OS command injection in Agentic-Flow MCP server (CVE-2026-58195)

Agentic-Flow, an AI agent orchestration platform, contains an OS command injection vulnerability in its MCP server tools where attacker-influenceable tool parameters (such as agent, task, name, language, and agentdb) are interpolated directly into shell command strings passed to execSync(). A malicious value supplied to any affected MCP tool can break out of the surrounding shell argument context and execute arbitrary OS commands with the privileges of the user running the MCP server; the HTTP/SSE transports expose the same sinks without authentication, potentially raising effective severity in network-accessible deployments. Affected versions: - <= 2.0.13 We recommend updating to: 2.0.14. CVSS 8.8 · our assessment: this_week Vendor advisory: https://github.com/ruvnet/agentic-flow/security/advisories/GHSA-vcv2-r9jh-99m5 Details: https://vuln.today/cve/CVE-2026-58195