The agent set’s the execution policy automatically at runtime.
The agent output with a little bit debug code to see where the plugin fails is the right thing what is needed now. I think that it fails at loading the SnapIn, but then we should see a minimum error message.