Use when a problem deserves two independent AI solution colleagues exploring alternatives together — the invoking agent coordinates a bounded multi-round discussion between two user-configured harness/model pairs and writes a final decision dossier.
