claude-howto/02-subagents

Subagents Examples

This folder contains example subagent configurations for Claude Code.

Installation

Copy these files to your project's .claude/agents/ directory:

cp *.md /path/to/your/project/.claude/agents/

Available Subagents

1. Code Reviewer (code-reviewer)

Comprehensive code quality and maintainability analysis.

Tools: read, grep, diff, lint_runner

Use Case: Automated code reviews focusing on security, performance, and quality.

2. Test Engineer (test-engineer)

Test strategy, coverage analysis, and automated testing.

Tools: read, write, bash, grep

Use Case: Creating comprehensive test suites with high coverage.

3. Documentation Writer (documentation-writer)

Technical documentation, API docs, and user guides.

Tools: read, write, grep

Use Case: Generating and maintaining project documentation.

4. Secure Reviewer (secure-reviewer)

Security-focused code review with minimal permissions.

Tools: read, grep (read-only)

Use Case: Security audits without modification capabilities.

5. Implementation Agent (implementation-agent)

Full implementation capabilities for feature development.

Tools: read, write, bash, grep, edit, glob

Use Case: End-to-end feature implementation.

How Subagents Work

Subagents are specialized AI assistants with:

• Isolated context windows - Clean slate for each task
• Customized system prompts - Specialized expertise
• Tool access control - Only necessary capabilities

Usage Example

Main agent delegates work to subagents:

User: "Build a new authentication feature"

Main Agent:
1. Delegates implementation to implementation-agent
2. Delegates security review to secure-reviewer
3. Delegates testing to test-engineer
4. Delegates documentation to documentation-writer
5. Synthesizes all results

File Structure

project/
├── .claude/
│   └── agents/
│       ├── code-reviewer.md
│       ├── test-engineer.md
│       ├── documentation-writer.md
│       ├── secure-reviewer.md
│       └── implementation-agent.md

When to Use Subagents

Scenario	Use Subagent	Why
Complex feature with many steps	✅ Yes	Separate concerns
Quick code review	❌ No	Unnecessary overhead
Parallel task execution	✅ Yes	Independent contexts
Specialized expertise needed	✅ Yes	Custom prompts
Long-running analysis	✅ Yes	Prevent context exhaustion
Single simple task	❌ No	Adds latency