agent-development

Create and validate Claude Code plugin agents with frontmatter, system prompts, and trigger scenarios.

16|1|Updated Mar 20, 2026
One-click install
npx skills add https://github.com/SocialGouv/iterion --skill agent-development-socialgouv
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agent-development
Source: https://github.com/SocialGouv/iterion/tree/main/vendor/github.com/SocialGouv/claw-code-go/internal/tools/bundled_skills/plugin-dev/skills/agent-development
Command: npx skills add https://github.com/SocialGouv/iterion --skill agent-development-socialgouv

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve? Writing Claude Code plugin agents that trigger reliably and behave autonomously requires precise frontmatter fields, well-structured system prompts, and carefully worded trigger descriptions, which are easy to get wrong without guidance. ## Core Features & Use Cases - Agent File Authoring: Provides the complete agent markdown format with required frontmatter fields (name, description, model, color, tools) and naming/length validation rules. - System Prompt Patterns: Supplies proven templates for analysis, generation, validation, and orchestration agents, plus triggering best practices with prose-based "When to invoke" scenarios. - Validation & Testing: Includes a validate-agent.sh script that checks frontmatter structure, required fields, and system prompt length, plus guidance for testing trigger conditions. - Use Case: When building a plugin that needs a code review agent, use this Skill to generate the agent file with correct frontmatter, a structured system prompt, and trigger scenarios, then validate it with the bundled script. ## Quick Start Create a Claude Code agent that reviews pull requests for code quality issues, with proper frontmatter and trigger scenarios.

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 system prompt body. You can generate it with AI assistance using the provided creation prompt or write it manually following the templates.

What frontmatter fields are required for a Claude Code agent?▼

Agents require name (lowercase with hyphens, 3-50 characters), description (trigger conditions starting with "Use this agent when"), model (inherit, sonnet, opus, or haiku), and color (blue, cyan, green, yellow, magenta, or red). The tools field is optional for restricting access.

How do I make an agent trigger reliably in Claude Code?▼

Write the description field as flat prose naming 2-4 trigger scenarios, and add a "When to invoke" body section with detailed prose bullets covering both explicit user requests and proactive triggering after relevant work. Avoid quoted user utterances; describe situations instead.

How do I validate a Claude Code agent file?▼

Run the bundled validate-agent.sh script with the agent file path. It checks frontmatter structure, required fields, name format and length, model and color values, and system prompt length, reporting errors and warnings.

What are the limitations of agent system prompts?▼

System prompts should stay between 20 and 10,000 characters, with 500-3,000 characters recommended. Prompts must be written in second person, define clear responsibilities, process steps, and output format, and should not exceed five trigger scenarios.