1. Security fundamentals
Optional. The habits everything else builds on: threat models, least privilege, logs. Skip it if you already work in security.
2. How AI agents work
What an agent is, how it calls tools, and the agents and MCP servers teams run today.
3. Threats to AI agents
How agents get attacked: prompt injection, poisoned tools, leaked credentials.
4. Inventory: find every agent
Which agents, MCP servers and extensions run in your company, and where.
5. Configuration: set what agents may do
What each agent may do without asking, and how to lock that down.
6. Identity: whose keys the agent holds
Which credentials and tokens an agent acts with, and how to scope them.
7. Runtime: stop bad actions as they happen
Hooks, sandboxes and policies that block a harmful action mid-task.
8. Drift: catch what changed
Notice when settings, versions or behaviour change after you set them.
9. Incident response: when an agent goes wrong
Work out what the agent touched, contain it and fix the cause.
Running an agent security program
Coming soon. Policy, ownership and reporting for agents across a company.
- AI acceptable use policy
- AI risk assessment and the NIST AI RMF
- Standards and control frameworks
- EU AI Act
- Agentic AI governance guidance
- Vendor and tool due diligence
Red teaming agents
Coming soon. Attack your own agents before someone else does.
- Threat-led testing
- Manual agent exploitation
- Evals for agent safety and security
- Automated scanners and frameworks
- Benchmarks
- CTFs and practice ranges