AI-Augmented Developer - Master Agentic Coding

Software Engineer (operational backend/frontend developers), Software Architect, Tech Leader in companies, IT services companies, startups, and scale-ups who want to boost their efficiency with AI while maintaining a high level of code quality., Teams concerned with maintainability, robustness, and best practices, looking to use AI to modernize their methods while improving the quality of their deliverables.

After this course participants should be able to:

• Accelerate all phases of the development cycle, from architectural design to resolving complex bugs, increasing their velocity while ensuring code quality, security, testing, documentation, and maintainability.
• Collaborate continuously with AI by adopting AI-augmented work methods that enhance productivity and efficiency.
• Master the ecosystem of leading tools for coding with AI and AI frameworks (like Claude Code, Gemini CLI, GitHub Copilot, Cursor AI), evaluate and adopt new emerging tools, understanding their respective strengths and limitations.
• Transform a classic team into an augmented team by orchestrating collaborative AI agents, establishing team standards (shared contexts, reusable prompts), and disseminating best practices for AI-assisted development that accelerate onboarding and standardize workflows.

01- Introduction to AI for Developers

• AI Refresher: Quick review, history of AI / ML / NLP / Generative AI, arrival of ChatGPT/LLM
• Market Models: GPT-5, Claude, Gemini - key differences
• The Art of Prompting: how to communicate effectively with AI to get quality code
• Evolution of tools: from ad-hoc assistance to collaborative agents
• Context and tokens: understanding the constraints
• Overview of tools: ChatGPT, Claude, etc
• Activities for Module 1:
• Prompt engineering: comparison of 3 approaches on a concrete case

02- Vibe Coding vs Agentic Coding

• Vibe Coding: definition, limits, and dangers
• Agentic Coding: vision of the augmented developer
• Continuous collaboration vs passive generation
• Impact on quality, maintainability, and documentation
• Activities for Module 2:
• Setting up the vibe coding tool
• Guided experiment: vibe code an application without looking at the generated code and explain the results, constraints, and advantages

03- Discovery and Practice of Agentic Coding

• Practical introduction: what is Agentic Coding for a developer?
• Live demonstration: traditional development vs augmented development
• Quality focus: how AI can improve code quality (conventions, patterns, best practices)
• Agentic workflow: Specify → Plan → Tasks → Implement → Validation
• Context management: the effective .md context file (locally for now)
• Structuring requests and iterations
• Best practices: documentation, testing, quality
• Activities for Module 3:
• Scenario 1: legacy - Refactoring legacy: getting started with the project to make it Agentic Coding compliant. Going as far as the migration plan
• Scenario 2: from scratch and evolutions - Project from scratch: architecture, technical choices, structuring; Development of a complete feature with the agentic workflow

04- Deep Dive on Agentic Tools

• The following points are covered and practiced according to the actual capabilities of the chosen agentic tool (Claude Code, GitHub Copilot, Gemini CLI, Codex CLI, etc.)
• Tool overview
• Key features, CLI commands, and shortcuts for each tool
• Custom agents: creating specialized sub-agents with isolated context for targeted workflows
• Custom commands / prompts: reusable prompt templates for recurring tasks
• Agent skills: composable knowledge loaded on demand
• Hooks: shell commands triggered on lifecycle events
• Plugins: distributable packages bundling agents, skills, hooks, and MCP servers
• Configuration files
• Context file
• Choosing the right type of feature for each scenario
• Activities for Module 4:
• Deep dive on the tool of your choice: create custom agents, commands, skills, hooks, plugins, etc.

05- Using MCP Servers

• Model Context Protocol (MCP): concepts and architecture
• Installation and use of MCP clients (examples: playwright, context7, Atlassian...)
• Concrete use case: MCP Playwright to facilitate testing with visibility of outputs in the browser
• Activities for Module 5:
• Installation of an MCP and complex interaction: screen capture + test generation

06- Augmented Team Development

• Using agent skills in a team context: learn to leverage existing reusable skills to consistently apply architecture rules, coding standards, and testing practices across the entire team. (Claude/Copilot/Gemini)
• Leveraging custom agents for roles: interact with role-specific agents (e.g., Architecture Guardian, Test Strategist, API Contract Agent) to get contextual guidance suited to your tasks. (Claude/Copilot)
• Running custom commands: practice invoking custom slash commands (e.g., /code-reviewer, /prepare-pr, /trace-flow) to speed up PR review, diff analysis, and test planning, etc. (Claude/Copilot/Gemini)
• Using marketplace plugins: install and use Claude Code plugins from marketplaces to extend functionality (commands, agents, skills, hooks, and MCP servers). (Claude)
• Configuring custom instructions: define repository-wide and path-specific rules to guide the AI. (Copilot)
• Using hooks for automated guardrails: apply pre-commit and post-edit hooks for quality gating. (Claude)
• Using hooks for automation: use Gemini hooks (e.g., BeforeTool, AfterTool, BeforeAgent) for validation and automation. (Gemini)
• Accessing repository knowledge: explore architecture docs, standards, and the business glossary. (Claude/Copilot/Gemini)
• AI-assisted collaborative code review: produce consistent and actionable comments. (Claude/Copilot/Gemini)
• Diff analysis and risk management in practice: interpret recommendations on risky changes. (Claude/Copilot/Gemini)
• Activities for Module 6:
• Hands-on: run /code-reviewer on a PR diff and analyze the recommendations. (Claude/Copilot/Gemini)
• Hands-on: interact with the Architecture Guardian agent to verify architectural compliance. (Claude/Copilot)
• Hands-on: interact with the topology-awareness skill to assess changes. (Claude/Copilot/Gemini)
• Hands-on: install a plugin from a marketplace and use its commands/agents/skills. (Claude)
• Hands-on: use /prepare-pr to generate a PR description and a checklist. (Claude/Copilot/Gemini)
• Hands-on: run /trace-flow on a backend endpoint and examine the generated explanation. (Claude/Copilot/Gemini)
• Hands-on: observe pre-commit and post-edit hooks in action on sample changes. (Claude)
• Hands-on: observe Gemini hooks (e.g., BeforeTool, AfterTool, BeforeAgent) in action. (Gemini)
• Hands-on: configure shared instructions and observe the impact on recommendations. (Copilot)
• Hands-on: consult shared references during a command/agent to improve context. (Claude/Copilot/Gemini)
• Hands-on: analyze a risky diff and discuss the merge decision. (Claude/Copilot/Gemini)

07- Risks, Responsibilities, and Perspectives

• The importance of the human-in-the-loop
• Maintaining classic development skills and loss of technical knowledge: myth or reality?
• Dependency on AI tools: risks and mitigation
• Security: code review, vulnerabilities, sensitive data
• Intellectual property and compliance
• Ethics and responsibility of the augmented developer
• Future perspectives

08- SFEIR Certified AI-Augmented Developer Certification Exam

• Online exam (MCQ type) with 20 practical questions covering the topics addressed during the training. Minimum passing score: 80%.

• Practical mastery of at least one programming language (TypeScript, Java, Python, JavaScript, C#, Go...).
• Daily experience with Git and the use of a modern IDE (VS Code, IntelliJ, WebStorm, etc.).
• Possess basic command-line and file editing skills.
• Teamwork experience: code review, collaborative workflows.
• A sensitivity to generative AI and prompt engineering is an asset to maximize the benefits of the training.
• A standard laptop (16 GB recommended) with rights to install software.
• Recent operating system (Windows 10+, macOS 10.15+, Linux).
• Git installed and configured with access to GitHub/GitLab.
• Recent version of NodeJS and npm installed and configured.
• An IDE of choice (VS Code, IntelliJ, WebStorm...).
• A stable internet connection.

None

None

• Official course book provided to participants.