What problem does it solve? Getting genuinely independent perspectives from an AI is hard — a single model voicing multiple viewpoints tends to blend them into one agreeable answer. This Skill runs a structured roundtable where distinct personas (installed BMAD agents or custom characters) argue, clash, and build on each other like real people, optionally backed by independent subagents so each voice truly thinks for itself. ## Core Features & Use Cases - Multi-persona roundtables: Run lively group discussions with installed agents or custom personas, with four execution modes (session, auto, subagent, agent-team) controlling whether voices are voiced inline or backed by real independent agents. - Custom party authoring: Create themed casts, one-off personas, named party groups with scenes, or distill customer data into an AI focus-group panel, saved via TOML customization overrides. - Per-party memory: Optional append-only memlogs let a party remember past sessions — grudges, alliances, callbacks — so dynamics carry forward across conversations. - Use Case: Run the built-in Code Review Crew, where a security engineer, an adversary, an edge-case hunter, a craftsman, and a pragmatist attack your code from different lenses and argue about what actually matters before it ships. ## Quick Start Start party mode and have the installed agents debate the tradeoffs of my proposed caching architecture.
