This paper refines Putnam’s Two-Level Games (TLG) by incorporating recent work on Principal-Agent approaches. I argue that Putnam’s conception of the domestic ratification process is underdeveloped. I assume that all governments are agents of domestic principals, which can vary across time and space. Thus, a “domestic win-set,” i.e. the overlap between a government’s and its principal’s win-sets, emerges. The constellation of win-sets (i.e. the government’s, the domestic principal’s, and the international) influences the likelihood of an agreement being signed and implemented. For example, if the government’s – but not the domestic principal’s – win-set overlaps with the international (Level I) one, the government will try to circumvent its principal. Furthermore, ceteris paribus, the breakdown of agreements is caused by changing win-sets of domestic principals: In cases of involuntary defection, a government fails to circumvent its principal. I apply this refined TLG framework to two cases: U.S. and North Korean policies over the latter’s nuclear weapons program from 1994 to 2008. Finding that shifting win-sets of the respective domestic principals (and, indeed, the principals themselves changing) were crucial to U.S. and North Korean behavior and the fate of several agreements, I conclude by outlining this framework’s benefits and avenues for further research.