Teach on Udemy

Turn what you know into an opportunity and reach millions around the world.

Learn More

Your cart is empty.

Keep shopping

Claude Code : Agent Skills, Agent Loop, Code Agents, MCP

Name: Claude Code : Agent Skills, Agent Loop, Code Agents, MCP
Rating: 4.4 (613 reviews)

Build Generative AI-augmented using Claude Code, Agent Skills, Agent Loop, MCP, Subagents, Slash Commands, AI Workflows

Created byFirstlink Consulting

Last updated 7/2026

English

English [Auto],Spanish [Auto],

What you'll learn

Design and Deploy Specialized AI Subagent Systems: Students will master the creation and configuration of specialized AI subagents
Master Claude Code's Memory Architecture: Students will understand how to create, organize, and maintain hierarchical memory files
Establish Enterprise-Ready AI Development Workflows: Students will design complete development workflows
Establish Enterprise-Ready AI Development Workflows: Students will design complete development workflows

Course content

23 sections • 109 lectures • 6h 44m total length

Introduction to Claude Code7:07
Discover what Claude Code is and how it brings AI-assisted development directly into your terminal. See how it can understand projects, generate code, troubleshoot issues, and streamline development workflows.
Install Claude Code3:46
Step-by-step walkthrough of installing Claude Code on your machine. Covers prerequisites, setup, and verifying your installation works correctly.
Using Claude Code with an IDE6:01
Connect Claude Code to your integrated development environment for a smoother coding experience.
See how terminal-based AI assistance and your code editor can work together during development.
Hands-On Project: Build an Expense Tracker6:20
Apply your Claude Code skills by building a practical expense-tracking application. Move from a basic project idea to working functionality using natural-language prompts and iterative development.

Introduction to Skills1:58
What are Skills1:52
What Skills Solve3:10
Skills vs MCP vs Agents1:32
Identify whether tasks require real-time external data, autonomous multi-step execution, or repeatable domain knowledge to decide if skills, MCP, or agents are best.
Install Skills on claude.ai6:01
Install skills on Cloud AI by enabling code execution, adding or selecting skills (including MCP Builder), and creating or uploading skills, then toggle them on to run locally.
Install Skill on Claude Code and Claude API3:01
Install skills on CloudCore and Cloud API by creating a .cloud directory with a skills file at project or user level, enabling team sharing and automatic discovery via API v1.
First Agent Skill - Hands On5:46
Skills with Claude Code - Hands On1:46

Architecture - Loading Skill3:25
Understand the skills architecture and on-demand loading in cloud, where the VM file system's skills directory feeds the context window based on user requests, loading only the requested skill.
Loading Skill - Step by Step2:57
Cloud loads only what's needed by matching pdf text to the pdf processing skill's description, loading the relevant instruction and references, illustrating progressive disclosure for efficient skill execution.
Design Principle - Progressive Disclosure2:45
Explore progressive disclosure as a design principle for dynamic skills loading: guide with references, domain organization, and conditional scripts to enable plan, validate, execute workflows.
Progressive Disclosure - Choosing the right file structure2:31
master progressive disclosure by choosing right file structure via two patterns: guide with references and domain-specific organization; follow five rules, including under 500 lines, one level deep, and descriptive names.

Plan the Skill3:11
Plan the skill before coding to define its domain, scope boundary, and necessary decisions for code review, including degree of freedom and pattern choices (template, example, workflow).
Code Review Skill - Walkthrough4:30
Execute Code Review Skill2:20
Execute a code review on a Python dbconnect file that reads a key-value config, connects to a database, and outputs a structured review with severity, strengths, and recommended next steps.
Quick Code Review1:12

Overview0:41
Skill Anatomy - The Frontmatter4:48
Design Principle - Right Degree of Freedom2:36
Design Principle - Skill Body Patterns2:47
Explore the three design patterns—template, example, and workflow patterns—and how to structure skills files with fixed sections, input-output pairs, and a copyable checklist.
3 Level of Loading Skills3:15
Describe three loading levels for skills, from startup front matter to domain expertise. Load level 1 at startup, level 2 on demand, level 3 as needed, with token optimization.
Skills with claude.ai4:38
Skill using API3:17

Agent Loop: What You'll Learn1:30
A quick roadmap for the module — five short lecture segments building toward a hands-on lab. See how "watching Claude Code work" becomes a concrete mental model you can code against by the end of the hour.
The Loop Behind Every Claude Code Session1:41
Every file read, edit, and test run you've seen Claude Code perform is the same underlying loop, running over and over. This lecture introduces that loop as the core mental model for the entire module.
The 5-Step Agentic Loop, Explained2:08
Break the loop into its five named steps — receive, evaluate, execute, repeat, return — and see exactly which SDK message type shows up at each stage, from the initial SystemMessage to the final ResultMessage.
The Loop Doesn't Know How Long It'll Take2:14
Compare a one-turn file lookup to a multi-turn refactor-and-test cycle using the same five-step loop. The key insight: the loop's shape adapts to the task, which is exactly why the SDK gives you dials to control it
Turns and the 5 Message Types1:36
Learn the precise definition of a "turn" and the five message types the SDK streams back — SystemMessage, AssistantMessage, UserMessage, StreamEvent, ResultMessage. Includes a full turn-by-turn trace of a real bug-fix session and the minimum code needed to handle each type safely.
Controlling the Loop: Permissions, Turns, Budget, and Effort2:20
Take the wheel with allowed_tools, disallowed_tools, and the five permission_mode values, then cap runaway agents with max_turns and max_budget_usd. Finish with effort levels — how to match reasoning depth to task complexity instead of defaulting to the most expensive setting.
What's Actually Eating Your Context Window2:29
The context window never resets mid-session. See exactly what accumulates — system prompt, CLAUDE.md, tool definitions, conversation history — and which of these get cached versus repriced on every single turn.
Compaction, Hooks, and Keeping Long Sessions Efficient1:56
When context nears its limit, the SDK compacts automatically — but a summary isn't a transcript. Learn why persistent rules belong in CLAUDE.md, how to steer what compaction preserves, and how subagents and hooks keep long-running agents cheap and observable.

Putting It All Together: One Script, Every Dial1:55
Walk through a single production-shaped script that combines everything from the module — allowed_tools, setting_sources, max_turns, effort, and full ResultMessage subtype handling — plus the five takeaways to carry into the lab.
Lab Setup: Installing the SDK and Your API Key6:41
Get your environment ready for the hands-on lab — install claude-agent-sdk, set ANTHROPIC_API_KEY, and preview the five message types you're about to see streaming live in your own terminal.
Lab: Watching the Agent Loop Run4:32
Write and run a script that streams every message type in real time, deliberately trigger an error_max_turns result and fix it, then compare effort="low" vs effort="high" on the same task to see the cost/depth tradeoff firsthand.

Introduction to Slash Command3:42
Examine slash commands in cloud code that use markdown prompts to automate workflows, save time, and standardize team processes. Learn project-based versus user-scoped commands and built-in versus custom options.
Slash Command Structure0:53
Master the slash command structure by learning the forward slash, command name, and positional arguments, illustrated with fix issue 123 high priority as the example.
Custom Slash Command - Quick Start2:05
Learn to pass multiple inputs to slash commands using positional parameters, demonstrated by a review file that prints a PR with a priority and assigns it to an author.
Slash Command Frontmatter4:10
Deep dive into frontmatter configuration for custom slash commands, including allowed-tools. Learn how frontmatter controls command behavior and permissions.
Bash in Slash Commands: Theory1:10
Understand the concepts behind embedding Bash commands inside custom slash commands. Covers how allowed-tools scoping works and why it matters for safety.
Bash in Slash Commands: Hands-On7:36
A practical, hands-on session writing slash commands that execute Bash. Reinforces the theory with real examples you can build and test.

Code Refactoring with Slash Commands4:39
Learn how to build a slash command that automates code refactoring tasks. Demonstrates how Claude Code can standardize and speed up common code cleanup work.
Git Commit Slash Command5:05
Build a custom slash command to automate git commit workflows. Shows best practices for scoping allowed-tools (e.g., Bash(git add:*), Bash(git commit:*)) safely.
MCP and Slash Commands4:56
Explore how slash commands can integrate with MCP (Model Context Protocol) servers. Covers connecting external tools/data sources into your custom commands.
Slash Command Workflow - Hands On4:01
See how slash commands fit into a real development workflow, standardizing repetitive tasks for you and your team. Ties together frontmatter, arguments, and Bash into one cohesive process.

Requirements

Modern development environment (Windows, macOS, or Linux), Claude Code installation, API access GitHub account for MCP server integration and Terminal/command-line access

Description

“This course contains the use of artificial intelligence to give better experience on a better voice quality”

July 4, 2026 - Agent Loop

June 26, 2026 - Agent Skills

Course Overview
Master Claude Code, Skills, Agent Loop, AI Subagents, MCP Servers, Slash Commands, Hooks, and GitHub Actions to build next-generation AI-augmented developer workflows. This course is your complete guide to the future of software engineering—where AI works alongside developers through automated pipelines, intelligent context handling, and well-orchestrated subagent systems.

This hands-on program teaches you how Claude Code transforms development productivity by acting as an AI teammate capable of executing workflows, running MCP servers, generating code, reviewing API designs, maintaining memory, and performing GitHub-integrated automations.

Whether you’re a software engineer, DevOps professional, QA engineer, or an AI enthusiast, this course equips you with the skills to build enterprise-grade AI workflows using Claude Code’s cutting-edge capabilities.

Learning Objectives

By the end of this course, students will be able to:

Understand Claude Code: Explain how Claude Code functions as both an MCP server and client, and its role in AI-assisted development
Agent Skills : Give your agents new abilities in minutes. Package expertise, workflows, and tools into reusable Skills — no rebuilding from scratch every time.
The agent loop : The five-step cycle — receive, evaluate, execute, repeat, return — that powers every Claude Code session, running once for a quick lookup or hundreds of times for a complex refactor. It's the same loop exposed directly through the Claude Agent SDK, giving you full visibility and control over how Claude reasons, calls tools, and returns results.
Implement Subagent Systems: Design and deploy specialized AI subagents for different aspects of the development lifecycle. We will go over understanding how Subagents work and what are their benefits
Slash Commands: Developers trigger powerful coding, debugging, and automation actions directly from chat. They streamline workflows by invoking predefined tools, hooks, and GitHub Action integrations with a simple command.
Workflow: An automated, multi-step process that Claude can run on your codebase—such as generating code, refactoring, testing, or syncing changes—based on defined steps in your project configuration. It acts like a programmable pipeline that Claude executes deterministically, ensuring repeatable and reliable automation triggered by slash commands or file changes.
Configure MCP Servers: We will take a deep dive on MCP server. Set up and integrate Model Context Protocol servers to extend Claude Code's capabilities. We will integrate with the MCP servers
Memory Systems: Create and maintain hierarchical memory structures that enhance AI assistance quality
Apply Best Practices: Implement enterprise-ready workflows using AI subagents while maintaining security and compliance standards

SECTION 1 — Introduction

Learn installation, extension setup, and create your first “Hello World” application inside the IDE.

SECTION 2 — Claude Architecture

Learn about Claude Code Architecture and how claude interacts with different components

SECTION 3 — Introduction to Claude Skills

Learn what Claude Skills are, how they load efficiently using a three-level filesystem model, and how they compare to MCP servers and Agents.
By the end you will have your first working Skill installed and triggering correctly in Claude Code or claude ai.

SECTION 4 — Building Custom Skills

Go deep on authorship — the exact YAML frontmatter rules, writing descriptions that trigger reliably using the WHAT + WHEN formula, and structuring the SKILL body using template, examples, and workflow patterns.
Covers installing Skills across claude ai, Claude Code, and the API, plus a hands-on walkthrough of Anthropic's pre-built Skills.

SECTION 5 - The Agent Loop: How Claude Code Actually Thinks and Acts

The 5-step loop behind every Claude Code session, and the 5 message types it streams back
How to control cost and depth with turns, budget, and effort settings
How context accumulates and compacts over long sessions
A hands-on lab building and running your own agent loop script

SECTION 6 & 7 — Slash Commands

Understand Slash Command structure, frontmatter definitions, and backend logic.
Learn how to create custom commands powered by Bash scripts to automate:

Code refactoring
Git commits & version control
MCP integration
Multi-step workflows

Hands-on labs reinforce practical usage.

SECTION 8 — Subagents: General Purpose

Dive deep into the Subagent
You’ll build intelligent multi-agent systems that perform tasks like:

Travel planning
Trip scheduling
Restaurant recommendations
Parallel & sequential orchestration

These modules teach AI planning and agent chaining.

SECTION 9 — Subagents for Developers

Use Subagents to accelerate real engineering workflows:
API review assistants, documentation reviewers, code auditors, and more.
Perfect for teams adopting AI pair-programming at scale.

SECTION 10 — Workflows with Subagents + Slash Commands

Build hybrid pipelines combining Slash Commands and Subagents.
Create deterministic workflows that Claude executes reliably for:

Code changes
Commit flows
Automated testing
Structured build steps

SECTION 11 — Claude Code Memory (Theory + Hands-On)

Master Claude’s hierarchical memory system, including:

Context preservation
User memory
Project root memory
Subdirectory memory
Memory priority
Memory access commands

This section teaches persistent multi-file context handling—critical for large projects.

SECTION 12 — All About MCP

A full deep dive into Model Context Protocol, covering:

MCP
Server components
Transports
End-to-end data flow

You will clearly understand how tools communicate with Claude.

SECTION 13 — Claude Code with MCP Servers

Integrate powerful servers:

Puppeteer MCP
Sequential Thinking MCP
GitHub MCP (with and without authentication)

Learn how MCP extends Claude Code into a full automation platform.

SECTION 13 — Hooks

Type Script formatting
Activity logging
Execution safeguards
Multi-step sequences
Debugging failing Hooks

SECTION 12 — Claude Code with GitHub Actions

Pull Request creation
Bug fixing
Code review workflows
Repo updates
Continuous automation pipelines

Who this course is for:

Software developers seeking to enhance productivity with AI assistance, Engineering managers exploring AI integration in development workflows DevOps engineers interested in AI-powered automation, Computer science students focusing on AI applications in software engineering

Claude Code : Agent Skills, Agent Loop, Code Agents, MCP

What you'll learn

Explore related topics

Course content

Introduction4 lectures • 23min

Claude Code Architecture1 lecture • 6min

Agent Skills8 lectures • 25min

Agent Skills Architecture4 lectures • 12min

Agent Skills - Hands on4 lectures • 11min

Custom Agent Skills7 lectures • 22min

The Agent Loop: How Claude Code Actually Thinks and Acts8 lectures • 16min

Agent Loop - Lab3 lectures • 13min

Slash Commands6 lectures • 20min

Slash Commands - Advanced4 lectures • 19min

Requirements

Description

Who this course is for: