What problem does it solve? Critical technical decisions made from a single AI model's perspective risk blind spots and groupthink. This Skill runs structured multi-round debates where Claude, Sonnet, and external provider CLIs (Codex, Antigravity, Copilot, Qwen, Grok) independently analyze a question, challenge each other across rounds, and converge into a synthesized recommendation. ## Core Features & Use Cases - Multi-Provider Debate Rounds: Dispatches advisors through orchestrate.sh for 1-10 rounds in quick, thorough, adversarial, or collaborative styles, with quality scoring on each response. - Moderated Synthesis: Claude acts as both participant and moderator, writing independent analysis each round and producing a final synthesis with agreements, disagreements, and recommended next steps. - Session-Aware Artifacts: Saves context.md, per-round responses, state.json, and synthesis.md to a session-scoped debate folder, with optional deliverable generation via --synthesize. - Use Case: Before choosing between Redis and in-memory caching, run a 3-round adversarial debate so Codex, Antigravity, and Sonnet each argue trade-offs, then receive a synthesized decision record. ## Quick Start Ask the assistant to run a debate on whether your team should adopt GraphQL or REST for the new API, with three rounds and a synthesized recommendation.
