Multi-Agent OS Agents

Overview

Agent definitions for the multi-agent-os plugin. These define specialized personas that can be spawned by the orchestrator.

Available Agents

Agent	File	Description
orchestrator	orchestrator.md	Master coordinator
sentinel-monitor	sentinel-monitor.md	Anomaly detection
qa-validator	qa-validator.md	Quality assurance
consolidator	consolidator.md	Output synthesis

Agent Categories

Coordination

• orchestrator — Root agent, manages delegation

Observability

• sentinel-monitor — Background monitoring

Quality

• qa-validator — Validation before completion

Synthesis

• consolidator — Merge parallel outputs

Agent Naming Convention

Orchestrator

Claude-Orch-Prime-{YYYYMMDD}-{4-hex}

Sub-Agents

Claude-{Role}-{prime-hex}-{sequence}

Parallel Agents

Claude-{Role}-{prime-hex}-{sequence}{a-z}

Agent Structure

Each agent file uses frontmatter:

---
name: agent-name
description: Brief description
---

# Agent Name

## Identity
[Naming format]

## Purpose
[What the agent does]

## When Invoked
[Trigger conditions]

## Capabilities
[What it can do]

## Output Format
[Expected output structure]

Spawning Agents

Agents are spawned via the Task tool:

Task(subagent_type="general-purpose", task="...")

Note: The frontmatter name is a conceptual identifier. Use actual Task tool types (general-purpose, code-reviewer, etc.) for spawning.

Agent Hierarchy

Orchestrator (root)
├── Sentinel Monitor (background)
├── Sub-Agent 1
│   └── Sub-Sub-Agent 1.1 (max depth 3)
├── Sub-Agent 2
└── Consolidator (synthesis)

---

Part of multi-agent-os plugin v1.0.0