AI curriculum

Course outline for Harness Engineering

Coding agents need controlled execution rather than unrestricted access. Design a harness with clear instructions, tool boundaries, recoverable state, and human approval.

Technology: AI-Assisted Software Engineering
Duration: 6 hours
Modules: 3 · 2 hours each
Level: Prerequisites listed

About Harness Engineering

Coding agents need controlled execution rather than unrestricted access. Design a harness with clear instructions, tool boundaries, recoverable state, and human approval.

Harness Engineering Course Objectives

Specify harness instructions and tool contracts.
Implement bounded repository access and session state.
Evaluate checkpoint recovery and approval behavior.

Pre-requisites

Programming and command-line experience.
Familiarity with agent tool calls and repository workflows.

Lab Setup

Computer with a programming runtime, Git, and a disposable workspace.
Local agent runtime or mocked tool calls; use a local sandbox where supported.
Keep production credentials out of the exercise.

Detailed Course Outline

Proposed modules

Module 1: Harness contracts

Agent harness architecture
Instructions and system prompts
Skills
Tools

Practical outcome: Define instructions and interfaces for a constrained agent.

Module 2: Execution boundaries

Context management
File and repository access
Sandboxed execution
Sessions

Practical outcome: Restrict execution to an isolated sample workspace.

Module 3: Recovery and oversight

Memory
Checkpointing
Evaluation
Human approval

Practical outcome: Test state recovery and required approval steps.

Practical exercise

Prototype a harness that performs a small change with a checkpoint and approval gate.

How we train

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