What is a Claude Code subagent?

A subagent is a specialized Claude assistant configured with a specific role, system prompt, and set of tools to perform dedicated tasks, such as code reviews, test running, or API design.

Can I restrict what tools a subagent can access?

Yes, the creation process includes a Capability Phase where you can define tool access strategies like 'Read-only', 'Research only', or manually whitelist/blacklist specific tools and MCP servers.

Where are the created subagent files stored?

You can choose to store agents at the project level (.claude/agents/), user level (~/.claude/agents/), or within a specific plugin directory for distribution.

How do I invoke a subagent once it's created?

After creation, you can verify your agent using the /agents command in Claude Code. It can then be triggered by name or proactively when Claude recognizes a task matching the agent's description.

Claude Subagent Creator

Name: Claude Subagent Creator
Author: filipowm

byfilipowm

0•

生産性とワークフロー

Guides users through a structured, 8-phase interactive process to design and build high-quality custom Claude Code subagents.

The create-agent skill transforms Claude into a specialized subagent architect, streamlining the creation of custom AI assistants within the Claude Code ecosystem. It leads users through a comprehensive design workflow—from identity and behavioral traits to complex technical configurations like scope management (project, user, or plugin levels), tool access strategies, and runtime permissions. By standardizing the design process, it ensures every custom agent follows best practices for prompt engineering and capability isolation, resulting in robust, specialized tools tailored to your specific development needs.

主な機能

010 GitHub stars

02Multi-level scope management (Project, User, and Plugin levels)

03Advanced runtime settings including memory, model selection, and lifecycle hooks

04Granular tool access and permission configuration

05Automated markdown file generation with optimized system prompts

06Structured 8-phase interactive design workflow

ユースケース

01Designing and packaging subagents as part of larger Claude plugins

02Building global productivity agents available across all your development environments

03Creating specialized project-level agents for specific code review or testing tasks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add filipowm/claude-plugins create-agent

For use in Claude.ai and ChatGPT

主な機能

010 GitHub stars

02Multi-level scope management (Project, User, and Plugin levels)

03Advanced runtime settings including memory, model selection, and lifecycle hooks

04Granular tool access and permission configuration

05Automated markdown file generation with optimized system prompts

06Structured 8-phase interactive design workflow

ユースケース

01Designing and packaging subagents as part of larger Claude plugins

02Building global productivity agents available across all your development environments

03Creating specialized project-level agents for specific code review or testing tasks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add filipowm/claude-plugins create-agent

For use in Claude.ai and ChatGPT