Research Software Engineers (RSEs) are at the coalface of ensuring that computational science is accurate, reliable and reproducible, and their views on making progress in this domain are therefore particularly valuable. This is not always recognized by the scientific community, however: work examining the challenge of developing research software tends to focus on the views of academics, and RSEs’ voices are rarely found in the evidence base. We report the results of a project that brings together researchers and RSEs, to determine how to help scientists publish their code. We interview scientists to identify challenges, and we interview RSEs to determine how to overcome them. We formalize the results into a set of recommendations, and evaluate them in a survey completed by 65 computational scientists. The survey shows that improving tool GUIs, linking internal repositories to external ones, and training in version control would all aid scientists in publishing reliable code. It also demonstrates that RSEs' views are valued by researchers: every recommendation received strong support. We conclude that RSEs can play a crucial role in scientific software policy, and their expertise should be officially recognized.