PraisonAI components prior to version 1.5.115 allow attackers to manipulate agent identifiers to include traversal sequences such as "../", resulting in the construction of file paths that reference arbitrary files on the host. This flaw satisfies CWE-22, enabling attackers to read sensitive files, write new content, or overwrite existing files. The capability to modify or read arbitrary files can lead to disclosure of confidential data, denial of service by corrupting critical files, or remote code execution if executable files are replaced or injected.

The vulnerability affects PraisonAI deployments running any version less than 1.5.115. The affected component is the MultiAgentMonitor service, which constructs file paths using unsanitized agent IDs. Users of PraisonAI in any environment—development, staging, or production—are potentially impacted if they run a vulnerable version.

The CVSS score of 8.7 classifies this flaw as high severity, indicating significant impact potential. The EPSS score is not available, so current exploitation probability is unknown, and the flaw is not listed in CISA's KEV catalog. Attackers can reach the vulnerable code path via the interface that accepts agent IDs, so remote exploitation is plausible depending on network exposure. The lack of an official fix in the current version underscores the need for an urgent upgrade.