AI Coding Engineering
Treat Claude Code and Codex as a maintainable system: global and project configuration, AGENTS.md, the current prompt, skills, and hooks.
Complete this path in order
- 01 Claude Code configurationClaude Code configuration belongs in layers: user habits, repo facts, directory rules, the current prompt, skills, and hooks. Share one project source of truth with Codex. About 19 min
- 02 How to write AGENTS.mdHow to write AGENTS.md so Codex and Claude Code share one project fact: keep repo-wide rules at the root, sink module rules, and verify what actually loaded. About 17 min
- 03 How to write Claude Code promptsHow to write Claude Code prompts for this change only: current state, sources, range, authorization, and an observable result. Do not copy project rules or skills. About 20 min
- 04 How to write Claude Code skillsHow to write Claude Code skills so a repeated workflow loads on the right task: keep methods in SKILL.md, facts in project rules, this-run variables in the prompt, and prove consumers. About 18 min
- 05 How to use Claude Code hooksHow to use Claude Code hooks with Codex: reuse an existing project check, pick the right event and scope, and prove failure feedback, missed write paths, and the security boundary. About 16 min