CVE-2026-17625

Langflow command handling did not neutralize OS special characters

IBM / Langflow · Langflow OSS · MCP platform

First documented disclosure · Review checkpoint

Summary

A privileged authenticated remote user could execute arbitrary OS commands.

Severity and exploitation evidence

High · 7.2/10 · CVSS 3.1. Severity source.

Exploitation not established: The reviewed sources document a weakness or controlled demonstration; malicious exploitation against users is not established.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

OS special characters were not properly neutralized before command execution.

Conditions required

Remote authenticated access with high privileges, as specified by the CNA CVSS vector.

Potential impact

Commands execute with the Langflow process authority, affecting confidentiality, integrity and availability.

Affected versions and configurations

1.0.0 through 1.10.3 according to IBM.

Fix and mitigation status

1.11.0 or later according to the IBM remediation table.

Practical next steps

Upgrade to 1.11.0 or later. IBM lists no workaround; restrict access to the affected configuration or MCP feature pending upgrade.

Operating lesson

Apply authentication and validation consistently before configuration becomes execution.

Evidence and limits

Historical backfill from the complete IBM bulletin, initially published July 31 and modified August 24. The exact CNA and issue-specific bulletin paragraph establish this identity and CVSS 3.1 assessment. The bulletin gives a shared 1.11.0 remediation table but no issue-specific commit or independent release-path verification for this CVE; no neighboring GHSA or patch is inferred. Mechanism detail is limited to the original description. No proof of malicious exploitation is established in these sources.

Sources