How do I restrict a subagent's tool access?

You can define a 'tools' field in the subagent's YAML frontmatter to grant only necessary permissions, such as 'Read' only or specific 'Bash' command patterns.

Where should I save my custom Claude subagents?

Project-specific agents should be saved in .claude/agents/my-agent.md, while personal agents for individual use belong in ~/.claude/agents/my-agent.md.

What is the difference between a subagent and a skill?

Subagents are specialized assistants with their own system prompts and focused expertise for complex workflows, while skills are context-aware capabilities that enhance Claude's core behavior automatically.

Which model should I use for my subagent?

Sonnet is recommended for most tasks, Opus for complex architectural reasoning, and Haiku for fast, low-latency tasks like formatting or simple validations.

Claude Subagent Builder

Name: Claude Subagent Builder
Author: pr-pm

bypr-pm

•

개발자 도구

Provides expert guidance and implementation patterns for creating, configuring, and troubleshooting specialized Claude Code subagents.

소개

The Agent Builder skill empowers developers to extend Claude Code's capabilities by creating specialized subagents tailored for specific domains like code review, debugging, or data analysis. It provides structured frameworks for system prompt engineering, tool access configuration, and model selection, ensuring that your custom AI assistants are focused, efficient, and reliable. Whether you are troubleshooting delegation patterns or designing a multi-agent workflow, this skill offers the implementation patterns and best practices needed to build high-performance subagents within the Claude ecosystem.

주요 기능

64 GitHub stars
Standardized subagent file structure and deployment patterns
Model selection strategies (Sonnet vs. Opus vs. Haiku)
Expert system prompt engineering for specialized AI roles
Troubleshooting for agent invocation and delegation issues
Granular tool access and permission configuration

사용 사례

Designing multi-agent workflows for complex data analysis tasks
Configuring a dedicated debugging agent with restricted bash access
Creating a specialized security-focused code reviewer subagent

소개

주요 기능

64 GitHub stars
Standardized subagent file structure and deployment patterns
Model selection strategies (Sonnet vs. Opus vs. Haiku)
Expert system prompt engineering for specialized AI roles
Troubleshooting for agent invocation and delegation issues
Granular tool access and permission configuration

사용 사례

Designing multi-agent workflows for complex data analysis tasks
Configuring a dedicated debugging agent with restricted bash access
Creating a specialized security-focused code reviewer subagent