Automated tasks usually consist of an action sequence, an optional guard and preconditions or postconditions. They are executed by the workflow server.
A guard is used to activate the automated task depending on some condition. For details about when conditions are reevaluated, see Section 4.1.4.1, “Common Features of All Tasks”.
Actions within an automated task usually modify workflow variables, manipulate resources, perform calculations and/or access external systems. However, they may not access the Client GUI, since they are not executed on the client side, as the workflow server uses a direct connection to the Content Management Server for automated tasks. If you want GUI interaction, you have to use a user task.
Several actions which are to be executed sequentially should be given as an action sequence within a single automated task, not as a sequence of automated tasks. This is easier to understand and will be executed faster. The general rule of identifying different tasks by potentially different users can also be applied here, if you consider automated tasks as being accepted and performed by a "robot".
An automated task completes as soon as all it actions have been executed and its optional postcondition is evaluated. If an action raises an exception or the postcondition evaluates to false, the automated task is aborted. The reason that led to the error should be fixed before the task is retried. As a last resort, the whole workflow can be aborted.
 |  |  |
 |