Author(s): Rick Hightower

Originally published on Towards AI.

Part 6 Subagents: Subagents are the Claude Code feature most engineers never touch, and skipping them is why long sessions slowly fall apart.

Your best Claude Code session quietly rots all afternoon, and you blame the model when the fix was one habit away. Claude Code subagents are the single highest-leverage habit separating casual users from experts. Most engineers never touch them. Here is what that costs you, and how to fix it in one afternoon.

After the lead, the article explains how Claude Code degrades over long sessions due to context pollution: file reads from exploratory detours accumulate in your main conversation and make answers fuzzier, even though the model isn’t “getting worse.” It then breaks down what a subagent is—an isolated context window with its own system prompt and tool restrictions—so the parent only receives the final summary rather than all intermediate tool calls. The piece covers the cost/benefit math (token exploration stays in the subagent), the key mechanics and constraints (no nesting, parent sees only outputs, some shared context like MCP/skills), and the three built-in subagents you can use immediately (Explore, Plan, and general-purpose) along with when each makes sense. It shows how to invoke subagents on purpose via prompt naming or delegation (including headless usage), how to create custom subagents using YAML/frontmatter (or the /agents UI), when to choose subagents versus worktrees or agent teams, and concludes with practical “do this today” steps—explicitly delegate with Explore, browse/create agents, and build one reusable custom subagent—arguing that subagents become a compounding habit that keeps your main conversation clean and your engineering judgment reusable.

Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI