Author: Olivier Vitrac, PhD, HDR — olivier.vitrac@adservio.fr
Institutional context: Adservio Innovation Lab · Applied Artificial Intelligence & Engineering Sciences
Academic level: Graduate / Professional Development
Estimated total duration: ~6–8 hours (excluding practical exercises)
This lecture series introduces the emergence of AI-assisted software engineering, from conversational code agents to autonomous auditing systems.
Students and professionals will explore how Claude Code (Pro/Max) and the Model Context Protocol (MCP) enable reproducible, auditable workflows in real-world development environments.
Starting with installation and toolchain integration, the series progresses toward the architecture of autonomous agents, large-context reasoning, and the design of technical audits that combine static, dynamic, and semantic analysis.
A final module focuses on evaluating technical debt, code quality, and compliance with industry standards.
By the end of the sequence, you will be able to:
Set up and operate Claude Code (Max) in VS Code and shell environments.
Differentiate between assistants, copilots, and fully autonomous agents.
Leverage the MCP protocol to integrate external tools (linters, test runners, CI/CD pipelines).
Design and execute AI-based code audits, including security and performance analyses.
Interpret and mitigate technical debt using both quantitative metrics and AI-driven recommendations.
Evaluate ethical and governance implications of machine-assisted programming.
| Type | Recommended knowledge |
|---|---|
| Programming | Intermediate proficiency in Python or similar languages |
| Software tools | Familiarity with VS Code, Git, and command-line interfaces |
| AI fundamentals | Basic understanding of LLMs, prompt engineering, and model inference |
| Systems | Linux or Unix-like environment (used for demonstrations) |
| Module | Title | Duration | Format | Link |
|---|---|---|---|---|
| Lecture 0 | Installation, setup, and shell/VS Code integration | 1 h 15 min | Demo + Slides | lecture0_install/slides.html |
| Lecture 1 | From Assistants to Agents – Practical AI Agents for Developers and Auditors | 2 h 00 min | Slides + Lab | lecture1_agents/slides.html |
| Lecture 2 | Principles and Strategies of Code Auditing – From Technical Debt to Claude Code | 2 h 30 min | Slides + Notebooks | lecture2_audit/slides.html |
| Supporting Docs | Audit prompt cheat-sheet, MCP examples, and demo notebooks | — | Reference | lecture2_audit/audit_prompts.html, lecture2_audit/claude_audit_demo.html |
Start with Lecture 0 to ensure a reproducible environment.
Follow Lecture 1 to understand architectural and conceptual frameworks.
Apply the knowledge in Lecture 2, focusing on traceability and compliance.
Extend through individual projects or audits on open-source repositories.
Briefly review Lecture 0 (installation nuances).
Focus on Lecture 1’s agent workflows and MCP usage.
Deep-dive into Lecture 2 for technical-debt quantification and automation.
Integrate learned workflows into CI/CD pipelines.
Skim installation (Lecture 0).
Concentrate on Lecture 1 sections describing model reasoning and context control.
Use Lecture 2 as a case study for LLM-based evaluation and explainability.
| Evaluation type | Description |
|---|---|
| Practical Audit Task | Run claude code audit . --rules security,style --out audit.json, interpret findings, and propose remediations. |
| Conceptual Quiz | Explain differences between static, dynamic, and semantic auditing. |
| Design Exercise | Write a YAML MCP tool spec connecting Claude Code to pytest and interpret output. |
| Reflection | Identify ethical implications of AI-assisted reviews in corporate or academic contexts. |
💡 You are encouraged to maintain a learning journal summarizing the prompts, outputs, and model behaviors encountered during each session.
Claude Code Pro / Max documentation – usage, API, and MCP specification
OpenAI Codex / GPT-4 Code Interpreter – comparative architectures
OWASP Top 10 – security standards
ISO/IEC 25010 – software quality model (maintainability, reliability, security)
Olivier Vitrac, PhD, HDR
Founder & Lead Architect — Generative Simulation Initiative
AI Specialist & Innovation Lead — Adservio Group
📧 olivier.vitrac@adservio.fr
../README.html — Project overview
../DELIVERY.html — Delivery and assessment conditions
To deepen your understanding and validate your progress, three complementary resources are provided within this lecture series:
📚 Supplementary Readings
An extended set of curated references covering agents, MCP protocols, Claude Code manuals, software auditing, AI governance, and foundational research papers.
These readings serve both as preparation for the lectures and as a follow-up path for independent study or group seminars.
🧠 Further Exercises
A collection of guided, hands-on activities for each lecture — from environment setup and MCP configuration to autonomous audit pipelines.
Each exercise includes learning objectives, deliverables, and a suggested rubric to support self- or instructor-based evaluation.
🧩 Test Your Knowledge
An interactive quiz interface designed to assess comprehension after each lecture.
Open it with a URL fragment specifying the target quiz, for example:
Lecture 0 → quiz.html#quiz=quizzes/lecture0_quiz.json
Lecture 1 → quiz.html#quiz=quizzes/lecture1_quiz.json
Lecture 2 → quiz.html#quiz=quizzes/lecture2_quiz.json
The quiz app randomizes answer order to reduce bias and supports direct linking to a specific question using#quiz=…&q=N.
Together, these resources transform the lecture suite into a full learning environment — combining conceptual knowledge, practical experimentation, and self-assessment consistent with university-level pedagogy.
| Component | License | Location |
|---|---|---|
| Code (scripts, templates, build utilities) | MIT or Apache-2.0 | LICENSES/LICENSE_CODE.txt |
| Lectures, docs, slides, synopses | CC BY-NC-SA 4.0 | LICENSES/LICENSE_DOCS.txt |
| Summary / clarification | mixed notice | LICENSES/NOTICE.txt |
Version 1.0 — November 2025
Initial integrated release generated via Claude Code (Max) for Adservio Innovation Lab.
Access to lecture source files
Log file (how the first draft lecture was assembled with Claude Code)