Subagent Configuration FAQs

Question 1

How does Claude know when to trigger a specific subagent?

Accepted Answer

Claude uses the 'description' field in the YAML frontmatter to determine delegation. By providing a clear, natural language description of the subagent's expertise and trigger phrases, Claude can automatically hand off tasks to the most qualified specialized agent.

Question 2

Where should I save my subagent configuration files?

Accepted Answer

For project-specific subagents that you want to share with your team via version control, save files in '.claude/agents/'. For personal subagents available across all your local projects, save them in the '~/.claude/agents/' directory.

Question 3

How do I control what a subagent is allowed to do?

Accepted Answer

Capabilities are managed through the 'tools' and 'permissionMode' fields. You can whitelist specific tools (like Read, Bash, or WebSearch) and set permissions to 'default' for confirmation prompts, 'acceptEdits' for auto-formatting, or 'bypassPermissions' for fully automated workflows.

Question 4

Which AI models can be assigned to a subagent?

Accepted Answer

You can optimize your workflow by selecting the best model for the task: use 'haiku' for fast and efficient operations, 'sonnet' for a balance of speed and logic, or 'opus' for complex reasoning. You can also set it to 'inherit' to match the main conversation's model.

Question 5

What is the Subagent Configuration skill in Claude Code?

Accepted Answer

Subagent Configuration is a specialized Claude Code capability that allows you to define custom AI personalities. By using YAML frontmatter in Markdown files, you can create agents with specific system prompts, restricted tool access, and optimized model settings tailored for tasks like code review or automated testing.

Subagent Configuration

Subagent Configuration

Key Features

Use Cases

Key Features

Use Cases