Claude Code Agent Development FAQs

Question 1

How do I trigger a custom agent in Claude Code?

Accepted Answer

Agents are triggered based on the 'description' field in the markdown frontmatter. It is best practice to include multiple examples with context and user requests to help Claude understand when to invoke the agent.

Question 2

What model should I use for subagents?

Accepted Answer

The recommended setting is 'inherit', which allows the agent to use the same model as the parent process. However, you can explicitly set it to Sonnet, Opus, or Haiku depending on the task's complexity.

Question 3

Where are agents stored in a project?

Accepted Answer

Custom agents are typically stored in the 'agents/' directory of a plugin or in '.claude/agents/' at the project level for auto-discovery by Claude Code.

Question 4

What is a Claude Code Agent?

Accepted Answer

Agents are autonomous subprocesses in Claude Code designed to handle multi-step tasks independently using specific system prompts, tools, and visual identifiers.

Question 5

Can I restrict which tools an agent can use?

Accepted Answer

Yes, you can use the 'tools' or 'disallowedTools' fields in the agent's frontmatter to specify an array of permitted or blocked tools, ensuring the agent follows the principle of least privilege.

Claude Code Agent Development

关于

主要功能

使用场景

Claude Code Agent Development

关于

主要功能

使用场景