How does Claude know when to use a specific sub-agent?

Claude uses the 'description' field in the sub-agent's YAML frontmatter to determine intent. Including 'use proactively' in the description helps Claude delegate tasks automatically.

Where are the created sub-agents stored?

Sub-agents are stored as Markdown files in the `.claude/agents/` directory for project-specific use or the `~/.claude/agents/` directory for global access across all projects.

Can I limit a sub-agent's ability to modify files?

Yes, you can specify exactly which tools a sub-agent can access using the 'tools' field, or set the 'permissionMode' to control how edits are handled.

What are sub-agent hooks used for?

Hooks allow you to trigger specific actions during the agent's lifecycle, such as automatically running a linter or test suite after a sub-agent performs a code edit.

AI Subagent Creator

Name: AI Subagent Creator
Author: GzuPark

byGzuPark

•

Productivity & Workflow

Generates specialized AI sub-agents with custom system prompts and tool permissions to automate domain-specific workflows within Claude Code.

The Subagent Creator skill enables developers to build a modular ecosystem of specialized AI assistants tailored to specific project requirements. By defining custom system prompts, tool access levels, and proactive trigger descriptions in a structured Markdown format, users can create agents for tasks like automated code reviews, security audits, or complex debugging. This skill streamlines the setup and management of these agents, ensuring they are correctly configured with YAML frontmatter and stored in the appropriate project or global directories for seamless multi-agent orchestration.

Key Features

01Automated creation of sub-agent Markdown files with YAML frontmatter.

02Proactive delegation triggers based on specialized agent descriptions.

031 GitHub stars

04Customizable tool access and permission modes for secure execution.

05Hierarchical storage supporting both project-specific and global user agents.

06Support for lifecycle hooks to run commands before or after tool use.

Use Cases

01Setting up specialized documentation assistants that follow specific organizational style guides.

02Building a dedicated code reviewer that triggers automatically after file modifications.

03Creating a security auditor sub-agent with restricted read-only access for scanning vulnerabilities.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add gzupark/claude-plugin-pack subagent-creator

For use in Claude.ai and ChatGPT

Download Skill