Cursor is a code editor built for programming with AI. Prior to 2.0 ,if a visited website contains maliciously crafted instructions, the model may attempt to follow them in order to “assist” the user. When combined with a bypass of the command whitelist mechanism, such indirect prompt injections could result in commands being executed automatically, without the user’s explicit intent, thereby posing a significant security risk. This vulnerability is fixed in 2.0.

Cursor is a code editor built for programming with AI. Sandbox escape via writing .git configuration was possible in versions prior to 2.5. A malicious agent (ie prompt injection) could write to improperly protected .git settings, including git hooks, which may cause out-of-sandbox RCE next time they are triggered. No user interaction was required as Git executes these commands automatically. Fixed in version 2.5.

Cursor is a code editor built for programming with AI. Prior to 2.3, hen the Cursor Agent is running in Auto-Run Mode with Allowlist mode enabled, certain shell built-ins can still be executed without appearing in the allowlist and without requiring user approval. This allows an attacker via indirect or direct prompt injection to poison the shell environment by setting, modifying, or removing environment variables that influence trusted commands. This vulnerability is fixed in 2.3.

Cursor is a code editor built for programming with AI. In versions 1.7.23 and below, a logic bug allows a malicious agent to read sensitive files that should be protected via cursorignore. An attacker who has already achieved prompt injection, or a malicious model, could create a new cursorignore file which can invalidate the configuration of pre-existing ones. This could allow a malicious agent to read protected files. This issue is fixed in version 2.0.

Cursor is a code editor built for programming with AI. In versions 1.7.44 and below, various NTFS path quirks allow a prompt injection attacker to circumvent sensitive file protections and overwrite files which Cursor requires human approval to overwrite. Modification of some of the protected files can lead to RCE. Must be chained with a prompt injection or malicious model attach. Only affects systems supporting NTFS. This issue is fixed in version 2.0.

Cursor is a code editor built for programming with AI. In versions 1.7.52 and below, manipulating internal settings may lead to RCE. Cursor detects path manipulation via forward slashes (./.cursor/./././././mcp.json etc.), and requires human approval to complete the operation. However, the same kind of manipulation using backslashes was not correctly detected, allowing an attacker who had already achieved prompt injection or some other level of control to overwrite sensitive editor files without approval on Windows machines. This issue is fixed in version 2.0.

Cursor is a code editor built for programming with AI. In versions 1.7.28 and below, an input validation flaw in Cursor's MCP server installation enables specially crafted deep-links to bypass the standard security warnings and conceal executed commands from users if they choose to accept the server. If an attacker is able to convince a victim to navigate to a malicious deeplink, the victim will not see the correct speedbump modal, and if they choose to accept, will execute commands specified by the attackers deeplink.

Cursor is a code editor built for programming with AI. Versions 1.6 and below are vulnerable to Remote Code Execution (RCE) attacks through Visual Studio Code Workspaces. Workspaces allow users to open more than a single folder and save specific settings (pretty similar to .vscode/settings.json) for the folders / project. An untitled workspace is automatically created by VS Code (untitled.code-workspace), which contains all the folders and workspace settings from the user's current session, opening up an entire new attack vector if the user has a .code-workspace file in path (either untitled created automatically or a saved one). If an attacker is able to hijack the chat context of the victim (such as via a compromised MCP server), they can use prompt injection to make the Cursor Agent write into this file and modify the workspace. This leads to a bypass of CVE-2025-54130 which can lead to RCE by writing to the settings section. This issue is fixed in version 1.7.

Cursor is a code editor built for programming with AI. In versions 1.6 and below, Mermaid (a to render diagrams) allows embedding images which then get rendered by Cursor in the chat box. An attacker can use this to exfiltrate sensitive information to a third-party attacker controlled server through an image fetch after successfully performing a prompt injection. A malicious model (or hallucination/backdoor) might also trigger this exploit at will. This issue requires prompt injection from malicious data (web, image upload, source code) in order to exploit. In that case, it can send sensitive information to an attacker-controlled external server. Some additional bypasses not covered in the initial fix to this issue were discovered, see GHSA-43wj-mwcc-x93p. This issue is fixed in version 1.7.

The configuration of Cursor on macOS, specifically the "RunAsNode" fuse enabled, allows a local attacker with unprivileged access to execute arbitrary code that inherits Cursor TCC (Transparency, Consent, and Control) permissions. Acquired resource access is limited to previously granted permissions by the user. Accessing other resources beyond previously granted TCC permissions will prompt the user for approval in the name of Cursor, potentially disguising attacker's malicious intent.

This issue was detected in 15.4.1 version of Cursor. Project maintainers decided not to fix this issue, because a scenario including a local attacker falls outside their defined threat model.

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.

ToDesktop before 2024-10-03, as used by Cursor before 2024-10-03 and other applications, allows remote attackers to execute arbitrary commands on the build server (e.g., read secrets from the desktopify config.prod.json file), and consequently deploy updates to any app, via a postinstall script in package.json. No exploitation occurred.