When Subagents Shine
The Core Question
You can create and design subagents — but should you? The whole decision comes down to one thing: whether the intermediate work matters to your main thread. If each step depends on what the previous step discovered, you want that work in your main thread. But if you just need an answer and don't care about the journey, delegate it.
Subagents shine when exploration is separate from execution
The sweet spot is work where you need a result, not a play-by-play; where the exploratory work would clutter your main thread; or where the task benefits from a fresh perspective or a custom system prompt. Three signals, all pointing to delegation.
Research: The Classic Use Case
Research is the textbook subagent task. Consider investigating how authentication works in an unfamiliar codebase. Your main thread needs to know where the JWT is validated — but it doesn't need to see every file searched along the way. A research subagent can read dozens of files, trace function calls, and explore code paths, all in its own context, then return a clean summary:
JWT validation happens in middleware/auth.js line 42,
called from the Express router in route/api.jsResearch is the classic fit: heavy exploration stays in the subagent, the main thread gets just the location it needs.
Code Reviews: Fresh Eyes
Code review is a subtler but powerful use case. Claude reviews code more effectively when the code is presented as if authored by someone else. If you built a feature over many turns with your main thread, asking that same thread to review it often produces weak feedback — Claude was involved in creating it, so it struggles to see it with fresh eyes.
A reviewer subagent sees the changes in a separate context. It runs git diff, reads the modified files, and applies its review criteria without the history of how the code was written. This separation also lets you encode project-specific review standards in the subagent's system prompt, ensuring consistent criteria across your team.
The fresh-context advantage
This is a benefit unique to delegation: the reviewer's isolated context is what gives it 'fresh eyes.' You literally cannot get this from the main thread that wrote the code — the separation is the feature, not a side effect.
Custom System Prompts: A Different Voice
Claude Code's default system prompt emphasizes concise, code-focused responses. Great for coding — wrong for everything else. A subagent can carry a completely different system prompt, making it genuinely better than the main thread for non-coding work.
- •Copywriting subagent — instructions about tone, audience, and style. The default code-focused prompt produces terse technical writing, which is not what you want for a landing page or email campaign.
- •Styling subagent — point it at your design-system files. When it runs, those files load into its context automatically, so it knows your color variables, spacing, and component patterns before writing any CSS.
Next
You've seen where subagents shine. Equally important is knowing where they hurt — the next lesson covers the three anti-patterns to avoid.
Key Takeaways
- ✓The core decision: does the intermediate work matter? If steps depend on each other, keep it in the main thread; if you just need the answer, delegate.
- ✓Subagents shine when you need a result not a play-by-play, when exploration would clutter the main thread, or when a fresh perspective / custom prompt helps.
- ✓Research is the classic fit: a subagent reads dozens of files in isolation and returns a clean summary (e.g. 'JWT validation in auth.js:42').
- ✓Code review benefits from a subagent's fresh, separate context — Claude reviews code better when it didn't author it in the same thread.
- ✓A reviewer subagent's system prompt can encode team-specific review standards for consistency.
- ✓Custom system prompts make subagents better than the main thread for non-coding work (copywriting tone/audience, styling against a design system).
Check Your Understanding
Test what you learned in this lesson.
Q1.What is the single question that decides whether to use a subagent?
Q2.Why does a reviewer subagent often give better feedback than the main thread?
Q3.Why is research a classic subagent use case?
Q4.Why does a custom system prompt make a subagent better for copywriting?
Practice This Lesson