design-council-orchestration

What problem does it solve?

It prevents high-risk design decisions from being made in a single viewpoint by orchestrating a structured, multi-role debate that produces a binding decision log.

Core Features & Use Cases

• Multi-agent design council orchestration: Spawns a roster of role-specialized Claude agents in parallel (e.g., security, performance, QA, platform, product) where the invoking Claude acts as CEO.
• Parallel, independent reasoning with peer cross-talk: Uses seat-to-seat DMs for disagreements (rather than sequential turn-taking through a single context).
• Decision-grade outputs: Produces a concise, arbitration-backed decision and a persistent artifact (brief + decision log) suitable for handoff and execution planning.
• Operational guardrails: Includes token-budget awareness, handshake verification, degraded roster fallback, and a stop/halting path to avoid runaway costs.
• Tracker integration via Beads (optional): Converts deferred items into structured tasks when beads are available, improving follow-through after the debate.

Quick Start

Ask your AI to "convene the design council to review this architecture decision and decide the right approach with a binding decision log."