CVE-2025-54135
[cursor] Arbitrary code execution from Cursor Agent through a prompt injection via MCP Special Files
Executive Summary
CVE-2025-54135 is a high severity vulnerability affecting appsec, ai-code. It is classified as OS Command Injection. Ensure your systems and dependencies are patched immediately to mitigate exposure risks.
Precogs AI Insight
"Precogs AI Analysis Engine identifies web application vulnerabilities through semantic code analysis, detecting injection flaws, broken authentication, and insecure data flows across your entire codebase."
What is this vulnerability?
CVE-2025-54135 is categorized as a critical Application Verification Flaw flaw. Based on our vulnerability intelligence, this issue occurs when the application fails to securely handle untrusted data boundaries.
Summary Cursor allows writing in-workspace files with no user approval. If the file is a dotfile, editing it requires approval but creating one if it d.
This architectural defect enables adversaries to bypass intended security controls, directly manipulating the application's execution state or data layer. Immediate strategic intervention is required.
Risk Assessment
| Metric | Value |
|---|---|
| CVSS Base Score | 8 (HIGH) |
| Vector String | N/A |
| Published | August 2, 2025 |
| Last Modified | August 2, 2025 |
| Related CWEs | CWE-78, CWE-829 |
Impact on Systems
✅ Unauthorized Access: Flaws in application logic can permit unauthorized interaction with protected APIs.
✅ Data Manipulation: Adversaries may alter critical application states, such as user roles or configurations.
✅ Service Disruption: Improper error handling or unvalidated inputs can lead to resource exhaustion.
How to fix this issue?
Implement the following strategic mitigations immediately to eliminate the attack surface.
1. Defense in Depth Implement multi-layered validation (client-side, API gateway, and server-side).
2. Least Privilege Ensure backend service accounts operate with the absolute minimum rights required.
3. Security Regression Testing Integrate automated semantic security scanning into the deployment pipeline.
Vulnerability Signature
// Generic Application Security Flaw (Node.js)
app.post('/api/update-profile', (req, res) =\> \{
// DANGEROUS: Mass Assignment / Object Injection
// Attacker can pass \{ "isAdmin": true, "email": "..." \}
User.update(\{ id: req.user.id \}, req.body);
// SECURED: Explicitly select permitted fields
const \{ email, displayName, bio \} = req.body;
User.update(\{ id: req.user.id \}, \{ email, displayName, bio \});
\});