Agents

This directory contains AI agents that can be used across all projects via the --plugin-dir flag.

Available Agents

Agent	Purpose
code-reviewer	Git-based PR review with enterprise-grade analysis
task-master	Worktree task execution with scope enforcement
multitasker	Parallel worktree management
agentic-workflow	Multi-agent coordination architecture

Agent File Format

All agents follow the Claude Code standard format with YAML frontmatter:

---
name: agent-identifier
description: Use this agent when [conditions]. Examples:

<example>
Context: [Situation]
user: "[Request]"
assistant: "[Response]"
<commentary>
[Why this agent triggers]
</commentary>
</example>

model: inherit
color: blue
tools: ["Read", "Write", "Bash", "Grep"]
---

[System prompt content]

Required Fields

Field	Description
name	Lowercase identifier with hyphens (e.g., code-reviewer)
description	Starts with "Use this agent when..." with examples
model	Usually inherit to use parent model
color	Visual identifier: blue, green, orange, red, purple

Optional Fields

Field	Description
tools	Array of allowed tools: ["Read", "Write", "Bash", "Grep", "Glob"]

Adding New Agents

1. Create a new .md file in this directory
2. Add YAML frontmatter with required fields
3. Include <example> blocks showing when the agent triggers
4. Write the system prompt content

The agent is immediately available in all projects using this plugin.

Usage

Agents are auto-discovered when using the plugin:

claude --plugin-dir ~/ai

Claude Code will automatically invoke agents based on their description field, or you can reference them explicitly.