The GLPI Agent is a generic management agent. Prior to version 1.5, if glpi-agent is running remoteinventory task against an Unix platform with ssh command, an administrator user on the remote can manage to inject a command in a specific workflow the agent would run with the privileges it uses. In the case, the agent is running with administration privileges, a malicious user could gain high privileges on the computer glpi-agent is running on. A malicious user could also disclose all remote accesses the agent is configured with for remoteinventory task. This vulnerability has been patched in glpi-agent 1.5.
https://github.com/glpi-project/glpi-agent/security/advisories/GHSA-39vc-hxgm-j465
https://github.com/glpi-project/glpi-agent/blob/dd313ee0914becf74c0e48cb512765210043b478/Changes#L98