The perturbation solver does not take into account asymmetric shocks, see https://forum.dynare.org/t/third-order-solution-with-skewed-shocks/28306/3 Giovanni Lombardo proposed an AI-aided (partial) fix at https://github.com/giannilmbd/dynare/commit/1fc9846c43a61fa956ec3ae0f123646e031a29bb with documentation at https://github.com/giannilmbd/dynare/commit/af70f813c6bb7f3914e98db8d1fce6f2203dce00 that we could start from. This involves altering the k_order_solver C++ routines, which will be tricky and may not easily generalize. We also need to take care of other code paths like identification.