When an Agent is run with Fleet, today any input that is supported by a package can be run. But there are cases where the administrator of the Agent itself would like to limit the type of inputs that can be run. Few examples:

• An agent should only be used for heartbeat pings
• An agent should not be allowed to run endpoint
• An endpoint should only be used to collect system metrics but not logs

The Agent needs a way to allow / block list certain inputs when it is run so Fleet cannot send down any configs which contain these inputs or the not allowed inputs are ignored. Which inputs are supported or not supported should also be sent up to Fleet to make Fleet smart around what can be configured and what not.