Goal-Oriented Skill Abstraction for Offline Multi-Task Reinforcement Learning | Read Paper on Bytez