agent-development

Create and validate Claude Code plugin agents with structured frontmatter and system prompts.

2|Updated Mar 21, 2026
One-click install
npx skills add https://github.com/ghoshsam/claude-code-skills --skill agent-development-ghoshsam
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agent-development
Source: https://github.com/ghoshsam/claude-code-skills/tree/main/skills/plugin-dev/agent-development
Command: npx skills add https://github.com/ghoshsam/claude-code-skills --skill agent-development-ghoshsam

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Writing Claude Code plugin agents requires precise frontmatter fields, triggering examples, and well-structured system prompts, and mistakes in any of these cause agents to fail to trigger or behave unpredictably. ## Core Features & Use Cases - Agent File Authoring: Provides the complete Markdown format with YAML frontmatter covering name, description, model, color, and tools fields. - Triggering Design: Guides writing description fields with 2-4 example blocks so agents activate under the right conditions. - System Prompt Templates: Supplies structured templates for responsibilities, analysis processes, output formats, and edge cases. - Use Case: When building a plugin that needs an autonomous code reviewer, use this Skill to generate a valid agent file with correct identifier rules, triggering examples, and a second-person system prompt, then validate it with the included scripts. ## Quick Start Create a Claude Code agent named code-reviewer that triggers on pull request feedback requests and follows the standard frontmatter and system prompt structure.

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 agent for a Claude Code plugin?

Create a Markdown file in the plugin's agents/ directory with YAML frontmatter containing name, description, model, and color fields, followed by a second-person system prompt. The description must include triggering conditions and 2-4 example blocks showing when the agent activates.

What frontmatter fields are required for Claude Code agents?

Agents require name, description, model, and color fields, with tools being optional. The name must be 3-50 lowercase characters with hyphens, and the description must contain triggering conditions plus example blocks.

How do I write a description that triggers an agent correctly?

Start with "Use this agent when" followed by specific conditions, then add 2-4 example blocks containing context, user request, assistant response, and commentary explaining why the agent triggers. Cover different phrasings of the same intent.

Which model should a Claude Code agent use?

Use inherit so the agent runs on the same model as the parent conversation, which is the recommended default. Choose sonnet, opus, or haiku only when the agent has specific capability, speed, or cost requirements.

How do I restrict which tools an agent can access?

Add a tools array to the frontmatter listing only the tool names the agent needs, such as ["Read", "Grep"] for read-only analysis. Omitting the field grants access to all tools, so apply least privilege.

Why is my agent name being rejected as invalid?

Agent identifiers must be 3-50 characters using only lowercase letters, numbers, and hyphens, and must start and end with an alphanumeric character. Underscores, spaces, leading or trailing hyphens, and names under 3 characters are invalid.