Anyone look into the way OpenViking organizes memory for skills and agents? Curious how the approach of openhands overlaps, bests or can be inspired by this