CVE-2026-33475
Langflow · Langflow
Unauthenticated remote shell injection in Langflow's GitHub Actions workflows allows attackers to execute arbitrary commands and exfiltrate CI secrets via malicious branch names.
Executive summary
Langflow is vulnerable to unauthenticated remote shell injection in its CI/CD pipelines, allowing attackers to steal sensitive secrets and compromise the software supply chain.
Vulnerability
This is a shell injection vulnerability in GitHub Actions workflows. It occurs because unsanitized GitHub context variables (like github.head_ref) are directly interpolated into run: shell commands, allowing an unauthenticated attacker to execute commands by submitting a pull request with a malicious branch name.
Business impact
An attacker can exfiltrate sensitive CI secrets, such as GITHUB_TOKEN, and potentially push malicious code or images to production repositories. This could lead to a massive supply chain compromise affecting all users of the Langflow product. The CVSS score of 9.1 reflects the critical risk to the integrity of the development lifecycle.
Remediation
Immediate Action: Update Langflow to version 1.9.0 or later, which refactors workflows to use environment variables instead of direct interpolation.
Proactive Monitoring: Audit GitHub Actions logs for suspicious activity, such as unusual curl commands or unauthorized access to secrets in workflows triggered by external forks.
Compensating Controls: Restrict GitHub Actions permissions and require manual approval for all workflows triggered by pull requests from external contributors.
Exploitation status
Public Exploit Available: Yes
Analyst recommendation
This is a critical supply chain vulnerability. We strongly recommend that all organizations using Langflow immediately update to version 1.9.0 and rotate any secrets that may have been exposed in their CI/CD environments.