sub-agent-patterns

SKILL.md

Delegate specialized tasks to isolated AI assistants with custom tools, models, and system prompts.

Sub-agents preserve main context by isolating verbose tool outputs and intermediate reasoning, enabling longer sessions and cleaner conversations

Three built-in agents available: Explore (Haiku, read-only codebase search), Plan (Sonnet, plan-mode research), and General-Purpose (Sonnet, full read/write access)

Create custom agents in .claude/agents/ with YAML frontmatter and markdown prompts; omit the tools field entirely to inherit all tools including MCP integrations

Orchestrate multi-phase workflows by adding the Task tool to an agent's configuration, enabling parallel execution and synthesis across specialists

Avoid Bash approval spam by removing Bash from tools lists unless the agent genuinely needs CLI execution; default to Write/Edit tools for file operations

Sub-Agents in Claude Code

Status: Production Ready ✅
Last Updated: 2026-02-02
Source: https://code.claude.com/docs/en/sub-agents

Sub-agents are specialized AI assistants that Claude Code can delegate tasks to. Each sub-agent has its own context window, configurable tools, and custom system prompt.

Why Use Sub-Agents: Context Hygiene

The primary value of sub-agents isn't specialization—it's keeping your main context clean.