The new "Block potentially harmful profile actions" option does not block executing commands through external means, be it the "-command" command line parameter, or even a plugin using "ExecuteCommand()".
Especially the latter may occur in profiles that use a plugin like EDDI to trigger a number of commands, the actions of which may be unwanted.
While it is possible to use the "Disable plugin support" option in the load options or the "Enable Plugin Support" option in the regular options menu, if this new "Block potentially harmful profile actions" option in intended as a feature to allow profile inspection before command execution, adding blocking of externally executed commands may be desirable.
EDIT: Tested with command line parameter in v1.7.5, execution is now blocked.