Where should I store agent configuration files?

Agent configurations should be saved as .md files within the 'agents/' directory of your plugin for automatic discovery and namespacing.

What is the best model to use for a sub-agent?

The recommended setting is 'inherit', which allows the agent to use the same model as the parent process, though you can specify 'sonnet', 'opus', or 'haiku' for specific needs.

How are agents triggered in Claude Code?

Agents are triggered via the 'description' field in their markdown frontmatter, which contains specific conditions and multiple usage examples to help Claude identify the right context.

What is a Claude Code Agent?

An agent is an autonomous subprocess within Claude Code that handles complex, multi-step tasks independently using its own system prompt and toolset.

Can I restrict what tools an agent can access?

Yes, the 'tools' field in the agent frontmatter allows you to define an array of specific tools, ensuring the agent follows the principle of least privilege.

Claude Agent Development

Name: Claude Agent Development
Author: arlenagreer

byarlenagreer

0•

開発者ツール

Provides a comprehensive framework and standardized patterns for building autonomous sub-agents within Claude Code plugins.

This skill serves as a definitive guide for developers looking to extend Claude Code's capabilities through autonomous agents. It provides structured templates and best practices for defining agent frontmatter, crafting precise triggering conditions using example-driven descriptions, and designing robust system prompts. By implementing these patterns, developers can create specialized, self-sufficient subprocesses that handle complex multi-step tasks—such as code reviews, security audits, or documentation generation—while maintaining clear visual organization and tool access control within their development environment.

主な機能

010 GitHub stars

02Standardized YAML frontmatter templates for agent metadata

03Advanced triggering logic using structured <example> blocks

04Step-by-step system prompt design patterns for autonomous behavior

05Model inheritance and UI color customization guidelines

06Principle-of-least-privilege tool access configuration

ユースケース

01Building specialized code reviewers that trigger automatically on specific file changes

02Developing multi-step documentation generators that run independently of the main thread

03Creating autonomous security analyzers with restricted tool access for safe audits

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add arlenagreer/claude_configuration_docs agent-development

For use in Claude.ai and ChatGPT

Download Skill