Симуляция совещания AI-агентов

★ 6.5 · ai-agents

multi-agent-meeting is a Claude Code skill that simulates a structured meeting among multiple AI agents with distinct specializations and personalities to reach a collaborative decision. The discussion unfolds across five phases: opening statements, open debate, focused argumentation, consensus convergence, and final decision generation. Each session configures 3–5 agents defined by name, domain expertise, personality traits, and speaking style — for example, a technical architect, product manager, or financial advisor — using role templates from `references/agent-roles.md` and output formatting from `references/meeting-record-format.md`. The skill produces a full meeting record including the live discussion log, key arguments, risk alerts, and recommended next steps. It fits architecture reviews, pricing strategy workshops, project risk assessments, and any scenario where a problem benefits from structured multi-perspective analysis.