What does this skill do, and when should you use it?
This skill migrates selected Claude Code project or global configuration into Codex files. It scans, plans, dry-runs, writes, validates, and reviews the migration. Supported surfaces include instructions, skills, commands, hooks, configuration, MCP servers, and subagents. It is intended for developers moving an existing Claude Code workflow to Codex while preserving unrelated Codex settings.
Runs the Python migration script with scan-only, plan, doctor, dry-run, write, and validate-target modes. It writes instructions to AGENTS.md, skills to .agents/skills/, configuration and MCP servers to .codex/config.toml, hooks to .codex/hooks.json, and custom agents to .codex/agents/. It also inspects the migration report, fixes actionable generated-artifact issues, and reruns validation checks.
- A developer migrating personal ~/.claude/ configuration into ~/.codex/.
- A team converting project .claude/ skills, commands, and agents into Codex files.
- A user transferring Claude Code MCP server configuration into Codex configuration.
- A maintainer reviewing generated artifacts and manual migration items after conversion.
How do you install this skill?
- Do not treat the no-renewed-confirmation rule as safe for every target; manually confirm and back up before using --replace or overwriting Codex artifacts.
- Verify that cli.py is complete and runnable before performing the dry-run, real migration, and validate-target sequence.
- Migrated permissions, hooks, tool allowlists, OAuth/MCP settings, and Claude-specific semantics may be preserved only as prompt guidance and require manual review.
- The repository is marked deprecated; confirm licensing and the future maintenance path before adoption.
- Shell / CLI
- Local filesystem
Python 3
This is a curated skill from openai/skills. Install it with: $skill-installer migrate-to-codex. Restart Codex after installation so it can load the skill.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/openai/skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/.curated/migrate-to-codex" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Select a global or project source and target, then ask Codex to perform the migration. Example dry run: python3 .codex/skills/migrate-to-codex/scripts/migrate-to-codex.py --source ./.claude/ --target ./.codex/ --dry-run. Remove --dry-run to write the selected target, then run --validate-target against it.
What are this skill's strengths and limitations?
- Covers instructions, skills, commands, hooks, configuration, MCP servers, and subagents.
- Provides scanning, planning, diagnosis, dry-run, validation, and self-healing steps.
- Specifies preservation of unrelated Codex configuration and protection of Claude Code source files.
- The README says the repository is deprecated and points to the OpenAI Plugins repository for current examples.
- No test suite or migration success-rate evidence is provided in the supplied source.
- Plugins and marketplaces are reported as manual migration work rather than fully migrated.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Codex Migration Assistant this page ✓ OpenAI · Official | 42 · Not recommended | ★ 28k | 4mo ago | — |
| Microsoft Agent Skills ✓ Microsoft · Official | Not yet reviewed | ★ 2.7k | 3mo ago | MIT |
| AI Tools Setup Assistant | 58 · Recommended | ★ 10k | 11d ago | MIT |
| Claude Code Plugin Marketplace Template | 56 · Use with care | ★ 70 | 4mo ago | MIT |
| VS Code Agent Customization Assistant ✓ Microsoft · Official | 55 · Use with care | ★ 194k | 11d ago | MIT |
How did FollowSkills review this skill?
The evidence scopes migration, forbids changes to source files and unrelated code, and preserves unrelated configuration. However, it explicitly permits creating, replacing, or deleting generated artifacts without renewed confirmation after target selection, and --replace can delete orphaned generated artifacts; rollback, stronger confirmation, and sensitive-data flow disclosure are incomplete, so points are deducted.
The migration order, self-healing loop, reporting, and target validation are clearly specified. However, the supplied static material shows cli.py ending at parser.add_arg, so complete executability cannot be established; there is also no committed test suite or key-path execution evidence. The score is therefore low.
The audience, migration scenarios, source surfaces, and several unsupported cases are described clearly, with multiple command modes. Trigger conditions, non-fit boundaries, Chinese-language support, and mainland-China environment fit are not established, and some use depends on reachable Codex, GitHub, or official documentation services, so points are deducted.
The skill has layered documentation, references, command examples, manual-review blocks, and known mapping limitations, with official OpenAI provenance information. License metadata is unknown, the repository README marks the repository deprecated, and versioning, changelog, and explicit maintenance ownership are unclear, so points are deducted.
The skill addresses scanning, planning, migration, remediation, validation, and final reporting, and could theoretically produce directly usable migration results. Static review cannot confirm validation or output correctness, the supplied script appears incomplete, and several Claude semantics remain manual guidance; the score stays below the static ceiling.
The source files and mapping rules are auditable, and validator functions are described. There is no skill-specific test suite, CI coverage, or third-party execution evidence, so conclusions rely mainly on static inspection and evidence strength is limited.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →