Modern Agentic AI Engineering and Security - Joseph Mlodzianowski - DCTFall26

Name of Training: Modern Agentic AI Engineering and Security
Trainer(s): Joseph Mlodzianowski
Dates: Oct 1-2, 2026
Venue: TBD
Early Bird Cost: $2500 SGD (w/GST)
Regular Cost: $2,950 SGD (w/GST)
Proficiency Exam Add-on: $450 SGD (w/GST)

Short Summary:

This two-day intensive bootcamp delivers practical, hands-on training tooling in an lab-driven bootcamp that will take a student from beginner to intermediate, with some advanced capabilities.

Course Description:

The course is designed to  starts with an overview of the main two AI coding agents, with a heavy emphasis on OpenAI Codex, and secondary on Claude Code. On the first day you will be immersed in to moving from coding to code to using coding agents to plan, build, verify, troubleshoot and fix simple to more complex issues. Modern AI coding agents can do more than write code: they can remember project context, read instructions, call tools, use app integrations, consult MCP servers, and coordinate specialized sub-agents. This course shows how to put those pieces together without creating an ungoverned automation mess.

Course Outline:

Chapter 1 — The Modern AI Software Engineering Stack

Students learn:

• How AI-assisted development has evolved beyond autocomplete
• The model–agent–tool–workflow architecture
• Codex interfaces and repository-aware development
• Repository instructions and external tool integrations
• Security, Sandboxing, approvals, least privilege, and human review
• How to analyze and safely approach an unfamiliar repository

Lab: Analyze, bound, and validate a repository change

Chapter 2 — Installing and Configuring Codex

Students learn:

• Installation and runtime considerations
• Authentication and credential boundaries
• User-level, project-levels and security configuration
• Configuration precedence and repository trust
• Advanced Sandbox and approval settings
• Diagnosing an effective Codex configuration without exposing secrets

Lab: Verify and explain a safe Codex setup

Chapter 3 — Context Engineering

Students learn:

• Why incomplete or excessive context causes failures
• Sources of repository and task context
• Selecting relevant files, tests, errors, and documentation
• Defining assumptions, constraints, and exclusions
• Constructing a structured context packet
• Balancing completeness with context discipline

Lab: Build and validate a bounded context packet

Chapter 4 — Mastering AGENTS.md

Students learn:

• How durable repository guidance works
• Instructions, discovery and precedence
• Repository-level and directory-level guidance
• Writing concise and actionable engineering instructions
• Security controls for Agents
• Handling conflicting, outdated, or oversized guidance
• Separating behavioral instructions from enforcement controls

Lab: Author and verify repository and subtree guidance

Chapter 5 — Steering Codex

Students learn:

• Engineering task postures such as Explore, Plan, Implement, Test, and Review
• Defining deliverables, constraints, validation, and stopping conditions
• Planning before complex or ambiguous implementation
• Security Considerations and Moving safely between task postures
• Escalation and retry boundaries
• Evidence-based completion decisions

Lab: Build and validate an evidence-gated steering plan

Chapter 6 — Cost-Effective Models and Modes

Students learn:

• Evaluating work by complexity, ambiguity, and risk
• Selecting appropriate reasoning and execution approaches
• Measuring latency, retries, context, and review effort
• Considering total engineering cost instead of model cost alone
• Defining escalation thresholds and budgets
• Avoiding unsupported cost-saving claims
• Security models and secure approaches

Lab: Build a measurable model-role decision 

Chapter 7 — Advanced Codex Development

Students learn:

• Breaking complex work into dependency-aware tasks
• Planning repository changes before implementation
• Using branches and isolated checkouts appropriately
• Implementing changes within bounded scope
• Security and Layered testing and validation
• Reviewing diffs, addressing findings, and preparing handoffs

Lab: Plan, implement, test, review, and hand off a bounded change

Chapter 8 — MCP Servers and Tool Integration

Students learn:

• MCP host, client, server, and tool architecture
• Capability discovery and tool contracts
• Transport, lifecycle, schema, and result handling
• Identity, data, side effects, and authorization boundaries
• Configuring and operating MCP integrations
• Monitoring, disabling, and safely retiring integrations

Core lab: Inspect, invoke, constrain, validate, and retire a local MCP tool

Optional Lab 8.1 — Five Engineering MCP Servers

Students explore:

• OpenAI documentation integration
• Version-aware library documentation
• GitHub repository workflows
• Browser-based engineering workflows
• Design-system and Figma workflows
• Canaries, monitoring, disablement, and removal

Lab: Deploy and monitor five engineering MCP servers

Recommended Lab 8.2 — Supply-Chain Security

Students learn:

• Server admission and ownership checks
• Source and dependency verification
• Version and artifact integrity
• Threat modeling before installation

Lab: Threat-model and evaluate an MCP server before admission

Supplemental Lab 8.3 — Identity and Authorization

Students learn:

• Authentication versus authorization
• Least-privilege scopes
• Token audience and identity separation
• Testing permitted and denied behavior

Lab: Design and verify MCP identity and authorization boundaries

Supplemental Lab 8.4 — Runtime and Secret Boundaries

Students learn:

• Transport and process isolation
• Environment and secret handling
• Filesystem and network restrictions
• Secure runtime configuration

Lab: Harden MCP transport, runtime, and secret boundaries

Supplemental Lab 8.5 — Tool and Side-Effect Safety

Students learn:

• Tool schema and contract review
• Input and output validation
• Read versus write operations
• Approval boundaries and dangerous side effects

Lab: Evaluate tool contracts, content, and side effects

Supplemental Lab 8.6 — Monitoring and Incident Response

Students learn:

• MCP health and security monitoring
• Detecting unexpected tools or behavior
• Security Controls and Best Practices
• Alerting and audit evidence
• Containment, recovery, and credential rotation

Lab: Monitor and respond to an MCP security incident

Chapter 9 — Claude Code as a Complementary Tool

Students learn:

• Comparing engineering tools using evidence
• Selecting the appropriate tool for a task
• Claude Code context and control concepts
• Portable repository instructions
• Avoiding conflicting simultaneous writers
• Creating governed handoffs between tools

Lab: Select and govern a complementary-tool workflow

Chapter 10 — AI Development Workflow Engineering

Students learn:

• Designing an end-to-end AI-assisted development lifecycle
• Defining states, owners, inputs, outputs, and gates
• Creating executable validation and evidence chains
• Managing permissions, failures, and escalation
• Rollback and release controls
• Workflow monitoring, economics, and improvement

Lab: Design and validate a governed AI delivery workflow

Chapter 11 — Multi-Agent and Review Workflows

Students learn:

• When parallel agents are beneficial
• Task independence and dependency graphs
• Defining agent roles, context, authority, and budgets
• Isolating parallel changes
• Identifying textual and semantic conflicts
• Independent review and finding adjudication
• Safe integration and stopping

Lab: Design and validate a safe parallel task and review workflow

Chapter 12 —  AI Agentic & Platform Engineering Rollup

Students apply the entire course through:

• A bounded enterprise change request
• Instruction and context analysis
• Surface, execution, cost, and tool decisions
• Minimal implementation
• Layered validation
• Security and least-authority controls
• Independent review and remediation
• Release, rollback, and residual-risk decisions
• A short technical defense

Agentic lab: Implement, validate, review, and defend an enterprise change

Optional Chapter 12.5 — AI Code Engineering Best Practices

What Students Should Bring:

• Laptop with at least 16 GB RAM (32 GB recommended) & 60gb free disk space
• Operating system with support and ability to run virtual machines (VMware Workstation, VirtualBox, or similar virtualization platform)
• A ChatGPT (Codex)  $20/Month plan and a separate free account plan and/or Claude Code $20 a month  and a separate free limited account plan.
• Administrative privileges on the laptop
• A Virtual Machine with Ubuntu 26.04 or newer installed and updated.

Difficulty Level:

Intermediate - The student has education, some experience in the field and familiarity with the topic being presented. The student has foundational knowledge that the course will leverage to provide practical skills on the topic.

Suggested Prerequisites:

• Students should have foundational Coding/Programing knowledge and beginner exposure to intermediate Codex and Claude Code familiarity.
• Understanding of Defense in Depth, Security Frameworks like NIST CSF 2.0

What Students Should Bring:

A laptop with browser access is ideal, preferably a personal laptop without network restricting tools.

What the Trainer Will Provide:

Details coming soon.

Trainer(s) Bio:

Joseph Mlodzianowski is an information security aficionado, author, keynote speaker and adventurer; he started multiple villages at RSA Conference, DEF CON, and BLACK HAT, among others, including founding the original Red Team Village at RSAC and Def Con. Joseph has extensive expertise in cybersecurity infrastructure and architecture with over 25 years leading large teams of Cybersecurity and Network engineers, project managers, and architects in the design, deployment, and monitoring of various ranges of technologies in highly complex and rigorous environments. He designed and deployed a number of multi-million dollar datacenters from inception to completion for the Department of Defense in support of cybersecurity program deployments and critical infrastructure. Joseph is a AI Datacenter Cybersecurity Architect for OpenAI. Previously a Secure Network Data-center Architect at the Department of Defense, and before that a Lead Cybersecurity Data Center Architect at Cisco Systems Federal, for thirteen years building very large datacenters for ML and AI systems, for the Department of Defense. His experience extends beyond Designing, architecting, and deploying nearly two hundred data centers in commercial, private and public sector as highly restrictive environments.  Follow him on X @cedoxx , Facebook and Linkedin.

Proficiency Exam Option:

Two-hour combination of hands-on project build and multiple questions with evidence.

Please reach out to training@defcon.org for any questions related to the proficiency exam and certificate option.

Registration Terms and Conditions: 

Trainings are refundable before Sept 1, 2026, minus a non-refundable processing fee of $350.

Between Sept 2, 2026 and Sept 28, 2026 partial refunds will be granted, equal to 50% of the course fee minus a processing fee of $350.

All trainings are non-refundable after Sept 28, 2026.

Training tickets may be transferred to another student. Please email us at training@defcon.org for specifics.

If a training does not reach the minimum registration requirement, it may be cancelled. In the event the training you choose is cancelled, you will be provided the option of receiving a full refund or transferring to another training (subject to availability).

Failure to attend the training without prior written notification will be considered a no-show. No refund will be given.

DEF CON Training may share student contact information, including names and emails, with the course instructor(s) to facilitate sharing of pre-work and course instructions. Instructors are required to safeguard this information and provide appropriate protection so that it is kept private. Instructors may not use student information outside the delivery of this course without the permission of the student.

By purchasing this ticket you agree to abide by the DEF CON Training Code of Conduct and the registration terms and conditions listed above.

Several breaks will be included throughout the day. Please note that food is not included.

All courses come with a certificate of completion, contingent upon attendance at all course sessions. Some courses offer an option to upgrade to a certificate of proficiency, which requires an additional purchase and sufficient performance on an end-of-course evaluation.