AI safety & securitymedium
Critical Cursor Flaws Could Let Prompt Injection Escape Sandbox and Run Commands
Vulnerabilities mentioned
All| CVE | Vulnerability | CVSS | EPSS | Signals | Published |
|---|---|---|---|---|---|
| CVE-2025-54135 | Cursor: Cursor allows writing in-workspace files with no user approval in versions below 1.3.9, If the file is a dotfile, editing it… | 9.8 | 2% | ||
| CVE-2025-54136 | Trusted MCP Config Swap Enables RCE in Cursor AI Code Editor ≤1.2.4 | 8.8 | 27% | ||
| CVE-2026-26268 | Cursor: Sandbox escape via writing .git configuration was possible in versions prior to 2.5. A malicious agent (ie prompt injection)… | 9.9 | <1% | ||
CVE-2026-50548+1 related CVE | Cursor: Cursor runs agent terminal commands in a sandbox by default, and the sandbox grants write access to the command's working… | 9.3 | 1% |
From the original article
4 min read at thehackernews.com
Two flaws in Cursor, an AI code editor, could let a single, ordinary-looking prompt break out of the editor's safety sandbox and run any command on a developer's computer. There is no click to fall for and no approval box to ignore. Cato AI Labs found the pair and named them DuneSlide .
By @TheHackersNews · thehackernews.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.