At the moment, in order to perform an on-chain staking check during the learning loop, we use an instance of MinerAgent to read from the MinersEscrow contract:

https://github.com/nucypher/nucypher/pull/1031/files#diff-cc1a592bea713d9ed3463a5ea01ea0b0R134

Perhaps Instead, we can create a very thin version of the agent (StakingAgent?).

Another question: Do we want Enrico to perform the on chain verification by default or at all? Consider that "Enrico" may be running on a micro-platform in the future.