Which harnesses do you use?

I usually use pi or minimal harnesses when I am working with Anthropic or OpenAI models. With local llms they seem to work well with maximalist harnesses like omp, Hermes etc.