Master Class: Agentic Software Engineering using Claude Code (ASECC) Online

Kursinhalt

Module 1: Claude Code Fundamentals

• Get Started with Claude Code
• Install and Configure Claude Code
• Choose Your Editor
• Write a CLAUDE.md Constitution
• Master Prompting Techniques
• Select the Right Model
• Configure the Permission Model
• Run Parallel Agents with Git Worktrees

Module 2: Claude Code Configure Harness

• Understand the Claude Code Harness
• Build Custom Slash Commands
• Automate Workflows with Hooks
• Connect MCP Servers to the Harness
• Create Reusable Skills
• Package and Distribute Plugins
• Persist Context with Memory
• Checkpoint and Resume Long Sessions
• Delegate Work to Sub-Agents

Module 3: Creating & Managing Skills

• Understand Skills vs Slash Commands
• Create a Skill from Scratch
• Compose Skills into Workflows
• Test and Publish Skills
• Evolve Skills with Metrics

Module 4: Consuming & Designing MCP Tools

• Master the Model Context Protocol
• Build a Custom MCP Server in C#
• Implement and Use MCP Applications

Module 5: Programmable Automation using the Claude Code CLI

• Run Claude Code in Headless Mode
• Parallelize Work with Git Worktrees
• Gate CI Pipelines with Sentinel Tokens
• Enforce Policy Hooks Across Parallel Worktrees
• Build Agents that Loop Until a Verifiable Goal is Met

Module 6: Plan and Decompose for Maximum Agentic Effectiveness

• Enter Plan Mode for Complex Tasks
• Reason Deeply with Extended Thinking and Persist Decisions in STATE.md
• Drive Specs with Spec Kit and BMAD

Module 7: Scale Task Execution with Sub-Agents

• Explore Sub-Agent Architecture and Context Isolation
• Implement a Backend Sub-Agent for .NET
• Configure a Frontend Sub-Agent for Angular
• Configure a QA Agent for Browser Verification
• Unit-Test Your Agent's Performance
• Orchestrate Sub-Agents with Decomposition Patterns

Module 8: Optimizing the Agentic Harness & Continuous Learning

• Optimize the Agentic Harness
• Compress Context with the Caveman Technique
• Build an Autoresearch Loop
• Design Memory Architectures
• Build a Second Brain with Obsidian

Module 9: Deploy Autonomous Action using Managed Agents

• Understand the Managed Agents Mental Model
• Build Your First Managed Agent
• Add Production Features to Managed Agents
• Add Human-in-the-Loop Controls

Module 10: Security, Hardening & Observability

• Map the Agent Threat Model
• Apply Guardrails and Least Privilege
• Secure Secrets and Credentials
• Trace Agents and Diagnose Failures

Module 11: Build Multi-Agent Pipelines using the Agent SDK

• Understand the Claude Agent SDK
• Call the Claude Agent SDK
• Run Parallel Agents with the SDK

Module 12: Bundling & Distributing Plugins

• Understand Why Plugins Exist
• Inspect Plugin Anatomy
• Convert a .claude/ Config to a Plugin
• Publish Plugins to Marketplaces

Module 13: Agentic DevOps & Automation

• Deploy with the Azure GitHub DevOps Agent
• Deploy to Azure with GitHub Actions
• Automate Deployments with Azure Developer CLI
• Provision Infrastructure with Bicep and Terraform
• Drive the Azure CLI from Claude Code
• Agentic Optimization: Web Vitals and SEO

Voraussetzungen

The exercises will be conducted in Python, TypeScript, and C#. Basic knowledge of these programming languages is helpful but not required.

Zielgruppe

Software Architects, Developer (All)