agent-conversation-protocols

What problem does it solve?

Multi-agent systems often devolve into chaotic, inefficient, or incorrect interactions because there is no clear protocol for turn-taking, delegation, conflict resolution, or termination; this Skill provides structured conversation designs that make agent interaction predictable, safe, and efficient.

Core Features & Use Cases

• Pattern selection guidance: Decision trees to pick between request/response, fan-out/fan-in, supervisor-worker, debate, critique, blackboard, and voting patterns based on task characteristics.
• Topology and termination rules: Maps topology × initiative × turn-order choices and explicit termination/gather policies to common workflows like code review, research synthesis, and dynamic routing.
• Failure mode detection & mitigation: Concrete detectors and fixes for delegation loops, sycophancy collapse, supervisor bottlenecks, context degradation, and blackboard bloat.
• Use Case: Design a 4-agent code review protocol that enforces security-first blocking review, parallel performance and style checks, and a debate phase to resolve conflicting recommendations.

Quick Start

Create a conversation protocol for a 4-agent code review that runs a blocking security review first, then parallel performance and style reviews, and a final debate-based conflict resolution step.