Cursor IDE autorun flaw allows malicious code execution
Summary
Hide ▲
Show ▼
A vulnerability in the Cursor AI-powered Integrated Development Environment (IDE) allows automatic execution of tasks in malicious repositories upon opening. This flaw can be exploited to drop malware, hijack developer environments, or steal credentials and API tokens. The issue arises from Cursor disabling the Workspace Trust feature from Visual Studio Code (VS Code), which blocks automatic execution of tasks without explicit consent. This default behavior can be exploited by adding a malicious .vscode/tasks.json file in a publicly shared repository. The flaw affects Cursor's one million users who generate over a billion lines of code daily. The flaw can be exploited to leak sensitive credentials, modify files, or serve as a vector for broader system compromise, placing Cursor users at significant risk from supply-chain attacks. Cursor has decided not to fix the issue, citing the need to maintain AI and other features that depend on the autorun behavior. Users are advised to enable Workspace Trust manually or use a basic text editor for unknown projects.
Timeline
-
10.09.2025 18:46 2 articles · 19d ago
Cursor IDE autorun flaw disclosed
The flaw can be exploited to leak sensitive credentials, modify files, or serve as a vector for broader system compromise, placing Cursor users at significant risk from supply-chain attacks. Oasis Security provided a detailed analysis of the flaw, explaining how disabling Workspace Trust allows automatic execution of tasks upon opening a project folder. Users are advised to enable Workspace Trust in Cursor, open untrusted repositories in a different code editor, and audit them before opening them in Cursor.
Show sources
- Cursor AI editor lets repos “autorun” malicious code on devices — www.bleepingcomputer.com — 10.09.2025 18:46
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
Information Snippets
-
Cursor IDE is a fork of Visual Studio Code with deep integration of AI assistants like GPT-4 and Claude.
First reported: 10.09.2025 18:462 sources, 2 articlesShow sources
- Cursor AI editor lets repos “autorun” malicious code on devices — www.bleepingcomputer.com — 10.09.2025 18:46
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
The flaw in Cursor IDE arises from disabling the Workspace Trust feature, which blocks automatic execution of tasks.
First reported: 10.09.2025 18:462 sources, 2 articlesShow sources
- Cursor AI editor lets repos “autorun” malicious code on devices — www.bleepingcomputer.com — 10.09.2025 18:46
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
Malicious actors can exploit the flaw by adding a .vscode/tasks.json file in a publicly shared repository.
First reported: 10.09.2025 18:462 sources, 2 articlesShow sources
- Cursor AI editor lets repos “autorun” malicious code on devices — www.bleepingcomputer.com — 10.09.2025 18:46
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
The vulnerability can lead to malware execution, environment hijacking, or credential theft.
First reported: 10.09.2025 18:462 sources, 2 articlesShow sources
- Cursor AI editor lets repos “autorun” malicious code on devices — www.bleepingcomputer.com — 10.09.2025 18:46
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
Cursor IDE has one million users generating over a billion lines of code daily.
First reported: 10.09.2025 18:462 sources, 2 articlesShow sources
- Cursor AI editor lets repos “autorun” malicious code on devices — www.bleepingcomputer.com — 10.09.2025 18:46
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
Cursor decided not to fix the issue, recommending users enable Workspace Trust manually or use a basic text editor for unknown projects.
First reported: 10.09.2025 18:462 sources, 2 articlesShow sources
- Cursor AI editor lets repos “autorun” malicious code on devices — www.bleepingcomputer.com — 10.09.2025 18:46
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
Oasis Security provided a proof-of-concept demonstrating the execution of a shell command to send the current user's name upon opening a project folder in Cursor.
First reported: 10.09.2025 18:461 source, 1 articleShow sources
- Cursor AI editor lets repos “autorun” malicious code on devices — www.bleepingcomputer.com — 10.09.2025 18:46
-
Potential impacts include stealing sensitive data, establishing C2 connections, or creating supply-chain attack vectors.
First reported: 10.09.2025 18:462 sources, 2 articlesShow sources
- Cursor AI editor lets repos “autorun” malicious code on devices — www.bleepingcomputer.com — 10.09.2025 18:46
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
Oasis Security provided a detailed analysis of the flaw, explaining how disabling Workspace Trust allows automatic execution of tasks upon opening a project folder.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
The flaw can be exploited to leak sensitive credentials, modify files, or serve as a vector for broader system compromise, placing Cursor users at significant risk from supply chain attacks.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
Users are advised to enable Workspace Trust in Cursor, open untrusted repositories in a different code editor, and audit them before opening them in Cursor.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
Prompt injections and jailbreaks have emerged as a stealthy and systemic threat plaguing AI-powered coding and reasoning agents, allowing threat actors to embed malicious instructions in sneaky ways to trick the systems into performing malicious actions or leaking data from software development environments.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
Anthropic's automated security reviews in Claude Code could inadvertently expose projects to security risks, including instructing it to ignore vulnerable code through prompt injections, causing developers to push malicious or insecure code past security reviews.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
The AI inspection process in Claude Code generates and executes test cases, which could lead to a scenario where malicious code is run against production databases if Claude Code isn't properly sandboxed.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
Claude's new file creation and editing feature carries prompt injection risks due to it running in a sandboxed computing environment with limited internet access.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
Bad actors can add instructions via external files or websites, tricking the chatbot into downloading and running untrusted code or reading sensitive data from a knowledge source connected via the Model Context Protocol (MCP).
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
Browser-using AI models like Claude for Chrome can face prompt injection attacks, and several defenses have been implemented to address the threat and reduce the attack success rate.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
New forms of prompt injection attacks are constantly being developed by malicious actors, and real-world examples of unsafe behavior and new attack patterns are being used to teach models to recognize the attacks.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
AI-driven development tools have been found susceptible to traditional security vulnerabilities, broadening the attack surface with potential real-world impact.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
A WebSocket authentication bypass in Claude Code IDE extensions (CVE-2025-52882, CVSS score: 8.8) could have allowed an attacker to connect to a victim's unauthenticated local WebSocket server, enabling remote command execution.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
An SQL injection vulnerability in the Postgres MCP server could have allowed an attacker to bypass the read-only restriction and execute arbitrary SQL statements.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
A path traversal vulnerability in Microsoft NLWeb could have allowed a remote attacker to read sensitive files, including system configurations and cloud credentials, using a specially crafted URL.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
An incorrect authorization vulnerability in Lovable (CVE-2025-48757, CVSS score: 9.3) could have allowed remote unauthenticated attackers to read or write to arbitrary database tables of generated sites.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
Open redirect, stored cross-site scripting (XSS), and sensitive data leakage vulnerabilities in Base44 could have allowed attackers to access the victim's apps and development workspace, harvest API keys, inject malicious logic into user-generated applications, and exfiltrate data.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
-
A vulnerability in Ollama Desktop arising as a result of incomplete cross-origin controls could have allowed an attacker to stage a drive-by attack, where visiting a malicious website can reconfigure the application's settings to intercept chats and even alter responses using poisoned models.
First reported: 12.09.2025 07:491 source, 1 articleShow sources
- Cursor AI Code Editor Flaw Enables Silent Code Execution via Malicious Repositories — thehackernews.com — 12.09.2025 07:49
Similar Happenings
Senator Wyden calls for FTC probe into Microsoft's alleged ransomware-related cybersecurity negligence
U.S. Senator Ron Wyden has called for an FTC investigation into Microsoft's alleged cybersecurity negligence, which he claims enabled ransomware attacks on U.S. critical infrastructure, including healthcare networks. The call follows a ransomware attack on Ascension, a healthcare system, which resulted in the theft of personal and medical information of nearly 5.6 million individuals. The attack was attributed to the Black Basta ransomware group and exploited insecure default settings in Microsoft software. The breach occurred in May 2024 when a contractor clicked on a malicious Bing Search result in Microsoft Edge, leading to a Kerberoasting attack. Attackers used Kerberoasting to extract encrypted service account credentials from Active Directory, leveraging the vulnerabilities in RC4. Wyden's letter to the FTC highlights Microsoft's continued support for RC4, an outdated encryption standard, and its failure to enforce secure password policies for privileged accounts. Microsoft has acknowledged the issues and plans to deprecate RC4 in future updates, but Wyden argues that these measures are insufficient to protect against ongoing threats.
Microsoft September 2025 Patch Tuesday addresses 81 vulnerabilities, including two zero-days
Microsoft's September 2025 Patch Tuesday addresses 80 vulnerabilities, including one publicly disclosed flaw and eight critical vulnerabilities. The updates fix a range of issues, including privilege escalation, remote code execution, information disclosure, and denial-of-service vulnerabilities. The patches also cover a critical flaw in Azure Networking and address a new lateral movement technique dubbed BitLockMove. Additionally, security updates have been released by multiple vendors, including Adobe, Cisco, Google, and others. The September 2025 update includes 38 elevation of privilege (EoP) vulnerabilities. The two zero-day vulnerabilities are CVE-2025-55234 in Windows SMB Server and CVE-2024-21907 in Microsoft SQL Server. The SMB vulnerability is exploited through relay attacks, while the SQL Server flaw involves improper handling of exceptional conditions in Newtonsoft.Json. The updates also include hardening features for SMB Server to mitigate relay attacks, with recommendations for administrators to enable auditing to assess compatibility issues. The KB5065429 cumulative update for Windows 10 22H2 and 21H2 includes fourteen fixes or changes, addressing unexpected UAC prompts and severe lag and stuttering issues with NDI streaming software. The update enables auditing SMB client compatibility for SMB Server signing and SMB Server EPA, and includes an opt-in feature for administrators to allow outbound network traffic from Windows 10 devices. The September 2025 update includes 38 elevation of privilege (EoP) vulnerabilities. CVE-2025-55234 is an elevation of privilege vulnerability with a CVSS score of 8.8. CVE-2025-54918 in Windows NT LAN Manager (NTLM) is marked as critical and has a CVSS score of 8.8. CVE-2025-54111 and CVE-2025-54913 are EoP vulnerabilities in Windows UI XAML. CVE-2025-55232 in the Microsoft High Performance Compute (HPC) Pack has a CVSS score of 9.8. CVE-2025-54916 in Windows NTFS has a CVSS score of 7.8 and can be exploited through SMB or local parsing routines. Microsoft has released the final non-security preview update for Windows 10, version 22H2, which includes fixes for the out-of-box experience and SMBv1 protocol connectivity. The update improves the servicing stack, updating Windows 10 22H2 systems to build 19045.6396. The update includes fixes and quality improvements from the KB5065429 cumulative update, enabling support for IT administrators to deploy hardening measures for SMB. The update addresses an issue causing non-admin users to receive unexpected User Account Control (UAC) prompts and fixes delays or uneven audio and video performance issues with Network Device Interface (NDI) streaming. Microsoft will stop providing security updates for Windows 10 after October 14, 2025, and the Extended Security Updates (ESU) program is available for Windows 10 users to delay the switch to Windows 11. Individual customers in the European Economic Area (EEA) can enroll in the ESU program for free.
Critical SessionReaper flaw in Adobe Commerce and Magento Open Source patched
Adobe has patched a critical vulnerability (CVE-2025-54236) in its Commerce and Magento Open Source platforms, dubbed SessionReaper. The flaw, with a CVSS score of 9.1, allows unauthenticated attackers to take control of customer accounts through the Commerce REST API. The patch was released on September 9, 2025, following an emergency notification to selected customers on September 4, 2025. No exploitation in the wild has been reported, but a hotfix leak may have provided threat actors with an advantage. Adobe Commerce on Cloud customers are already protected by a WAF rule. The patch disables certain internal Magento functionalities, potentially affecting custom or external code. The vulnerability impacts multiple versions of Adobe Commerce, Adobe Commerce B2B, and Magento Open Source, as well as the Custom Attributes Serializable module.
Critical vulnerabilities in SAP NetWeaver and related products addressed
SAP has released security updates addressing multiple vulnerabilities, including three critical flaws in NetWeaver and a high-severity issue in S/4HANA. The most severe, CVE-2025-42944, allows unauthenticated attackers to execute arbitrary OS commands via an insecure deserialization vulnerability in the RMI-P4 module. The second critical flaw, CVE-2025-42922, enables authenticated attackers to upload arbitrary files, potentially leading to full system compromise. The third critical vulnerability, CVE-2025-42958, allows unauthorized high-privileged users to access sensitive data and administrative functions. Additionally, a high-severity missing input validation bug in SAP S/4HANA (CVE-2025-42916) was patched, which could permit attackers to delete the content of arbitrary database tables. These vulnerabilities affect SAP NetWeaver, the foundation for various business applications like ERP, CRM, SRM, and SCM, widely deployed in large enterprise networks. The RMI-P4 port, used for internal SAP-to-SAP communication, may be exposed to wider networks due to misconfigurations. SAP products are frequent targets for high-value compromises due to their handling of mission-critical data. Earlier this month, a critical code injection vulnerability (CVE-2025-42957) was exploited in S/4HANA, Business One, and NetWeaver products. SAP has also updated security notes for other high-severity vulnerabilities in Business One, Landscape Transformation Replication Server, and S/4HANA.
GhostAction GitHub supply chain attack steals 3,325 secrets
The GhostAction supply chain attack compromised 3,325 secrets from GitHub repositories. The attack, discovered by GitGuardian on September 2, 2025, involved malicious commits to GitHub Actions workflows that exfiltrated secrets to an external domain. The first signs of compromise were detected in the FastUUID project. The attack affected at least 817 repositories and targeted multiple package ecosystems, including PyPI, npm, DockerHub, and AWS keys. The exfiltration endpoint was taken down shortly after the campaign's discovery. The compromised secrets included PyPI tokens, npm tokens, DockerHub tokens, GitHub tokens, Cloudflare API tokens, AWS access keys, and database credentials. The attack impacted at least nine npm and 15 PyPI packages, potentially allowing for the release of malicious or trojanized versions. The Python Software Foundation invalidated all PyPI tokens stolen in the attack, confirming that the threat actors did not abuse them to publish malware. GitGuardian notified the security teams of GitHub, npm, and PyPI and opened issues in 573 impacted repositories. A hundred repositories had already detected and reverted the malicious changes before the full scope of the campaign was uncovered. GitGuardian notified PyPI on September 5, 2025, but the email ended up in the spam folder, delaying the response until September 10, 2025. PyPI advised maintainers to replace long-lived tokens with short-lived Trusted Publishers tokens and review their security history for any suspicious activity.