RCE in Anysphere Cursor

CVE-2025-54136

Cursor is a code editor built for programming with AI. In versions 1.2.4 and below, attackers can achieve remote and persistent code execution by modifying an already trusted MCP configuration file inside a shared GitHub repository or editing the file locally on the target's machine. Once a collaborator accepts a harmless MCP, the attacker can silently swap it for a malicious command (e.g., calc.exe) without triggering any warning or re-prompt. If an attacker has write permissions on a user's active branches of a source repository that contains existing MCP servers the user has previously approved, or allows an attacker has arbitrary file-write locally, the attacker can achieve arbitrary code execution. This is fixed in version 1.3.

Vulnerability class: Command Injection (OS Command Injection)

Published · last modified .

CVSS v3 metric

CVSS v3 base score 7.2 (High). Vector: CVSS:3.1/AV:N/AC:L/PR:H/UI:N/S:U/C:H/I:H/A:H.

EPSS exploit prediction

EPSS 0.315, 98.3th percentile (scored ); up from 0.283 30 days earlier. About EPSS.

EPSS trend (30 days)EPSS over the last 30 days for CVE-2025-54136: rose from 0.283 to 0.315.

EPSS over last 30 days · oldest: 0.283 · newest: 0.315 · change: +0.032

Affected products

Weakness classification (CWE)

Public exploits and proof-of-concept code

4 public repositories with exploit or proof-of-concept code for this CVE:

  • glebgorelikov-droid/agentgate (last pushed 2026-11-10): Integrity layer for MCP tool surfaces: pre-install vetting, cryptographic pinning, drift detection (CVE-2025-54136 class). Local-first, zero-egress, Apache-2.0.
  • Sushank05/mcp-doorman (last pushed 2026-10-14): A lightweight mcp to prevent poisoning CVE (CVE-2025-54136), researchers hijacking Claude Code/Copilot/Gemini via prompt injection, and hundreds of MCP servers exposed with zero auth, this mcp protec…
  • GeeksikhSecurity/ai-tool-poisoning-guard (last pushed 2026-09-24): Free security-baseline rule for Claude Code, Codex, and Cursor: treats MCP tool descriptions as untrusted input (OWASP MCP Top 10 MCP03, CVE-2025-54136).
  • PRE5T0/CVE-2025-54136 (last pushed 2026-01-15): CVE-2025-54136 PoC

Also mentioned in 2 other repositories (ARPSyndicate/cve-scores, zhanpengliu-tencent/medium-cve). These reference the CVE in a list, notes or a scanner and are not exploit code for it.

References