Deep Dive Claude Code | Gui Ferreira

Learn how to use and configure Claude Code, from custom skills and hooks to MCP servers and CI/CD integration.
AI coding assistants are rapidly moving beyond simple autocomplete, and Claude Code represents a major shift toward autonomous agentic workflows directly in your terminal. Getting the most out of it requires understanding how it actually reasons, manages context, and executes commands. This course is a deep dive into configuring and extending Claude Code for real-world engineering projects. You will start with the fundamentals of how the agent thinks, how to balance Model and Effort settings, and how to control costs while integrating it into your IDE with multimodal inputs. From there, we will go under the hood to explore configuration at scale, detailing the .claude directory, how Claude's memory works, and how to manage context using modular CLAUDE.md files, .claude/rules/, and Sessions with Checkpoints.
The core of the course focuses on extensibility and execution control. You will build custom Agent Skills utilizing bash arguments and file inclusion, and intercept the execution lifecycle using Pre-Tool, Post-Tool, and Prompt Hooks. We will cover how to orchestrate Built-in Subagents and create custom ones, as well as how to connect Claude to external systems by configuring MCP Servers and using MCP tools and resources. Because autonomous agents introduce unique risks, we dedicate a module to the permission model, bash sandboxing, and mitigating Prompt Injection. Finally, you will master autonomous operations, learning how to drive dynamic workflows, run the agent in Headless Mode for CI/CD integration with GitHub Actions, and package your custom extensions as reusable Plugins.
Homepage

---