Sub-Agents: The Feature Nobody Uses (But Should)

Quick poll: do you know Claude Code can spawn sub-agents? Agents that run in parallel, each focused on a specific task, reporting back to a main orchestrator?

If not, you're in the majority. And you're leaving serious productivity on the table.

What Are Sub-Agents?

A sub-agent is a focused Claude Code instance that handles one specific job. Think of them as specialists you delegate to:

• A code reviewer that checks your changes
• A test writer that generates test cases
• A security scanner that audits for vulnerabilities
• A documentation writer that updates docs

You define them in agent files, give them specific instructions, and let them run.

The Delegation Pattern

Here's the workflow that changed my game:

1. Write the feature code yourself (or with Claude's help)
2. Kick off 3 sub-agents in parallel: code review, test generation, docs update
3. Each agent focuses on its specialty and writes output
4. You review the combined results

What used to take 45 minutes of switching between tasks now takes 10 minutes of parallel execution.

Getting Started

Start simple. Create one agent file for code review. Put it in .claude/agents/code-reviewer.md. Give it clear instructions about what to check. Use it after every feature.

Then expand. Add a test writer. A security checker. A refactoring assistant. Each agent compounds the value of the others.

Read the full agents guide to set up your first sub-agent in under 5 minutes.