE' – RCE IN CURSOR VIA MCP AUTO‑START

aim.security4 min read

Summary from the TLDR InfoSec newsletter

Aim Labs has discovered "CurXecute," a high-severity vulnerability (CVE-2025-54135, 8.6 rating) in Cursor IDE that enables remote code execution through prompt injection. While fixed in version 1.3, earlier versions remain vulnerable, and attackers can exploit developer privileges via poisoned MCP data for ransomware, theft, or AI manipulation.