Agent Design FAQs

Question 1

What is the primary purpose of the Agent Design skill?

Accepted Answer

The Agent Design skill is used to define and configure 'AgentProfiles'. It allows you to programmatically set an agent's identity, system prompts, memory files, available tools, and complex orchestration logic for specialized AI tasks.

Question 2

How do subagents work within this skill?

Accepted Answer

Subagents allow the main agent to delegate specific tasks to specialized model instances (like Sonnet or Haiku). You can define unique prompts and tool constraints for each subagent, enabling complex multi-step workflows like code reviews or test generation.

Question 3

Can I integrate third-party tools using the Agent Design skill?

Accepted Answer

Yes. This skill natively supports the Model Context Protocol (MCP), allowing you to connect both bundled local servers and external MCP servers (like GitHub, Slack, or Filesystem) to extend the agent's capabilities.

Question 4

What is the difference between a systemPrompt and an agentMDFile?

Accepted Answer

The systemPrompt defines the agent's core personality and behavioral 'DNA,' while the agentMDFile (like CLAUDE.md) provides project-specific context, tech stack conventions, and workspace knowledge that acts as the agent's contextual memory.

Question 5

When should I use custom Agent Commands?

Accepted Answer

Use custom commands when you need to trigger specific, repeatable workflow patterns. By defining a command like '/review-pr', you can orchestrate a sequence of skills, subagents, and tools to run automatically via a single prompt.

Agent Design

Agent Design

主な機能

ユースケース

主な機能

ユースケース