Where are custom agents stored?

They are stored as markdown files in .claude/agents/ for project-specific use or ~/.claude/agents/ for global user-level use across multiple projects.

What is a Claude Code sub-agent?

A sub-agent is a specialized AI instance within Claude Code configured with specific instructions, tools, and a limited context to handle focused tasks without polluting the main conversation.

How do I trigger a custom agent?

Agents can be triggered automatically when Claude detects a task matching the agent's description, or explicitly by saying 'Use the [agent-name] agent to [task]'.

Can I control which tools an agent can access?

Yes, you can specify exactly which tools (such as Bash, Read, Edit, or external MCP tools) an agent is permitted to use via the tools field in the agent's YAML configuration.

What is the benefit of context isolation?

Context isolation ensures that the specialized agent only sees relevant information for its task, preventing the main conversation window from hitting token limits or becoming confused by irrelevant data.

Sub-Agent Builder

Name: Sub-Agent Builder
Author: resolve-io

byresolve-io

•

Productivity & Workflow

Empowers developers to create and manage specialized AI sub-agents with custom tool permissions and domain-specific expertise within Claude Code.

About

The Sub-Agent Builder skill modularizes complex software engineering tasks by enabling the creation of focused AI sub-agents. It provides a structured framework to define specialized roles—such as code reviewers, debuggers, or data analysts—with granular control over tool access (Bash, File Operations, MCP) and isolated context windows. By automating the generation of markdown-based configuration files and system prompts, it ensures that these agents are triggered intelligently based on task descriptions, facilitating a more predictable, scalable, and maintainable AI-assisted development workflow.

Key Features

Context isolation to prevent main conversation pollution
Intelligent routing based on descriptive trigger conditions
Automated agent configuration with YAML frontmatter
Granular tool permission management for security and focus
Hierarchical scoping for project-level or global user-level agents
1 GitHub stars

Use Cases

Deploying a dedicated 'Security Auditor' agent to scan code changes for vulnerabilities before every commit
Creating a 'Documentation Expert' that strictly adheres to specific style guides and organizational standards
Building a 'DevOps specialist' with restricted tool access to debug CI/CD pipeline failures and suggest infrastructure fixes

About

Key Features

Context isolation to prevent main conversation pollution
Intelligent routing based on descriptive trigger conditions
Automated agent configuration with YAML frontmatter
Granular tool permission management for security and focus
Hierarchical scoping for project-level or global user-level agents
1 GitHub stars

Use Cases

Deploying a dedicated 'Security Auditor' agent to scan code changes for vulnerabilities before every commit
Creating a 'Documentation Expert' that strictly adheres to specific style guides and organizational standards
Building a 'DevOps specialist' with restricted tool access to debug CI/CD pipeline failures and suggest infrastructure fixes