A project-based course on harness engineering for AI coding agents like Codex and Claude Code. It covers environment design, state management, verification, and control systems to make agentic coding tools reliable. The curriculum includes theoretical lectures, hands-on projects, and copy-ready templates (AGENTS.md, feature_list.json) for constraining agent behavior, maintaining context across sessions, and verifying work through full-pipeline tests.

2m read timeFrom walkinglabs.github.io
Post cover image
Table of contents
Get started ​The Core Mechanism of a Harness ​What you will learn ​Next steps ​

Sort: