A methodology to deal with shape optimization in flow problems is described. The scheme is based on the solution of a continuous adjoint problem. The computation of the objective function sensitvities is performed by using only the flow and adjoint solutions on the boundary. This property makes the scheme totally independent of the numerical solvers (flow and adjoint). In addition, the scheme allows the use of incomplete gradients by only taking into account the geometrical sensitivities, and/or the boundary conditions of the adjoint problem. The error generated by this procedure is documented in the paper. Finally, an improved pseudo-shell approach to move the surface to be optimized, and to impose the geometric restrictions is proposed and tested. Some applications for control and design problems are shown.