First, it plots the classic RDD graph: a scatterplot with the running variable in the x-axis and the outcome in the y-axis, the discontinuity dividing the graphic in the middle and the treated and control observations on plotted on both sides.   

 Second, I created a function that takes in the outcome variable, the running variable and the treated individuals and spits out a plot with point estimate and confidence interval (y-axis)  over different bandwidths (x-axis), highlighting the Imbens-Kalyanaraman bandwidth. This is the most transparent way of presenting RDD results, in my opinion.

Third, you should show that a set of covariates not affected by the running variable (i.e., are balanced on both sides of the discontinuity). This can be shown in a single plot.