create-subagents

Create and configure Claude Code subagents with XML system prompts and tool inheritance.

3|Updated Nov 28, 2025
One-click install
npx skills add https://github.com/rayk/lucid-toolkit --skill create-subagents-rayk
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: create-subagents
Source: https://github.com/rayk/lucid-toolkit/tree/main/.claude/skills/sub-agent-builder
Command: npx skills add https://github.com/rayk/lucid-toolkit --skill create-subagents-rayk

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Complex tasks often require specialized expertise and isolated execution, which a single AI agent might struggle with. This skill empowers you to design, build, and orchestrate focused subagents, delegating specific parts of a workflow to autonomous, non-interactive AI instances.

Core Features & Use Cases

  • Subagent Creation: Learn to define subagents with focused roles, limited tool access, and custom system prompts.
  • Multi-Agent Orchestration: Understand how to use the Task tool to launch and manage specialized subagents for research, code generation, or analysis.
  • Use Case: For a software development project, you can create a code-reviewer subagent to automatically audit code changes for quality and security, while a doc-generator subagent handles documentation updates, all orchestrated by your main Claude conversation.

Quick Start

  1. Run /agents command
  2. Select "Create New Agent"
  3. Choose project-level (agents/) or user-level (~/agents/)
  4. Define the subagent:
    • name: lowercase-with-hyphens
    • description: When should this subagent be used?
    • tools: Optional comma-separated list (inherits all if omitted)
    • model: Optional (sonnet, opus, haiku, or inherit)
  5. Write the system prompt (the subagent's instructions)

Frequently Asked Questions about create-subagents

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create specialized AI subagents to delegate different tasks in my workflow?▼

Subagents are focused AI instances you configure with custom system prompts and limited tool access to handle specific responsibilities. Use the `/agents` command, select project-level or user-level storage, define a lowercase-hyphen name, choose tools and model, then write the system prompt. Each subagent runs autonomously within your orchestrated workflow.

What's the difference between creating a single agent versus multiple subagents for complex tasks?▼

A single agent handles all tasks with one instruction set; subagents split responsibilities by expertise and tool scope. Multi-agent orchestration isolates failures, improves focus, and lets you delegate specialized work—like code review or documentation—to independent instances launched via the Task tool without blocking your main conversation.

Can I use subagents to automate code review and documentation in the same project?▼

Yes. Create separate subagents with focused roles: a `code-reviewer` with code analysis tools and a `doc-generator` with documentation tools. Your main agent orchestrates both using the Task tool, delegating code audits and documentation updates to their respective specialists while coordinating outputs.

How do I configure tool access and model selection for each subagent?▼

During subagent creation, specify tools as an optional comma-separated list; omit the list to inherit all tools. Select model explicitly (`sonnet`, `opus`, `haiku`) or use `inherit` to match your parent agent. Tool inheritance and model choice control what capabilities and compute each subagent receives.

What happens if I don't specify tools or a model when creating a subagent?▼

Without explicit tools, the subagent inherits all available tools from its parent context. Without a model specification, it inherits the parent's model. Defaults ensure the subagent runs immediately but with broader scope; specify both to enforce isolation and cost control.

Do subagents work with project-level and user-level storage, and what's the difference?▼

Yes, subagents store at both scopes: project-level (`agents/`) isolates agents to a single project; user-level (`~/agents/`) makes agents reusable across projects. Choose project-level for task-specific specialists and user-level for general-purpose subagents you reference from multiple workflows.