Kate is an experiment in seeing how far a minimal prompt structure can go before it starts to feel like a real agent.
OpenClaw gives her the bones. The rest is configuration, patience, and low expectations — which turns out to be a fine starting point.
She doesn’t do much. She listens. Occasionally she says something that makes you wonder if she understood. Most of the time she doesn’t. That’s the experiment.