Show GN: Geas - AI Agent가 실수하지 않도록 엄격한 업무 프로토콜을 지키게 하기

As AI agents like Claude Code become essential to professional work, this account reveals their hidden pitfall: they can enter infinite loops of refinement and scope creep, progressively deviating from original goals. The proposed 'Geas' framework shifts focus from prompting to constraint-based control through strict protocols, offering a crucial methodology for ensuring AI remains disciplined and focused in complex, long-running tasks.