Loading vulnerability…
CurXecute is a vulnerability in Cursor versions earlier than 1.3.9 that permits the AI agent to create workspace files without user approval. Although modification of an existing dotfile requires approval, creation of a new dotfile does not. An attacker who can deliver an indirect prompt injection can exploit this discrepancy to cause Cursor to create a sensitive MCP configuration, enabling subsequent remote code execution without user confirmation.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
Patch, then assume compromise.
No valid public exploits. Mallory filtered out 3 candidates as fakes, detection scripts, or README-only repos.
All candidate exploits were filtered out by Mallory's validation.
Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.
Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.
22 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
A pair of vulnerabilities enabling remote code execution through an IDE's Model Context Protocol (MCP) implementation.
A high-severity indirect prompt-injection vulnerability in Cursor versions prior to 1.3.9. A malicious analyzed file could cause creation of a Cursor MCP configuration and lead to remote code execution without user confirmation; the article associates it with CWE-78 and CWE-829.
An earlier Cursor prompt-injection-related vulnerability that allowed command execution by rewriting Cursor MCP configuration even after the user rejected the edit.
A vulnerability/attack chain affecting AI coding assistant and MCP-integrated developer environments, where untrusted repository files such as hidden AI configuration files can trigger malicious command execution or credential exfiltration when a project is opened.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.