Which model should I specify for my agent?

It is generally recommended to use 'inherit' so the agent uses the same model as the parent session, though you can specify 'sonnet', 'opus', or 'haiku' for specific needs.

Can I limit the tools available to a specific agent?

Yes, you can use the 'tools' field to specify an allowlist of tools or 'disallowedTools' to prevent the agent from invoking specific capabilities or other agents.

How do I ensure my agent triggers at the right time?

Triggering is controlled by the 'description' field in the YAML frontmatter. It should include 'Use this agent when' conditions followed by 2-4 concrete examples with commentary.

What is the difference between a command and an agent in Claude Code?

Commands are for user-initiated actions, while agents are autonomous subprocesses designed to handle complex, multi-step tasks independently based on triggering conditions.

What are agent-level hooks?

Hooks like PreToolUse, PostToolUse, and Stop allow you to run validation, logging, or cleanup tasks specifically scoped to the agent's lifecycle.

Claude Code Agent Development

Name: Claude Code Agent Development
Author: OshanKHZ

byOshanKHZ

•

개발자 도구

Streamlines the creation of autonomous agents for Claude Code plugins using standardized structures and best practices.

소개

The Agent Development skill is an essential resource for developers building extensions for Claude Code. It provides a comprehensive framework for designing autonomous subprocesses, covering everything from YAML frontmatter configuration and identifier validation to sophisticated system prompt engineering. By following these patterns, developers can ensure their agents trigger accurately via well-defined examples, adhere to the principle of least privilege with tool restrictions, and maintain high-quality outputs through structured behavioral boundaries and lifecycle hooks.

주요 기능

Comprehensive guidance for defining triggering conditions with example blocks
3 GitHub stars
Advanced tool management including allowed and disallowed tool configurations
Visual customization options using color-coding and model inheritance
Expert system prompt design patterns for specialized agent personas
Standardized Markdown file structure with YAML frontmatter templates

사용 사례

Building specialized autonomous agents for automated code reviews and testing
Implementing security-constrained agents with limited tool access for sensitive tasks
Designing complex multi-step workflows that require independent subprocesses

소개

주요 기능

Comprehensive guidance for defining triggering conditions with example blocks
3 GitHub stars
Advanced tool management including allowed and disallowed tool configurations
Visual customization options using color-coding and model inheritance
Expert system prompt design patterns for specialized agent personas
Standardized Markdown file structure with YAML frontmatter templates

사용 사례

Building specialized autonomous agents for automated code reviews and testing
Implementing security-constrained agents with limited tool access for sensitive tasks
Designing complex multi-step workflows that require independent subprocesses