Where should I store my custom Claude agents?

Store agents in '.claude/agents/' for project-specific team workflows shared via Git, or in '~/.claude/agents/' for personal agents available across all your local projects.

What is a Claude Code Agent?

Agents are specialized AI assistants that run in separate context windows within Claude Code, enabling focused, autonomous task execution without cluttering your main conversation history.

Can I limit the commands an agent can run?

Yes, you can define granular tool access in the YAML configuration, such as restricting Bash to specific commands like 'npm test' or 'git diff' to ensure safe execution.

How does Claude know when to use a specific agent?

Claude uses the description field in the agent's YAML frontmatter to determine when to trigger it. High-quality descriptions include roles, expertise, and specific trigger terms like 'review', 'debug', or 'audit'.

Claude Code Agent Builder

Name: Claude Code Agent Builder
Author: mike-coulbourn

bymike-coulbourn

•

Productivity & Workflow

Simplifies the creation and optimization of specialized, context-aware AI agents for autonomous Claude Code workflows.

The Agent Builder skill provides a comprehensive framework for designing and deploying custom AI assistants within the Claude Code environment. By guiding users through structured YAML configuration, granular tool scoping, and strategic system prompt engineering, it ensures that custom agents are highly discoverable and capable of executing complex tasks like security audits, automated testing, and deep architectural reviews in isolated context windows. This skill is essential for teams looking to standardize development workflows or individuals wanting to create persistent, domain-specific AI experts for their local environment.

Key Features

01YAML frontmatter generation for agent metadata and discovery optimization.

023 GitHub stars

03Support for both project-level (shared) and personal (global) agent scopes.

04Granular tool permission configuration including scoped Bash command access.

05Strategic system prompt templates for self-contained, multi-step workflows.

06Advanced discovery tuning using proactive trigger terms and role definitions.

Use Cases

01Standardizing team-wide code review and quality gates via Git-shared agents.

02Creating task-specific agents for automated debugging and regression testing.

03Building a dedicated security auditor agent for continuous vulnerability scanning.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add mike-coulbourn/claude-vibes agent-builder

For use in Claude.ai and ChatGPT

Download Skill