Claude Code Mastery: Build AI Agents, MCP & Automation

Identify how slash commands provide in-session controls, use the key commands demonstrated in the lesson, and view usage and cost information with /cost.

Use /compact to condense a long conversation while retaining its useful context, and recognize when compacting is appropriate before continuing a task.

Use /clear to begin a fresh session when earlier conversation context is no longer needed, and distinguish resetting a session from compacting it.

Compare Standard, Auto-Accept, and Bypass permission modes, understand how each affects approval of actions, and select an appropriate mode for the task.

Use Plan Mode to examine a task, outline an approach before making changes, and review the proposed steps before moving into implementation.

Execute a planned task, inspect the resulting changes, and use context-management commands after longer workflows to retain the project information needed for the next step.

Inspect the hidden configuration structure used by Claude Code, identify the main project-level components introduced in this part, and understand how they organize coding guidance.

Continue through the remaining project configuration components, connect each component to its role, and see how the configuration workflow is applied to a coding project.

Write project-level instructions, organize context for projects and nested folders, and explain how Claude Code loads instructions and memory when handling a task.

Compare the available instruction and settings scopes, decide where each rule or preference belongs, and understand when each setting applies.

Pin a selected model version and manage prompt versions so a project can use known model behavior and consistent instructions.

Apply reusable Claude Rules in a practical workflow, defining guidance that helps Claude Code follow project conventions and produce more consistent changes.

Explain what a context window is, how it fills during a coding session, and how accumulated context can begin to affect the quality and relevance of Claude Code’s responses.

Recognize when a session has accumulated outdated or irrelevant context, decide when to start fresh, and identify which project details are useful to carry into the next session.

Keep requests focused, prioritize relevant context, and apply token-budgeting habits to reduce unnecessary input during Claude Code tasks.

Use /resume, /fork, /context, /rename, /rewind, and /status to revisit, branch, inspect, rename, restore, and monitor Claude Code sessions.

Use Auto Memory to retain useful project knowledge, prune unnecessary tool output, and combine context-management practices to keep future sessions focused.

Explain how Claude Skills package reusable instructions, identify the components of a Skill, and create a first Skill to guide Claude through a repeatable task.

Implement an Agent Skill step by step and apply it in a practical Claude Code workflow to make a specialized task reusable.

Explain how Hooks respond to workflow events, examine PreToolUse and PostToolUse triggers and their schema, and configure notifications and guardrails.

Identify the Hook event types introduced in this lesson and match them to practical use cases in a Claude Code workflow.

Explore advanced Hook events, understand when each event is triggered, and select an appropriate event for a specific workflow action.

Complete the Hook event reference and apply the events as part of a repeatable workflow for automation and process control.

Explain the purpose of the Model Context Protocol and compare MCP with connectors to understand how Claude Code can access external systems and capabilities.

Follow a practical Gmail connection demo to see how MCP links Claude Code with an external service and how that connection is used in a workflow.

Distinguish local and hosted MCP deployments, identify where each server runs, and understand the basic connection model for each option.

Continue the deployment walkthrough and follow the steps for connecting Claude Code to an MCP server in the selected deployment environment.

Complete the local and hosted MCP architecture walkthrough and compare how each deployment model supports Claude Code connections.

Follow the custom MCP server walkthrough to create a server that exposes capabilities for Claude Code to use in an external-system workflow.

Distinguish MCP resources, tools, and prompts; explore how server capabilities are discovered; and follow the practical examples, including the AWS example.

Explain the core elements of agent architecture, distinguish an agent’s structure from its workflow, and see how multi-agent systems extend the basic model.

Examine how a multi-agent system divides work among agents and how their responsibilities fit into a coordinated architecture.

Analyze more advanced multi-agent architecture patterns, focusing on how agents coordinate, exchange work, and contribute to a shared outcome.

Connect subagents to the broader agent architecture, and explain how the Agent SDK and code harness support the construction and coordination of agent workflows.

Explore managed agents and their workflow, including how pre-action and post-action hooks can respond to events during an agent’s execution.

Trace the agentic loop from deciding on an action to making a tool call and observing its result, using managed agents as the practical context.

Define subagents, explain their benefits and common use cases, and identify when delegating a task to a specialized agent can help.

Divide work across multiple subagents running in parallel, then understand how their separate results can support the larger task.

Use /agents to invoke and manage agents, and understand how to work with available agent definitions during a Claude Code task.

Begin implementing a custom subagent by defining its specialized role and instructions for a focused development task.

Complete the custom subagent implementation and apply it to delegate a focused task within a Claude Code workflow.

Compare agent teams with subagents, identify how their roles and coordination differ, and consider which approach fits the project introduced in the lesson.

Begin a practical project with an agent team by organizing the work across agents and starting the coordinated implementation.

Continue and complete the agent-team project, following how the agents’ contributions are coordinated into a working result.

Explain the role of plugins in Claude Code, review the plugin workflow demonstrated in the lesson, and install a plugin to extend the working environment.

Browse the plugin marketplace, find a plugin relevant to a workflow, and follow the steps to add and manage marketplace plugins in Claude Code.

Compare the scheduling methods covered in the lesson and use /loop and the other demonstrated approaches to schedule Claude Code tasks.

Explain how headless mode supports non-interactive Claude Code operation, and explore how Claude Routines can be used to organize automated tasks.

Set up the autonomous task workflow demonstrated in the lesson and understand how it allows scheduled Claude work to continue while your laptop is closed.

Build a Claude Messages API request using system prompts and message content, and explore response formats, token use, streaming, and image input.

Compare batch and real-time processing, explore prompt caching and API providers, and use token counts and cost modeling to assess API usage.

Explain tokens and context, explore sampling and non-deterministic outputs, and understand how these concepts relate to Claude’s Fast Mode.

Compare effort settings and model choices, apply zero-shot, single-shot, and multi-shot prompting, and explore adaptive and extended thinking.

Improve prompt results by setting explicit criteria, adding few-shot examples, requesting structured JSON, and applying a prompt-refinement loop.

Turn software-development requirements into clear prompts by specifying the task, relevant context, constraints, and the expected result.

Identify prompt-injection, personal-data, and data-leakage risks, and explain how permission controls help protect Claude Code workflows.

Apply permission settings in a practical workflow, configure sandbox boundaries, and observe how sandboxing limits the actions available to Claude Code.

Explain how security Hooks can enforce workflow guardrails, and begin building a Hook that applies security checks to Claude Code actions.

Complete the security Hook implementation and follow how it applies the defined checks during a Claude Code workflow.

Distinguish authentication from authorization, understand how API keys and permissions affect access, and write tool descriptions that help agents choose suitable tools.

Follow a practical build of an agent tool, compare built-in and custom tools, and learn how to make a custom capability available in an agent workflow.