Skip to content

CVE-2026-30615: Windsurf MCP Prompt Injection RCE

Analysis indicates that the vulnerability originates from insecure processing of MCP configurations. The application fails to sanitize inputs provided in JSON format, allowing command injection.

The attack vector leverages the MCP adapter configuration logic. By injecting a malicious JSON configuration containing arbitrary command and argument values, the LiteLLM process executes these inputs on the host system.

title: Windsurf MCP Prompt Injection RCE Detection
logsource:
category: process_creation
detection:
selection:
CommandLine: '*mcp*register*'
condition: selection