There has been quite a lot of fuss over the years to explain why the Demon does not represent a violation of the Second Law of thermodynamics. Many papers bring information or complexity theory on board, explaining either that the recording or the discarding of information supplies the missing entropy, or else re-defining a new entropy which will in increase despite the Demon's efforts.
Why do all these authors have such a strong expectation that some argument must be found to restore the second law? Given what I think is a standard understanding of entropy, there is simply no reason to expect this.
Once an agent's preferred macro-parameters are settled, phase space is divided into chunks and entropy refers to the volumes of these chunks. A different agent, with access to a differing level of experimental control, would have a differing set of macro-variables and a different partitioning of phase space. Such an agent could definitely implement what appears to be, on the first agent's partition, a decrease in entropy. Couldn't they?