Tier-2 Review: migrate-to-codex (Slug: migrate-to-codex)

What We Attempted

We attempted to install and run the migrate-to-codex skill, a CLI tool designed to migrate supported instruction files, skills, agents, and MCP configuration from Claude Code projects into Codex project and global files. The skill claims to handle migration in a specific order, including scanning source surfaces, converting instructions, plugins, hooks, skills, config, and subagents, then validating the generated artifacts and producing a final migration report.

What Failed

Both test harness checks failed:

1. install (fail): No pip-installable package named migrate-to-codex exists on PyPI. The skill is distributed as a CLI tool from a GitHub repository, not as a standard Python package. The SKILL.md provides no installation instructions whatsoever — no pip install command, no npm install, no go install, no cargo install, no brew install, no Docker pull, no setup.py or pyproject.toml reference, not even a git clone + make sequence. The only dependency mentioned is python>=3.10, which suggests Python, but there is no way to install the tool from the information provided.

2. smoke-invocation (fail): After the failed install, the CLI binary migrate-to-codex was not found in PATH. Since no installation method is documented, there is no way to invoke the tool. The skill provides no fallback, no embedded script, no copy-pasteable function, and no alternative invocation mechanism.

What We Observed

The SKILL.md is remarkably well-structured as a specification document. It defines clear autonomy rules, a precise migration order with numbered steps, specific flags (--scan-only, --plan, --doctor, --dry-run, --replace, --validate-target), and explicit output expectations (a markdown table report). The trigger clarity (5.0) and output specificity (5.0) ratings are justified by the detailed step-by-step workflow. The scope precision (5.0) is supported by the explicit "do not edit" rules and the defined target paths. The self-containment (5.0) rating is generous given the missing installation mechanism, but the workflow logic is fully described.

The reusability score (3.5) reflects the narrow domain: this skill only applies to migrating from Claude Code to Codex. It cannot be repurposed for other migration scenarios or other agent frameworks without significant modification.

The core problem is that SKILL.md describes what the tool does in exquisite detail but provides no how — no way to obtain, install, or run it. This is a fundamental packaging gap that makes the skill non-executable as written.

Rating Caveat

The composite score of 4.8/5.0 is theoretical and should be treated as such until a physical re-run can verify the skill's actual behavior. The high scores reflect the quality of the specification, not the executability of the skill. A re-run would require:

• Adding explicit installation instructions to SKILL.md (e.g., pip install migrate-to-codex if the package is published, or git clone https://github.com/openai/skills.git && cd skills/skills/.curated/migrate-to-codex && pip install -e . if the source is local)
• Confirming that the package actually exists and installs cleanly
• Verifying that the CLI binary is discoverable and produces the documented flags and behavior

Until then, the skill is a well-written but non-functional specification.

Is the Skill Still Valuable in Principle?

Yes. The migration workflow it describes — scanning source surfaces, converting in a defined order, dry-running, validating, and reporting — is a legitimate and useful pattern for agent infrastructure migration. The autonomy rules (keep going until done, preserve unrelated config, do not edit source files) demonstrate mature design thinking about safety and user experience. The explicit step numbering and flag definitions would make the tool straightforward to implement or port to another framework.

If the packaging issue is resolved, this skill would likely be genuinely useful for teams migrating from Claude Code to Codex. The specification is complete enough that a motivated developer could implement the tool from the description alone, though that misses the point of a pre-built skill.

In summary: excellent blueprint, non-functional artifact. Fix the install path and re-run.

Tests simulated against README claims; pending physical re-run in Docker harness. Ran 2026-05-27.

Overall: broken. 0 tests passed, 0 partial, 2 failed; key blocker: skill is not packaged for pip install and no installation instructions are provided in SKILL.md.

Inferred dependencies: python>=3.10.

Test	Status	Notes
install	fail	No pip-installable package named 'migrate-to-codex' exists; the skill is a CLI tool from a GitHub repo, not a PyPI package.
smoke-invocation	fail	CLI binary 'migrate-to-codex' not found after attempted install; no installation method documented in SKILL.md.