Claude Code: Generate, Debug, and Improve Code with AI Assistance

Claude access will be provided for workshop activities.

Prerequisite: This is a beginner-level Claude Code course, not a programming course. Participants should have prior experience with at least one programming language and be comfortable working from the command line.

Coding with AI assistance is no longer just about autocomplete. In this hands-on course, participants work directly with Claude Code, Anthropic's command-line AI tool, to tackle real coding tasks through natural conversation. Whether you are writing a function from scratch, tracking down a stubborn bug, or looking for ways to make existing code cleaner and easier to maintain, this course offers practical experience with AI as a collaborative coding partner.

Across two sessions, participants explore the full Claude Code workflow: setting up the environment, prompting effectively, generating and refining code, diagnosing errors, and applying improvements. The course is designed for people who write code regularly and want to work more efficiently, as well as those who are newer to coding and want to use AI assistance to build confidence and capability. No prior experience with Claude Code is required.

Learners will have the opportunity to:
1. Set up and navigate the Claude Code command-line environment and configure it for their development context
2. Generate working code from natural language prompts and refine output through iterative conversation
3. Identify and debug errors in existing code with AI-guided analysis and suggested fixes
Improve code quality by exploring techniques for refactoring, documentation, and readability
4. Practice applying Claude Code to realistic tasks such as writing functions, building small scripts, and reviewing code snippets
5. Use Claude Code within a Git-based workflow to generate commit messages, summarize diffs, and support code review and collaboration tasks
6. Explore ways to extend Claude Code through custom commands, external tool connections, and automation patterns, and examine its limitations and responsible use in a development context

Topics:

Session 1: Getting Started with Claude Code
What Claude Code Is (and Isn't)
- Where Claude Code fits: CLI-native, agentic, operates directly on your codebase
- How it differs from chat-based AI and inline copilots (Copilot, Cursor)

Project Configuration
- How Claude Code reads your project context ¿ file tree awareness and codebase understanding
- Configuring with CLAUDE.md for project-level instructions and coding standards
- Setting permissions and trust levels for file edits and command execution

The CLI Interface and Core Commands
- Navigating the conversational CLI ¿ entering prompts, reviewing diffs, accepting or rejecting changes
- Key slash commands and context management

Prompting for Code Generation
- Writing effective prompts: task + context + constraints
- Iterative refinement through follow-up prompting
- Hands-on: generating functions, scripts, and route handlers

Multi-Step Tasks and File Operations
- Creating, modifying, and scaffolding across multiple files in a single conversation
- Chaining tasks: generate a function, write its test, run the test
- Using Claude Code to read and summarize unfamiliar codebases

Session 2: Debugging, Improving, and Applying Claude Code
Debugging with Claude Code
- Submitting broken code, error messages, and stack traces for diagnosis
- The agentic debugging loop: Claude Code runs your code, observes the failure, and proposes a fix
- Hands-on: debug pre-broken scripts with progressively harder issues

Refactoring and Code Quality
- Refactoring for readability, performance, and maintainability
- Auto-generating comments, docstrings, documentation, and type annotations
- Evaluating code against best practices configured in CLAUDE.md

Git Integration and Workflow Patterns
- Generating commit messages, summarizing diffs, and writing PR descriptions
- Real workflow patterns: test generation, dependency upgrades, codebase onboarding

Extending Claude Code
- MCP (Model Context Protocol) -- connecting Claude Code to external tools like GitHub, databases, and docs
- Custom slash commands for reusable team workflows
- Headless/CI mode for non-interactive automation

Limitations and Responsible Use
- Context window limits, security considerations, and cost awareness
- Verifying output and avoiding over-reliance
- Building Claude Code into your daily development workflow

Prerequisite: This is a beginner-level Claude Code course, not a programming course. Participants should have prior experience with at least one programming language and be comfortable working from the command line.