quinte
"QUINTE — Five-agent structured debate architecture. Hermes+cc+cw+Reasonix+oh-my-pi, all DeepSeek v4-pro max reasoning. R1 independent analysis → R2 cross-review → R3 final verdict."
Resources
9Install
npx skillscat add eric-stone-plus/quinte Install via the SkillsCat registry.
SKILL.md
QUINTE — Multi-Agent Debate Architecture
Five independent AI agents debate questions through structured rounds: independent analysis, adversarial cross-review, and final verdict. Dramatically higher confidence than any single model.
Architecture
| Agent | Engine | R1 | R2 | Role |
|---|---|---|---|---|
| Hermes | v4-pro · xhigh | ✅ | ✅ | Orchestrator + final verdict |
| Claude Code | v4-pro · max | ✅ | ✅ | Broadest coverage, structured reports |
| CodeWhale | v4-pro · max | ✅ | ✅ | Deepest research, concurrency analysis |
| Reasonix | v4-pro · max | — | ✅ | R2 pure reasoning judge |
| oh-my-pi | v4-pro · xhigh | ⚡ | ✅ | Hot spare (cc timeout 180s) |
Invocation
All agents via terminal(background=true), direct CLI. No delegate_task.
Built With
License
MIT — protocol and orchestration. Individual agents carry their own licenses.