AI curriculum
Course outline for Agent Security
An agent's access must remain narrower than its ability to propose actions. Design identity, permission, and isolation controls for delegated tool execution.
About Agent Security
An agent's access must remain narrower than its ability to propose actions. Design identity, permission, and isolation controls for delegated tool execution.
Agent Security Course Objectives
- Define agent identity and delegated authority.
- Specify least-privilege tool permissions.
- Test denied actions and credential handling boundaries.
Pre-requisites
- Familiarity with agent tools and APIs.
- Basic authentication and authorization knowledge.
Lab Setup
- Computer with a local mock API and a disposable execution workspace.
- Use dummy credentials and synthetic identities; never use production tokens or accounts.
Detailed Course Outline
Proposed modulesModule 1: Identity and access
- Agent identity
- Authentication
- Authorization
Practical outcome: Define identities and access decisions for agent actions.
Module 2: Delegated permissions
- Delegation
- Least privilege
- Credential handling
Practical outcome: Restrict delegated authority and keep credentials outside model context.
Module 3: Execution trust
- Tool permissions
- Sandboxing
- Inter-agent trust
Practical outcome: Specify isolation and trust checks for tool execution.
Practical exercise
- Build or simulate a permission gate and verify that unauthorized delegated actions are rejected.
How we train
Contact us for full course details, including duration, delivery options and lab requirements.
Ask about exercises, instructor feedback and the prior knowledge you need. Tool-specific courses marked provisional may become modules in a broader course.
Discuss your learning goals