How does Claude know when to use a specific agent?

Claude uses the 'description' field in the agent's configuration as a trigger. If a task matches that description, Claude can automatically delegate the work to that sub-agent.

What is the difference between project and user agents?

Project agents (.claude/agents/) are specific to a codebase, while user agents (~/.claude/agents/) are global and available across all your projects.

What is a Claude Code sub-agent?

A sub-agent is a specialized instance of Claude with its own system prompt, tool permissions, and context window, designed to handle specific recurring tasks.

How do I share agents with my team?

By creating agents in the .claude/agents/ directory within your project and committing them to version control, your entire team can use the same agent configurations.

Can I restrict which tools a sub-agent can use?

Yes, you can define specific tool access (e.g., Read-only, Bash, or Edit) in the agent's YAML configuration to ensure the principle of least privilege.

Agent Builder

Name: Agent Builder
Author: resolve-io

byresolve-io

•

Productivity & Workflow

Constructs specialized Claude Code sub-agents with custom tool access and domain-specific expertise to streamline complex development workflows.

About

Agent Builder is a foundational skill within the PRISM framework that allows developers to modularize their AI assistance by creating dedicated sub-agents. It solves the problem of context pollution and task ambiguity by defining focused agents for recurring roles like code review, debugging, or data analysis. By configuring specific tool permissions, model selections, and trigger conditions, users can build a library of reusable AI experts that operate within isolated context windows, ensuring more precise and maintainable software development outcomes.

Key Features

Automatic task routing based on agent descriptions
Project-level and global user-level agent scoping
Granular tool permission control for security and focus
Guided sub-agent creation via the /agents interface
Isolated context windows to prevent main conversation pollution
1 GitHub stars

Use Cases

Creating a dedicated debugging agent with full bash access for root cause analysis
Building domain-specific agents for SQL optimization or infrastructure-as-code
Setting up an automated code reviewer that triggers before every git commit

About

Key Features

Automatic task routing based on agent descriptions
Project-level and global user-level agent scoping
Granular tool permission control for security and focus
Guided sub-agent creation via the /agents interface
Isolated context windows to prevent main conversation pollution
1 GitHub stars

Use Cases

Creating a dedicated debugging agent with full bash access for root cause analysis
Building domain-specific agents for SQL optimization or infrastructure-as-code
Setting up an automated code reviewer that triggers before every git commit