Multi-Agent Orchestration System Design

# Multi-Agent Orchestration System Design

You are a senior AI systems architect designing a multi-agent orchestration system for [ORGANIZATION_NAME] to handle [WORKFLOW_NAME] end-to-end.

## System Architecture

Define the full agent roster with roles, system prompts, input/output formats, and error handling for each:

**Orchestrator Agent:**
- Role: Receive tasks, delegate to specialists, aggregate results, manage failures
- System prompt template:
- Input/output schema:

**Specialist Agent 1 — [SPECIALIST_1_ROLE]:**
- Scope and responsibilities:
- System prompt:
- What triggers a retry vs. escalation:

**Specialist Agent 2 — [SPECIALIST_2_ROLE]:**
[Same structure]

**Reviewer Agent:**
- Role: Quality-gate all outputs before delivery
- Evaluation criteria for [WORKFLOW_NAME]:
- Rejection threshold and what happens on rejection:

## Inter-Agent Communication Protocol

Define the message format (JSON schema) passed between agents, including:
- Task ID and parent task reference
- Payload structure
- Status codes
- Confidence or quality scores

## Error Handling Framework

Design recovery logic for:
1. Agent timeout → fallback behavior
2. Conflicting outputs from two agents → arbitration rule
3. External API failure → degraded mode
4. Human escalation needed → notification format and context package

## Orchestrator Prompt

Write the full master orchestrator system prompt for [WORKFLOW_NAME].

## Observability

Define logging, tracing, and alerting requirements for a production deployment.