agent-builder

Create custom Claude Code sub-agents with YAML frontmatter and controlled tool access.

Updated Aug 23, 2026
One-click install
npx skills add https://github.com/resolve-io/_prism --skill agent-builder-resolve-io
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agent-builder
Source: https://github.com/resolve-io/_prism/tree/main/skills/agent-builder
Command: npx skills add https://github.com/resolve-io/_prism --skill agent-builder-resolve-io

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill guides you to create custom Claude Code sub-agents with specialized expertise, enabling reuse, controlled access, and isolation from the main conversation for focused tasks.

Core Features & Use Cases

  • Specialization: Create agents with targeted domains (e.g., code review, debugging, data analysis).
  • Isolation & Reuse: Separate context windows and reuse agents across projects.
  • Controlled Access: Granular tool access and model selection per agent.
  • Templates & Outputs: Generate agent files with YAML frontmatter and system prompts.

Quick Start

Use the built-in agent creator: run /agents, choose Create New Agent, then save to .claude/agents/ or ~/.claude/agents/. Then customize and test.

Frequently Asked Questions about agent-builder

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

FAQPage Schema
How do I create a custom sub-agent with specialized expertise in Claude Code?

Create a sub-agent by running /agents, selecting Create New Agent, then configuring YAML frontmatter with name, description, and optional tools and model fields. Save to .claude/agents/ or ~/.claude/agents/, customize the system prompt, and test to isolate focused tasks with controlled permissions.

What's the difference between running tasks in the main conversation versus using a sub-agent?

Sub-agents provide isolated context windows, separate tool access, and reusability across projects while preserving main conversation state. This enables specialization—code review, debugging, data analysis—without mixing contexts or losing conversation history.

Can I control which tools a sub-agent can access?

Yes. Sub-agents support granular tool access and model selection per agent through YAML configuration. Define tools in the agent's frontmatter to restrict capabilities to the specific workflow, preventing unintended tool usage.

How do I reuse a sub-agent across multiple projects?

Save agents to ~/.claude/agents/ for user-level availability across all projects, or .claude/agents/ for project-specific agents. Agent hierarchies and templates enable reuse while maintaining separate instances and permissions per deployment context.

What happens to my main conversation when I delegate work to a sub-agent?

The main conversation state is preserved. Sub-agents operate in isolated context windows, allowing you to delegate specialized tasks like code review or data analysis without disrupting your primary workflow or losing conversation history.

Do I need prerequisites or dependencies installed to build sub-agents?

No external dependencies are required. Sub-agents are built using Claude Code's built-in agent creator with YAML frontmatter configuration. You only need Claude Code running and access to the .claude/agents/ directory structure.