Developed with Claude Code
Course Objective
This course is designed to help you progress from basic use of Claude Code to becoming a confident “agent-driven” engineer: in-depth work with context, customization for your own tech stack using MCP, hooks, and skills, building repeatable workflows for the entire team, and integration with CI/CD and issue tracking systems.
Learning Format
Online/offline/hybrid
Practical case studies, workshops, interactive sessions
Support for participants during training via a private chat
What Will Graduates Gain?
The problem the course solves
Developers who start using Claude Code often get stuck at the “initial plateau”: they quickly solve small tasks but don’t know how to adapt the tool to their stack, don’t use MCP, hooks, and skills, and don’t know how to scale workflows to the team. The course provides a systematic transition from individual use to the engineering discipline of working with AI agents.
Prerequisites
- 1+ year of commercial development experience
- Proficiency with Git and the CLI
- Basic experience using Claude Code or another AI agent (Cursor, Copilot, Cody)
- Access to your own working repository (preferably medium-sized: 50k+ lines of code)
- Node.js 18+; Anthropic account
Program
- Rapid recap: model + tools + context
- Process architecture: tool use, sub-agents, side effects
- Quick setup, IDE integrations, settings, OAuth, and API keys
- How the model indexes a project; context window limitations
- CLAUDE.md across 3 levels: global / project / directory
- Naming conventions and references to "sources of truth"
- Using /init as a starting point, followed by manual refinement
- Plan mode, auto-edit, accept-edits
- Allowlists and denylists for tools and bash commands
- Working with secrets and sensitive data
- Sandboxing and network access restrictions
- Built-in commands: /help, /init, /review, /clear, /compact, /cost, /status
- Custom slash commands: format, storage location, arguments
- A template library for common team workflows
- Delegation concepts: explorer, planner, reviewer, tester
- When sub-agents improve outcomes—and when they do not
- Creating and using sub-agents
- End-to-end workflow: discover → plan → implement → test → review → merge
- Working with tickets and issue trackers
- Task readiness checklists
- What Model Context Protocol is and when it makes sense
- Ready-to-use servers: GitHub, Jira/Linear, Slack, Postgres, Sentry
- MCP configuration in Claude Code
- Hook types: pre/post tool use, on stop, on session start
- Use cases: linting before commits, preventing edits to specific files, Slack notifications
- Best practices and anti-patterns
- The concept of skills and how they differ from slash commands and sub-agents
- SKILL.md structure and automatic activation
- Creating domain-specific skills (e.g., database migrations, DTO generation)
- Coverage strategies, mutation testing, and property-based testing
- Integration testing with real dependencies
- Snapshot testing and E2E testing with Playwright/Cypress
- Self-review templates, /review, and team checklists
- Identifying security issues and language-specific review patterns
- Integration with GitHub pull requests (via MCP or CI)
- Running Claude Code in CI/CD pipelines (headless / non-interactive)
- Automated fixes for Dependabot and Snyk issues
- Automatic changelog and release notes generation
- Distributing CLAUDE.md, slash commands, skills, and hooks through repositories
- Onboarding new developers with the help of AI agents
- Measuring impact: cycle time, defect rate, review time
- Anti-patterns: "AI does everything", "vibe coding", and over-engineered pull requests
- Presenting an end-to-end case: real-world problem → delivered outcome
- Live coding demonstration
- A 30/60/90-day adoption plan for introducing the tool into a team
- Feedback and discussion
Education provides
