Skip to main content

ShipSafe

ShipSafe
ShipSafe
67% of Cursor-built apps have critical vulnerabilities

Security Scanner forCursor Apps

We scanned 100 real Cursor-built apps. 67% had at least one critical vulnerability. IDOR, inverted auth, hardcoded secrets, and three CVEs published in 2025 alone.

See What Cursor MissedRead the Full Research
Free scan2 minutesNo card needed

The Short Answer: Cursor Is Safe, But Its Code Isn’t

Cursor the editor is SOC 2 Type II certified and offers Privacy Mode to keep your code off its servers. The security risk isn't the tool — it's the code the tool generates.

A Stanford University study (Perry et al., 2024) found that developers using AI code assistants produce significantly less secure code, with roughly 45% of AI-assisted code containing vulnerabilities. Our own research confirms this: scanning 100 real Cursor-built repositories, we found 67% had at least one critical vulnerability, with an average of 3.2 security issues per app.

On top of code-level risks, three CVEs were published against Cursor in 2025:

  • CVE-2025-54135 (CurXecute) — Malicious Slack messages processed by Cursor could rewrite MCP configs and execute arbitrary commands
  • CVE-2025-54136 (MCPoison) — Shared repository configs enabled persistent team-wide compromise
  • CVE-2025-59944 — Case-sensitivity bypass allowed attackers to modify protected configuration files

The bottom line: use Cursor for speed, but scan before you ship. Read our full analysis of 100 Cursor-built apps for the complete data.

Common vulnerabilities

Common Vulnerabilities in Cursor Projects

These are the security issues we find most frequently in codebases built with Cursor. For a full breakdown, see our vulnerabilities by platform reference.

Critical

IDOR with Sequential IDs

Cursor frequently generates API routes using auto-incrementing IDs without ownership checks. An attacker changes /api/invoices/42 to /api/invoices/43 and accesses another user's data. Found in 43% of the 100 Cursor apps we scanned.

Critical

Inverted Auth Conditions

A single misplaced negation operator flips your auth logic: authenticated users get blocked, anonymous requests get through. We found this in 31% of Cursor-built apps. It passes manual testing because you test while logged in.

High

Frontend-Only Admin Checks

Cursor generates admin panels where role checks only happen in React components. The API endpoints behind them accept requests from anyone. Present in 28% of scanned apps.

High

Hardcoded Secrets in Source

Cursor pulls API keys from your context window and embeds them directly in code. These get committed to Git history permanently. According to GitGuardian, 12.8 million secrets were exposed on GitHub in 2024 alone. Found in 22% of our scanned repos.

Critical

MCP & Prompt Injection Risks

CVE-2025-54135 (CurXecute) showed that malicious content processed by Cursor can rewrite MCP configs and execute arbitrary commands. CVE-2025-54136 (MCPoison) enabled team-wide compromise through shared repository configs.

High

Auto-Run Code Execution

Cursor ships with Workspace Trust disabled by default (CVE-2025-59944). A malicious .vscode/tasks.json in a cloned repo can execute code the moment you open the folder — no click required.

How ShipSafe Secures Your Cursor Project

1

Connect Your Repository

Link your GitHub repo or paste your code. ShipSafe supports any Cursor-generated project regardless of framework.

2

Automated Security Scan

Our scanner analyzes every file for the vulnerability patterns that Cursor's AI commonly introduces, including auth logic, data access, and secret management.

3

Get Fix Suggestions

Receive a prioritized report with severity ratings and concrete code fixes you can apply immediately, no security expertise needed.

Frequently Asked Questions

Is Cursor safe to use for production code?
Cursor the editor is SOC 2 Type II certified and secure. The risk is in the code it generates: according to a Stanford study (Perry et al., 2024), roughly 45% of AI-assisted code contains vulnerabilities. Our scan of 100 Cursor-built apps found 67% had at least one critical vulnerability. Use Cursor for speed, but always scan before deploying to production.
What are the security risks of using Cursor?
There are two categories of risk. First, code-level vulnerabilities: IDOR (found in 43% of apps we scanned), inverted authentication logic (31%), frontend-only admin checks (28%), and hardcoded secrets (22%). Second, tool-level risks: three CVEs were published in 2025 — CurXecute (CVE-2025-54135), MCPoison (CVE-2025-54136), and a file protection bypass (CVE-2025-59944). Update Cursor regularly and scan your code with a tool like ShipSafe.
How do I secure code generated by Cursor?
Five steps: 1) Add security instructions to your .cursorrules file. 2) Review every auth-related code path manually. 3) Run an automated security scanner like ShipSafe before every deploy. 4) Use server-only imports for sensitive operations. 5) Never commit secrets to Git — use environment variables and a secrets manager.
How is ShipSafe different from other vibe coding scanners?
ShipSafe scans your actual GitHub repository — not just a URL. It uses AI to analyze authentication flows, authorization boundaries, and secret management across your entire codebase. Most competitors only run regex-based pattern matching on surface-level code. ShipSafe performs 17 security checks including OWASP Top 10 vulnerabilities, and generates plain-English reports with fix suggestions.
Does ShipSafe work with Cursor Composer and Agent mode?
Yes. ShipSafe scans your actual codebase regardless of which Cursor feature generated it — tab completion, inline editing, Composer multi-file generation, or Agent mode. The scanner analyzes the output, not the tool that produced it.

67% of Cursor Apps Have Critical Vulnerabilities. Is Yours One of Them?

Paste your GitHub URL. 2 minutes. Plain-English report — no security background needed. Find out if yours is cooked before your users do.

Scan My Cursor App Free

No credit card required. See all plans