CVE-2026-5023: Command Injection in codebase-mcp

The command injection vulnerability in codebase-mcp allows an attacker with local access to execute arbitrary commands on the system. This could involve gaining persistent access, stealing sensitive data, modifying system configurations, or even pivoting to other systems on the network. The ability to execute OS commands directly bypasses normal application security controls, making it a high-impact vulnerability. While the description specifies local exploitation, successful command execution could be leveraged to escalate privileges or compromise the underlying infrastructure, significantly expanding the blast radius.

1. Reserved2026-03-27
2. Published2026-03-29
3. Modified2026-03-30
4. EPSS updated2026-04-04

Due to the rolling release model of codebase-mcp, a direct upgrade to a patched version is not currently available. Mitigation strategies should focus on restricting local access to the codebase-mcp component and implementing robust monitoring. Consider implementing strict access controls to limit which users or processes can interact with the vulnerable function. Implement input validation and sanitization to prevent malicious commands from being injected. Monitor system logs for suspicious command execution patterns, particularly those related to the getCodebase/getRemoteCodebase/saveCodebase function. Consider using a Web Application Firewall (WAF) or proxy to filter potentially malicious requests, although this may not be effective for locally triggered commands. There are no specific Sigma or YARA rules available at this time.