Agent SDK App

claude-code-agent-teams-exercises

by panaversity

Claude Code Agent Teams exercises — master multi-agent orchestration through 11 real business challenges

A comprehensive set of practical exercises for developers learning Claude Code's Agent Teams feature. This package provides 11 hands-on and design-based projects to master multi-agent coordination, task pipelines, and communication protocols. It's designed to teach agent team thinking and execution through real-world business problem-solving scenarios.

View on GitHub ↗

Key features

8 hands-on exercises for running real agent teams with business data
3 capstone projects for complex multi-agent problem-solving
Design exercises to plan team architectures without API costs
Covers task pipelines, communication, and quality gate workflows
Includes a universal 4-criteria scoring rubric for evaluation

Top contributors

mjunaidca3

README

View on GitHub ↗

Claude Code Agent Teams: Business Problem-Solving Exercises

By Panaversity -- Master Multi-Agent Orchestration Through Real Business Challenges

Welcome! This package contains 8 exercises + 3 capstone projects (11 total) for learning Claude Code's Agent Teams feature through real-world business problem solving. You'll create research teams, build task pipelines for event planning, run feature prioritization debates, and manage quality review cycles -- all using multi-agent coordination.

Package Structure

agent-teams-exercises/
├── README.md                                          ← You are here
├── EXERCISE-GUIDE.md                                  ← Full guide with budget paths
├── scoring-rubric.md                                  ← Universal 4-criteria rubric
├── module-1-your-first-team/
│   ├── exercise-1.1-market-research-sprint/           (3-agent team analyzes a market opportunity)
│   └── exercise-1.2-team-architecture-workshop/       (design team structures for 5 scenarios)
├── module-2-task-pipelines/
│   ├── exercise-2.1-event-planning-pipeline/          (budget→venue→logistics dependency chain)
│   └── exercise-2.2-pipeline-blueprint/               (design dependency graphs for 3 workflows)
├── module-3-communication/
│   ├── exercise-3.1-feature-prioritization-debate/    (3 advocates + 1 decision-maker)
│   └── exercise-3.2-communication-protocol-design/    (plan message flows for 5 scenarios)
├── module-4-quality-gates/
│   ├── exercise-4.1-client-proposal-pipeline/         (draft→review→approve cycle)
│   └── exercise-4.2-review-workflow-design/           (plan approval pipelines for 3 deliverables)
└── module-5-capstones/
    ├── capstone-A-business-plan-assembly/             (full business plan with 5 agents)
    ├── capstone-B-customer-feedback-triage/           (triage 200 reviews with agent team)
    └── capstone-C-your-business-challenge/            (apply to your own domain)

How to Get Started

Prerequisites

Claude Code with Agent Teams enabled:

export CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1

A working Claude Code session (CLI or VS Code)

Quick Start

Navigate to the exercise folder: cd module-1-your-first-team/exercise-1.1-market-research-sprint
Launch Claude Code
Read the INSTRUCTIONS.md and start working

Exercise Types

Every module contains two exercises:

Type	What You Do	API Cost
Exercise A (Hands-on)	Run real agent teams with provided business data	$$
Exercise B (Design)	Design team architectures on paper -- no agents needed	Free

Design exercises (B-type) teach team THINKING without API costs. You can learn the concepts for free, then selectively run hands-on exercises.

Recommended Paths

Full Path (All Exercises)

Session	Module	Exercises	Domain	Budget
1	Module 1: Your First Team	1.1, 1.2	Knowledge Work	Low-Med
2	Module 2: Task Pipelines	2.1, 2.2	Corporate	Low-Med
3	Module 3: Communication	3.1, 3.2	Entrepreneurship	Low-Med
4	Module 4: Quality Gates	4.1, 4.2	Freelancer	Low-Med
5	Module 5: Capstones	A, B, or C	Mixed	Med-High

Budget-Friendly Path (Design First)

Complete all design exercises for free, then run select hands-on exercises:

Exercise 1.2 (design) → 2.2 (design) → 3.2 (design) → 4.2 (design)
Then pick 1-2 hands-on exercises to run with agents

Speed Path (Essentials Only)

Exercise 1.1 (learn teams) → 2.1 (learn pipelines) → 3.1 (learn communication) → Capstone A or B

Scoring Rubric

All exercises use the same 4-criteria rubric (see scoring-rubric.md):

Criteria	What It Measures
Comprehensiveness (1-5)	Did the analysis cover all required angles?
Actionability (1-5)	Could someone act on these recommendations?
Evidence Quality (1-5)	Are conclusions backed by data from starter files?
Team Coordination (1-5)	Did agents effectively share and build on each other's work?

Module exercises target: 14/20 Capstones target: 16/20

Tool Note

Claude Code (Terminal): Required for hands-on exercises (A-type). Agent Teams is a terminal feature.
Cowork (Desktop): Works for design exercises (B-type) but does not support multi-agent team orchestration.
Design exercises: Can be completed with any text editor -- no Claude Code needed.

Built for Panaversity's AI-Native Development Curriculum