A proof rests on assumptions about the legitimate devices that the implementation must satisfy: a secure perimeter no adversary can enter, classical interfaces without side channels into it, a quantum interface that cannot be used to probe internal settings, quantum mechanical models of sources and detectors, and quantified deviations between model and device taken into account in privacy amplification (device-independent proofs are an alternative that still needs the perimeter and interface assumptions). Any restriction assumed on the adversary (no quantum memory, short-term, bounded or noisy memory) must be clearly stated and consistent, and holds only during key establishment; claims against specific attacks do not give meaningful security statements, which should be made against the adversary's whole toolbox.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.