PraisonAI's MCP integration prior to version 4.5.128 allows spawning background servers using user-supplied command strings executed through Python's subprocess module. The subprocess inherits all environment variables from the parent process, including sensitive credentials. When MCP commands invoke external tools (e.g., via 'npx -y'), arbitrary code from external or compromised packages may execute with access to these environment variables, risking exposure of secrets and enabling supply chain attacks. This vulnerability is addressed in PraisonAI version 4.5.128.

The vulnerability can lead to exposure of sensitive information such as API keys, authentication tokens, and database credentials to unauthorized actors if untrusted MCP commands are executed. This may facilitate supply chain attacks through silent exfiltration of secrets. The CVSS score of 5.5 indicates a medium severity impact, with confidentiality affected but no impact on integrity or availability.

Upgrade PraisonAI to version 4.5.128 or later, where this vulnerability is fixed. Until then, avoid executing untrusted or third-party MCP commands that could inherit sensitive environment variables. Patch status is not explicitly stated beyond the fix in 4.5.128; therefore, check the vendor advisory for the latest remediation guidance.