When AI assistants generate scientific scripts or simulations, common failure modes immediately appear. Last week, RSE Faruk Diblen released rseng-agent-skills to fix this. It packages 67 Research Software Engineering practices into native agent skills so AI coding agents follow the research software best practices and open-science standards while writing code, not after the fact. Check it out 👉 https://github.com/fdiblen/rseng-agent-skills