# Codex Agents Md Authoring > Author or audit Codex AGENTS.md files, including user-level and project-level instruction scope, plan persistence, large-plan splitting, hierarchy, context budget, and what belongs in skills, rules, hooks, or subagents instead. Use when this capability is needed. - Skill: `tomevault-io/codex-agents-md-authoring` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/codex-agents-md-authoring` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/codex-agents-md-authoring/raw - Safety review: pending (external: skill-scanner PASS, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Security - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/codex-agents-md-authoring --- # codex-agents-md-authoring Use before editing `AGENTS.md` files or rules that change how Codex reads standing instructions. ## Procedure 1. Read `references/agents-md-scope.md` before changing instruction scope. 2. Read `references/plan-splitting.md` when a task has multiple workflow types, subsystems, or phase contracts. 3. Keep `AGENTS.md` short, stable, and cross-project safe. 4. Move repeatable procedures to skills, command policy to `.codex/rules`, and hard enforcement to hooks. 5. Do not store volatile task state in `AGENTS.md`; put it in timestamped plans or workspace artifacts. 6. For large plans, require a parent plan that lists child plan paths. --- > Source: [ryugen04/dotfiles](https://github.com/ryugen04/dotfiles) — distributed by [TomeVault](https://tomevault.io).