Agent teams

The 202 Agent Dev Team: How to Put Four Claude Agents on One Build

How to install a marketplace of specialized Claude Code agents and put four of them on one build

4min read
14numbered steps
7sections

What this actually is

Every serious Claude Code user eventually hits the same wall. You have one model, one context window and one set of instructions, and you keep asking it to be a frontend engineer, a database designer and a security reviewer in the same session. Every extra requirement dilutes the last one, so the output drifts toward average.

wshobson/agents is a marketplace that fixes the shape of that problem rather than the prompt. It ships 202 role-specific agents, each with its own instructions, its own conventions and its own definition of done, generated from a single Markdown source into whatever harness you run.

  • Repo: github.com/wshobson/agents
  • 39.4k stars, and it is free
  • 202 agents across architecture, languages, infrastructure, security, data, machine learning, docs, business and SEO
  • 94 plugins, 183 skills, 105 commands, 16 orchestrators
  • Runs in Claude Code, and also Codex, Cursor, OpenCode, GitHub Copilot and Antigravity

The mistake almost everyone makes first

They try to install all of it. Do not.

The unit of installation is the PLUGIN, not the agent. Skills and agents come along with the plugin, and Claude auto-discovers which one matches your task from the description you type. That discovery is a matching problem, so a bloated install makes it worse, not better: you are handing the selector 202 candidates when your repo only ever needed six.

Install two or three domain plugins that match what you actually build. Add more when a real task has no match, never in advance

Setting it up

  1. Register the catalog once. In Claude Code, run /plugin marketplace add wshobson/agents
  2. Install a domain plugin, not everything. For example /plugin install backend-development
  3. Add one or two more only if your work genuinely spans them. Common honest pairs are backend-development plus security-scanning, or frontend plus unit-testing.
  4. Run /plugin on its own to list every command you now have. This is the discovery step people skip, and it is the only place you see the real names.
  5. Try one command against a real file, not a toy. For example /unit-testing:test-generate src/api/users.py

Free, for your email

Get the full setup steps + the copy-paste setup prompt

The rest of this guide (4 more sections) opens right here. The setup prompt, one paste into Claude that sets it all up, comes by email only. Use your real address.

Unlocks: The two ways to call an agent, and when each is right · Putting four agents on one build · Four things that are true and are not obvious · A first week that actually works

50,000+ follow where these guides come from. No spam, reply “unsubscribe” any time.

Read next