agent-development

Create autonomous agents for Claude Code plugins using markdown files with YAML frontmatter.

Updated Dec 21, 2025
One-click install
npx skills add https://github.com/s-hiraoku/claude-code-harnesses-factory --skill agent-development-s-hiraoku
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agent-development
Source: https://github.com/s-hiraoku/claude-code-harnesses-factory/tree/main/.claude/agents/agent-development
Command: npx skills add https://github.com/s-hiraoku/claude-code-harnesses-factory --skill agent-development-s-hiraoku

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill guides creating autonomous agents for Claude Code plugins using markdown files with YAML frontmatter.

Core Features & Use Cases

  • Agent File Structure: Markdown with YAML frontmatter and system prompt body.
  • System Prompt Design: Second-person language, structured process, output format, quality standards.
  • Best Practices & Boundaries: Least privilege, color consistency, testing triggers, clear boundaries.

Quick Start

Create a new agent file named my-agent.md in the project scope and specify name, description, and system prompt.

Frequently Asked Questions about agent-development

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create an autonomous agent for Claude Code plugins?

Create autonomous agents by writing markdown files with YAML frontmatter. Define required fields (name, description) and optional fields (model, color, tools), then write a second-person system prompt that specifies behavior, output format, and boundaries for your plugin integration.

What structure should my agent markdown file follow?

Agent files use markdown with YAML frontmatter at the top containing metadata, followed by the system prompt in the body. Frontmatter must include name and description; optional fields include model, color, and tools. The system prompt uses second-person language and defines processing logic.

How do I write an effective system prompt for agent behavior?

Write system prompts in second person that clearly define the agent's task, expected input formats, output structure, and quality standards. Structure the prompt as a logical sequence of instructions, apply least-privilege principles, and test triggers to verify the agent behaves consistently during plugin deployment.

What are the best practices for agent design and boundaries?

Enforce least-privilege access, maintain consistent color schemes, define clear operational boundaries, and establish testing triggers before deployment. Document prerequisites and constraints so plugin consumers understand scope limitations and expected agent behavior.

Can I specify which Claude model version my agent should use?

Yes, the optional model field in YAML frontmatter lets you specify the Claude model version. Include this in your frontmatter along with other optional fields like color and tools to customize agent behavior and appearance for plugin integration.

Do I need to configure tools and colors for my agent?

Tools and colors are optional fields in frontmatter. Define tools if your agent needs access to external functions, and specify color for visual consistency in the Claude Code plugin interface. Both are configuration best practices for complete agent setup.