AI Red Team Academy🌍 International
Online course · Certificate included

AI Agent Security: Build and Defend

Learn how AI agents work, where their security boundaries are, and how to architect them so they hold up. Covers agent internals, threat modelling, prompt injection, access control, sandboxing, container isolation, network policies, secret management, and more. Earn a verifiable Certificate of Completion.

  • 8 modules
  • 45 lessons
  • ~8 hours

Course content

01Module 1: Foundations — How AI Agents Work6 lessons · 📝 Module quiz
  1. Welcome & Course OverviewFree5m
  2. LLMs From the Outside: Tokens, Context Windows and MessagesFree12m
  3. Agent Architecture: Loops, Tools and Memory15m
  4. RAG: Retrieval-Augmented Generation12m
  5. MCP, A2A and Agent Communication Protocols10m
  6. Security Frameworks: OWASP Top 10 for LLMs, MITRE ATLAS, NIST AI RMF10m
02Module 2: The Threat Landscape5 lessons · 📝 Module quiz
  1. Threat Modelling AI Agents: Trust Boundaries and Data Flows15m
  2. Prompt Injection: Why Instructions and Data Share One Channel15m
  3. How Jailbreaks Work: Competing Objectives and Safety Gaps12m
  4. Agent-Specific Risks: Excessive Agency, Confused Deputy, Data Leakage12m
  5. Case Studies: Real-World AI Security Incidents10m
03Module 3: Secure Agent Architecture Patterns6 lessons · 📝 Module quiz
  1. Core Principle: The LLM Is an Untrusted Component10m
  2. Least-Privilege Tool Design12m
  3. The Dual-LLM Pattern: Privileged Planner + Quarantined Reader12m
  4. Structured Outputs, Action Selectors and Schema Validation10m
  5. Human-in-the-Loop and Approval Gates8m
  6. Guardrails as Defence in Depth — Not a Security Boundary10m
04Module 4: Authentication and Authorization6 lessons · 📝 Module quiz
  1. Authentication vs Authorization: Getting the Basics Right12m
  2. Identity for AI Agents: Service Accounts, Workload Identity and OIDC15m
  3. Scoped Tokens and Short-Lived Credentials12m
  4. RBAC and ABAC Applied to Agent Actions12m
  5. Multi-Tenant Isolation: What Must Never Be Shared12m
  6. OAuth 2.0 Patterns for Agent-to-Service Communication10m
05Module 5: Sandboxing and Code Execution5 lessons · 📝 Module quiz
  1. Why Agents Run Code and What Can Go Wrong10m
  2. The Isolation Spectrum: Processes → Containers → MicroVMs → WASM15m
  3. Linux Sandboxing Primitives: Namespaces, seccomp, AppArmor15m
  4. Container Hardening Checklist: Capabilities, Read-Only FS, No-New-Privileges12m
  5. Lessons from Sandbox Escapes: What Went Wrong and Why10m
06Module 6: Containers, Orchestration and Tenant Isolation5 lessons · 📝 Module quiz
  1. Docker Hardening Baseline: CIS Benchmarks and Image Hygiene12m
  2. Kubernetes Pod Security Standards and RuntimeClass15m
  3. Ephemeral Per-Session Sandboxes and Cleanup10m
  4. Supply Chain Security: Pinned Images, SBOMs and Signing10m
  5. Tenant Boundaries: Filesystems, Caches, Vector Stores, Memory12m
07Module 7: Network Policies and Secret Management6 lessons · 📝 Module quiz
  1. Egress Control: Default-Deny and Domain Allowlists12m
  2. Building SSRF-Safe Fetch Tools: Resolve, Validate, Block12m
  3. Cloud Metadata Protection: IMDSv2 and Blocking 169.254.169.25410m
  4. Rule #1: Secrets Never Enter the LLM Context10m
  5. The Credential Broker Pattern: Injecting Secrets Outside the Sandbox12m
  6. Secret Managers, Vault, and Workload Identity Federation10m
08Module 8: Testing, Monitoring and Operations6 lessons · 📝 Module quiz
  1. Security Testing Tools: Garak, PyRIT, Promptfoo12m
  2. Evaluating Guardrails: Llama Guard, NeMo, LLM Guard10m
  3. Logging and Observability for AI Agents10m
  4. Incident Response for AI Systems10m
  5. Compliance: Data Protection (NDPA, Kenya DPA), AI Regulations10m
  6. Course Summary and Next Steps5m

What you get from this course

Practical, in-demand skills

Build, break and secure AI agents.

A Certificate of Completion

Your name, a unique ID, a public verify page.

Proof you can share

LinkedIn, WhatsApp, job applications.

Learn at your own pace

No deadlines. Study on your phone.

Topics you'll cover

  • Foundations — How AI Agents Work
  • The Threat Landscape
  • Secure Agent Architecture Patterns
  • Authentication and Authorization
  • Sandboxing and Code Execution
  • Containers, Orchestration and Tenant Isolation
  • Network Policies and Secret Management
  • Testing, Monitoring and Operations

Your certificate

Certificate of Completion

This certifies that

Your Name

has successfully completed the course

AI Agent Security: Build and Defend
17 November 2026Date issuedAI Red Team AcademyIssued byXXXX-XXXX-XXXXCertificate ID

Verify at https://en.airedteam.africa/verify/…

Sample — yours shows your name, date and a verifiable ID. A Certificate of Completion shows you finished the course and passed its quizzes. It is not a professional accreditation.