This hands-on workshop is designed to showcase intermediate-to-advanced features of Claude Code through structured, hands-on tracks. Participants will master basic workflows AND genuinely advanced patterns like parallel development, production automation, and multi-agent orchestration.
By the end, attendees should understand how to:
- Apply PRDs (Product Requirements Documents) to an existing repo with minimal human intervention.
- Use subagents to generate structured task plans and execute them.
- Integrate MCP servers for documentation access, browser testing, and visual validation.
- Run a full loop: spec → code → tests → PR.
- Master advanced workflows: parallel development with worktrees, production-ready hooks, and multi-agent coordination.
- Duration: 1.5 hours total (flexible)
- Pre-reqs:
- Laptops with GitHub access.
- Node.js + npm installed (if choosing track 1).
- Git CLI and
gh(GitHub CLI) installed and authenticated. - Claude Code setup in their IDE/editor with MCP enabled.
- Repos: Fork TodoMVC React (if choosing track 1).
Demonstrate a fully automated end-to-end dev cycle where Claude Code executes against a well-specified PRD.
- PRD Examples:
- See example PRD markdown files here
- Subagents Examples:
- See example agent markdown files here
- Alternatively, try creating a new agent (following the formatting guidelines in the Claude docs)
- Custom Slash Command Examples:
- See example custom slash commands here
- MCP Server Docs:
- Advanced Guides:
-
Clone the ToDo Repo
git clone https://github.com/tastejs/todomvc.git cd todomvc/examples/react -
Install Dependencies
npm install
-
Run the App
npm run serve
-
Open in Your Editor
- Open the
examples/reactfolder in VS Code (or your preferred IDE). - Add your chosen/created
PRD-*.mdandagent-*.mdfiles to this folder.
- Open the
-
Start Claude Code
- Launch Claude Code in your IDE.
- Ensure any MCP servers are configured (Playwright MCP, Context7, etc.) as needed .
- Follow the guide for creating/updating a
CLAUDE.mdfile.
- Pick one of the provided PRD example markdown files and modify it (or create a new one from scratch) and copy it into the repo root (i.e.,
examples/react/) - Pick one of the provided subagent markdown files and modify it (or create a new one from scratch) and copy it into the repo (i.e.,
examples/react/.claude/agents/)- Alternatively (recommended) do this at a later stage via the interface by calling
/agentsafter Claude Code has been initialized.
- Alternatively (recommended) do this at a later stage via the interface by calling
- Add MCP servers to Claude Code (optional, as desired)
- Start Claude Code in YOLO mode and add a
CLAUDE.mdfile. - Prompt Claude Code to spin up a subagent and generate a task list based on the PRD.
- Read Docs (
CLAUDE.md, repo docs). - Implement Feature.
- Unit + Browser Tests (e.g., via Playwright MCP).
- Create PR via GitHub CLI.
- Review PR in GitHub (or chosen tool).
- Merge PR.
- Review the updates in the app running locally.
Allow experienced developers to apply Claude Code concepts to their own repositories instead of the provided TODO app. This gives participants a chance to explore subagents, MCP servers, and task automation in a codebase that matters to them.
- Participants must have:
- An active repo they’re comfortable experimenting with.
- Tests set up (unit or integration).
- GitHub CLI installed for PR creation.
Developers can try:
-
Spec-driven dev:
“Here’s a new feature request: [paste PRD or Jira ticket]. Please use the generate-tasks subagent, follow CLAUDE.md conventions, and implement the feature with tests and a PR.”
-
Refactoring task:
“Please refactor all class components to functional components with hooks, add tests to confirm unchanged behavior, and create a PR.”
-
Testing enhancement:
“Add browser tests for user login using Playwright MCP. Make sure they run successfully before creating a PR.”
Master advanced Claude Code problem-solving techniques that dramatically improve your daily development workflow: debugging stuck conversations, managing context in large codebases, and sophisticated prompting for complex scenarios.
25 minutes per section (pick 1-2 based on your current challenges)
- Completed Track 1 or 2, OR
- Existing Claude Code experience with basic workflows
- Experience with getting "stuck" or frustrated with Claude responses
Skill: Use git worktrees to develop multiple features simultaneously with focused Claude Code sessions
Hands-On Exercise:
- Set up 2 worktrees for TodoMVC with parallel features (categories + search)
- Run focused Claude Code sessions in each worktree
- Develop features independently, then integrate
- Success Criteria: Complete 2 features faster than sequential development
Real Scenario: "Add dark mode AND user preferences without context switching overhead"
Skill: Quickly recover when Claude gets confused, over-analyzes, or produces poor results
Hands-On Exercise:
- Intentionally create a "stuck" scenario with vague requirements
- Practice recovery techniques: reset & redirect, scope limiting, progressive disclosure
- Transform confused conversation into productive implementation in under 5 minutes
- Success Criteria: Master 3 recovery patterns applicable to your daily workflow
Real Scenario: "Claude is over-analyzing instead of implementing" → Working feature in minutes
Skill: Guide Claude effectively in large, complex codebases without information overload
Hands-On Exercise:
- Take a complex feature request for TodoMVC (or your own codebase)
- Practice context layering, selective file reading, and pattern recognition
- Guide Claude to understand and modify code without getting lost
- Success Criteria: Claude makes meaningful changes with no confusion or backtracking
Real Scenario: "Add authentication system" → Clean implementation that follows existing patterns
Skill: Guide Claude effectively in large, complex codebases without information overload
Hands-On Exercise:
- Take a complex feature request for TodoMVC (or your own codebase)
- Practice context layering, selective file reading, and pattern recognition
- Guide Claude to understand and modify code without getting lost
- Success Criteria: Claude makes meaningful changes with no confusion or backtracking
Real Scenario: "Add authentication system" → Clean implementation that follows existing patterns
Skill: Craft advanced prompts that get superior results for complex, real-world scenarios
Hands-On Exercise:
- Learn constraint-driven development, result-driven prompting, and multi-phase problem solving
- Compare basic prompts vs advanced patterns on the same task
- Practice production-mindset vs prototype-mindset prompting
- Success Criteria: Advanced prompts produce dramatically better results than basic ones
Real Scenario: Transform "make it better" → Specific, measurable improvements with clear success criteria
- Parallel Development:
guides/Parallel_Development_with_Worktrees.md+examples/worktrees/ - Problem-Solving:
examples/problem-solving/- Copy-paste patterns for common scenarios - Production Automation:
examples/hooks/- Quality gates and team integration
- Parallel Development: Worktree workflows that enable focused, simultaneous development
- Problem-Solving Focus: Skills you need daily but can't learn from basic documentation
- Immediate Impact: Techniques that improve your workflow from day one
- Real Scenarios: Based on actual frustrations developers face with Claude Code
- Transferable Skills: Patterns that work across any codebase or technology stack
- Subagent Chain: Planning → coding → testing.
- Visualization MCP: Hook Claude into a screenshot/preview MCP.
- Repo-wide Refactor: Try automated, large-scale changes.
- Each group demos their PR(s).
- BYOC participants share wins + blockers.
- Highlight:
- Strengths of Claude Code (task decomposition, test execution).
- Where human oversight remains essential.
- How this maps to real enterprise workflows.