Council
An agentic pressure-test for questions, decisions, and ideas.
Five advisors with locked thinking styles run as parallel subagents — each in its own isolated context, so none of them sees the others while it writes. They optionally argue across 1–3 debate rounds, then a Chairman subagent (a pure judge, no new arguments) renders the verdict.
Why subagents
If one model writes all five voices in a single context, it knows what the other voices will say while writing each one
[Description truncada. Veja o README completo no GitHub.]