Formalizing Potential of Agents: The KARO Framework Revisited