Where are Claude Code agent files stored?

Agent files can be stored globally in ~/.claude/agents/ for use across all projects, or locally in .claude/agents/ for project-specific team sharing.

Can I restrict the tools available to a subagent?

Yes, you can define specific tools in the agent's frontmatter. For example, omitting 'Write' or 'Bash' creates a read-only agent for safer code exploration.

How do I trigger a specific subagent?

Claude invokes subagents via the Task tool using the 'subagent_type' parameter, or you can reference specific agents within your instructions or CLAUDE.md files.

What Claude models can be assigned to subagents?

Subagents can be configured to use Haiku for fast/simple tasks, Sonnet for a balance of speed and intelligence, or Opus for high-level reasoning.

What are the built-in agent types in Claude Code?

Built-in types include 'Explore' for codebase navigation, 'Bash' for command execution, 'Plan' for implementation planning, and 'general-purpose' for multi-step tasks.

Subagents Configuration Reference

Name: Subagents Configuration Reference
Author: Dunc4nJ

byDunc4nJ

•

学習とドキュメンテーション

Provides a comprehensive reference for creating, configuring, and deploying specialized subagents within the Claude Code environment.

概要

The Subagents Reference skill acts as an essential guide for developers looking to extend Claude's capabilities through specialized AI assistants. It provides precise details on agent file structures, frontmatter configurations, and tool restrictions, enabling users to spawn task-specific agents for code reviews, security audits, or test writing. Whether you need to understand built-in agent types like Explore and Plan or create custom agents with restricted tool access, this skill ensures you have the documentation necessary to orchestrate complex multi-agent workflows effectively.

主な機能

Tool restriction and model selection guidelines
Custom agent implementation examples
1 GitHub stars
Comprehensive frontmatter field definitions
Built-in agent type directory (Explore, Bash, Plan)
Standardized agent file format templates

ユースケース

Defining project-specific security review agents
Scaling complex tasks by delegating to specialized subagents
Configuring read-only agents for codebase exploration

概要

主な機能

Tool restriction and model selection guidelines
Custom agent implementation examples
1 GitHub stars
Comprehensive frontmatter field definitions
Built-in agent type directory (Explore, Bash, Plan)
Standardized agent file format templates

ユースケース

Defining project-specific security review agents
Scaling complex tasks by delegating to specialized subagents
Configuring read-only agents for codebase exploration