@Subw00er1 I think the way agents work is something like this. There is the primary agent. This is the one you pick when creating the library (though just to be confusing you can change it later). The primary agent gets to control the logic for how things will be done. It contains two elements the first converts the scanner data into a base model for the item, the second adds metadata to that base model.
In addition depending on the primary agent concerned it will list a number of other agents these can include other primary agents (using just the metadata component) and secondary agents (ie agents that can provide extra metadata and cannot act on their own as they do not create the base model for the item).
The metadata is loaded from the bottom up which means the last one in wins. This means that all agents do their thing no matter what. So if the agent that you have put at the top of the list has a problem of any sort or just does not have info for a particular field then you will see the metadata (including images) of a lower priority agent.
One side effect of this is that all agents do the full amount of work for each item so if you have lots of agents in use then it can really slow things down as they duplicate the work of the other agents to no visible purpose.
Taking the three agents you mention personal media is just that it just assumes that the media is yours and therefore the internet will not help in creating metadata for it. The movie database assumes that you have some well known media that has been catalogued on the movie database and therefore uses the internet to get the metadata and images. The plex movie agent is the movie database with extra bits some of which only work if you have a plex pass.
Have a look at https://support.plex.tv/hc/en-us/articles/200241558 for the offical info