Specialised subagents in Claude Code
Our verdict
Subagents keep the main session's context clean and give specialised tasks their own instructions and tools. A practical first step into multi-agent setups with almost no code.
- Who it's for
- Claude Code users on larger projects
- Leverage
- Parallel, focused helpers without context bloat
- Effort
- Low
- Date added
What it does
Claude Code lets you define subagents, each with its own system prompt, tools and context window, that the main agent can delegate tasks to.
How builders use it
- Create a code-reviewer subagent with a strict checklist.
- Create a test-runner subagent that only runs and fixes tests.
- Commit subagent definitions to share them.
What to expect
Cleaner sessions and more consistent specialised work.
Copy this: Reviewer subagent prompt
You are a senior code reviewer. Review only the diff on the current branch.
Check: correctness, error handling, security (auth, injection, secrets), tests for new behaviour, and unnecessary changes.
Output: blocking issues, then suggestions, each with file:line. Do not edit files.Related discoveries
- Multi-Step Agent SetupsUse case
Deterministic guardrails with Claude Code hooks
Run your own scripts at agent lifecycle events: format, lint, block risky commands.
Effort: Medium
- Coding & Shipping AgentsUse case
Give your agent a conventions file
A short CONVENTIONS.md stops agents from fighting your style every session.
Effort: Low
- Research & AutomationUse case
Turn a docs site into agent context
Crawl the docs once, save as Markdown, point your coding agent at it.
Effort: Low
- Vibe-Coding & Prompt-to-AppTool
Context7: up-to-date docs for AI editors
Feeds current, version-specific library docs into your coding assistant.
Effort: Low
Want more like this?
Practical discoveries, without the noise.