[How Coding Agents Work][06] The Codex Repo Map and Crate Graph

How Coding Agents Work · Module 06 of 29: The Codex Repo Map and Crate Graph Part I — One Concrete Turn How 109 crates and 1.5M lines are organized, which twenty crates matter for understanding the agent, and how to navigate the codebase without reading Rust. After this module you can find the code for any harness behaviour in this repo within a minute or two. Lessons in this module: 1. The shape of the repo: crates grouped by harness responsibility 2. The twenty crates that matter, and why the other eighty-nine don't for our purposes 3. The two-layer pattern: standalone crates hold the contracts, core/ holds the wiring — and why tools/ and core/src/tools/ are different things 4. Reading this codebase without reading Rust: naming conventions and file patterns 5. Where the tests live, and why test files are the best documentation in the repo Prerequisites: module 4. About the course: Built from the OpenAI Codex harness at commit 8d01cd42fa.