Create Subagent FAQs

Question 1

Can subagents run tasks in parallel?

Accepted Answer

Yes. By using the 'Task' tool invocation approach, you can dispatch multiple subagents simultaneously. This is ideal for 'Divergent Exploration'—where multiple agents investigate the same problem from different angles—or for executing independent tasks at the same time.

Question 2

When should I use a subagent instead of a regular skill?

Accepted Answer

You should use a subagent when a task requires a specific persona (like a Code Reviewer or Debugger), focused instructions that benefit from context isolation, or unique tool permissions that differ from your main session. Use skills when you simply want to extend Claude's knowledge without needing a separate context.

Question 3

What is a Claude Code subagent?

Accepted Answer

A subagent is a specialized AI assistant that Claude Code can delegate specific tasks to. Each subagent operates within its own context window, preventing the main conversation from becoming cluttered while utilizing a custom system prompt and specific tools tailored to its expertise.

Question 4

What are subagent permission modes?

Accepted Answer

Permission modes define how much autonomy a subagent has. 'Default' requires standard prompts for actions, 'read-only' limits the agent to analysis without file modifications, and 'bypassPermissions' allows the agent to execute tasks automatically without user interruption.

Question 5

How do I create a persistent subagent?

Accepted Answer

Persistent subagents are created by defining a Markdown file in the `.claude/agents/` directory. These files use frontmatter to configure the agent's name, description, model, and tools, followed by a system prompt that defines the agent's role and workflow.

Create Subagent

Create Subagent

主な機能

ユースケース

主な機能

ユースケース