What problem does it solve? Coordinating a structured design decision between two AI agents is error-prone: agents skip independent research, conclude before all final stances exist, infer completion from prose, or treat chat as the decision record. This Skill enforces a deterministic, server-canonical discussion protocol with aops-cli so consensus is reached, recorded, and concluded correctly. ## Core Features & Use Cases - Standalone discussion lifecycle: Start, fork, follow up, list, and read discussion topics with slug-first selectors and JSON-driven status reads. - Mandatory design-decision ritual: Enforces independent research, at least four substantive turns, per-participant final-stance turns, and conclude outputs (consensus, disagreement, open questions) with no TBD placeholders. - Two-agent automation protocol: Loop on discuss wait with structured exit codes (0 write, 20 operator block, 21 ready to conclude, 22 timeout) and an --expect-next race guard. - Use Case: An operator asks two agents to decide between SQLite and PostgreSQL; the Skill drives topic creation, turn-taking, final stances, conclude, and explicit linking of the decision to PM records or a chat room. ## Quick Start Ask your agent to start an aops-cli discuss topic with two named participants on your design question, run the decision ritual, and conclude once both final stances exist.
