CVE-2026-34938 MervinPraison CVE debrief

Who should care

Users of PraisonAI versions prior to 1.5.90 should apply the patch to prevent arbitrary OS command execution. This vulnerability has a critical CVSS score of 10 and should be prioritized for immediate attention. Security teams and operators should review the official advisory and CVE record to validate affected systems and apply the patch.

Technical summary

The execute_code() function in praisonai-agents runs attacker-controlled Python inside a three-layer sandbox. A vulnerability allows bypassing this sandbox by passing a str subclass with an overridden startswith() method to the _safe_getattr wrapper, enabling arbitrary OS command execution on the host. This issue was patched in version 1.5.90. Affected users should apply the patch to prevent exploitation.

Defensive priority

High priority for PraisonAI users to apply the patch and monitor for potential exploitation attempts.

Recommended defensive actions

• Apply the patch to update PraisonAI to version 1.5.90 or later
• Monitor for potential exploitation attempts
• Review and update inventory of affected systems
• Implement compensating controls to detect and prevent exploitation
• Review the official advisory and CVE record to validate affected scope and severity
• Track exceptions and retest remediated assets
• Confirm whether affected product deployments exist in managed environments

Evidence notes

The CVE record and NVD detail provide information on the vulnerability and patch. The vendor advisory on GitHub provides additional context on the issue and mitigation. Evidence limits suggest that further verification is required to confirm affected scope and severity. Defenders should review the official advisory and CVE record to validate affected systems and apply the patch. Limited source detail indicates that additional research may be needed to fully understand the vulnerability.

Official resources