Council

When you need multiple models to weigh in on the same technical decision, this orchestrates a structured investigation with model diversity first, perspective diversity second. Default roster is Opus for correctness and edge cases, GPT for implementation feasibility, and an open-source model as contrarian critic. Each agent gets the same brief but a different angle, then you synthesize their reports into one recommendation. Best for judgment calls where you want evidence from different reasoning systems: architecture tradeoffs, risky merges, incident analysis, or "should we pursue this alternative" questions. It's explicit orchestration, not magic consensus, so you still own the final call.

Install to Claude Code

npx -y skills add warpdotdev/common-skills --skill council --agent claude-code