The various events that occur at a crack on an interface are explored, and described in terms of a simple graphical construction called the crack stability diagram. For simple Griffith cleavage in a homogeneous material, the stability diagram is a sector of a circle in the space of stress intensity factors, ${\mathit{K}}_{\mathrm{I}}$/${\mathit{K}}_{\mathrm{II}}$. The Griffith circle is limited in both positive and negative ${\mathit{K}}_{\mathrm{II}}$ directions by nonblunting dislocation emission on the cleavage plane. For a branching plane inclined at an angle to the original cleavage plane, both cleavage and emission (which blunts the crack) can be described as a balance between an elastic driving force and a lattice resistance for the event. We use an analytic expression obtained by Cotterell and Rice for cleavage, and show that it is an excellent approximation, but show that the lattice resistance includes a cornering resistance, in addition to the standard surface energy in the final cleavage criterion. Our discussion of the lattaice resistance is derived from simulations in two-dimensional hexagonal lattices with UBER force laws with a variety of shapes. Both branching cleavage and blunting emission can be described in terms of a stability diagram in the space of the remote stress intensity factors, and the competition between events on the initial cleavage plane and those on the branching plane can be described by overlays of the two appropriate stability diagrams. The popular criterion that ${\mathit{k}}_{\mathrm{II}}$=0 on the branching plane is explored for lattices and found to fail significantly, because the lattice stabilizes cleavage by the anisotropy of the surface energy. Also, in the lattice, dislocation emission must must always be considered as an alternative competing event to branching.