migrate-to-codex
Migrate supported instruction files, skills, agents, and MCP config into Codex project and global files.
Security Assessment
Detected risks:
About migrate-to-codex
migrate-to-codex is a migration tool that converts supported Claude Code configuration — instruction files, slash commands/skills, subagents, hooks, and MCP server config — into the equivalent Codex project and global files (AGENTS.md, .codex/, .agents/, ~/.codex/). It solves the problem of moving a working Claude Code setup onto Codex without hand-porting each surface, and it maps sources to targets explicitly (for example .claude/agents to .codex/agents, .mcp.json to .codex/config.toml MCP servers, .claude/settings.json hooks to .codex/hooks.json).
The skill drives a Python migrator with inspection-first flags — `--scan-only`, `--plan`, `--doctor`, and `--dry-run` — before writing anything, then converts surfaces in a fixed order and validates the generated config and agents with `--validate-target`. It reads a `references/differences.md` guide to account for Claude-vs-Codex behavior gaps, reports plugins and marketplaces as manual-review work, and produces a final markdown migration report per scope. It is designed to run autonomously once a target is selected: it will create, edit, replace, or delete generated Codex artifacts in that target without pausing for confirmation, using `--replace` to remove orphaned generated skills or agents. Importantly, it is constrained to only touch generated Codex files and explicitly must not edit source Claude Code files, unrelated project code, secrets, or other repositories, and it preserves unrelated existing Codex config entries.
It targets developers moving from Claude Code to Codex. The main thing to be aware of is its autonomous file writing and deletion within the chosen target directories, which is scoped to migration-generated artifacts rather than arbitrary files.
FAQ
What does it migrate?
Instruction files (CLAUDE.md/AGENTS.md), slash commands and skills, subagents, hooks, and MCP server configuration, into Codex targets like AGENTS.md, .agents/skills/, .codex/config.toml, .codex/hooks.json, and .codex/agents/.
Will it change my Claude Code configuration?
No. It must not edit source Claude Code files (.claude/, ~/.claude/, .mcp.json, .claude.json), unrelated project code, secrets, or other repositories; it only writes generated Codex artifacts.
Does it ask before writing or deleting files?
Once you select a target it runs autonomously, creating, editing, replacing, or deleting generated Codex artifacts in that target without confirmation. Deletion of orphaned generated skills/agents happens via --replace.
How can I preview the migration first?
Use the inspection flags --scan-only, --plan, and --doctor, and --dry-run, before writing a target, then --validate-target after edits.
Are plugins migrated automatically?
No. Claude plugin trees and marketplaces are reported as manual migration work rather than converted automatically.
All Files
19 filesInstall migrate-to-codex
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
openai/skills