Can I add project-specific rules to these agents?

Yes, the agent instructions allow you to define custom domain knowledge, rules, and response formats specific to your codebase's conventions.

Where are these agent definitions stored?

Agents are stored as markdown files within the .claude/agents/ directory of your repository for easy management and version control.

How does the main Claude instance know when to use an agent?

The orchestrator uses the 'description' field in the agent's markdown frontmatter, which includes specific context examples, to determine when a task should be delegated.

Which Claude models can I use for agents?

You can specify Claude 3.5 Haiku for fast, mechanical tasks, Sonnet for balanced performance, or Opus for tasks requiring the highest level of reasoning.

What is a Claude Code sub-agent?

A sub-agent is a specialized process defined within your project that handles specific tasks like searching, testing, or reading documentation in isolation from the main chat context.

Claude Code Agent Creator

Name: Claude Code Agent Creator
Author: hejhi

byhejhi

开发者工具

Generates specialized sub-agents for Claude Code to handle isolated tasks like codebase exploration, verification, and implementation.

关于

The Create Agent skill empowers developers to extend Claude Code's capabilities by defining custom sub-agents tailored for specific workflows. By creating structured markdown files in the .claude/agents directory, users can delegate noisy or repetitive tasks—such as API lookups, testing, or deep file reads—to specialized processes. This architectural approach keeps the main conversation context clean, leverages different Claude models (Haiku, Sonnet, Opus) for cost and speed efficiency, and ensures consistent implementation of project-specific patterns and domain knowledge.

主要功能

Context isolation to prevent main-thread noise during complex tasks
Domain-specific instruction sets for project-specific patterns
Model-specific configuration for cost and performance optimization
0 GitHub stars
Standardized markdown templates for rapid agent definition
Predictable response formatting for reliable tool orchestration

使用场景

Building API explorers to search and extract documentation from external libraries
Automating repetitive codebase exploration and deep-search tasks
Creating specialized verifiers to execute and report on complex test suites

关于

主要功能

Context isolation to prevent main-thread noise during complex tasks
Domain-specific instruction sets for project-specific patterns
Model-specific configuration for cost and performance optimization
0 GitHub stars
Standardized markdown templates for rapid agent definition
Predictable response formatting for reliable tool orchestration

使用场景

Building API explorers to search and extract documentation from external libraries
Automating repetitive codebase exploration and deep-search tasks
Creating specialized verifiers to execute and report on complex test suites