Get a technical second opinion through Codex CLI, with read-only consultations, optional plan-review hooks, and scoped edits only after an explicit implementation request.
codex
Use when Claude Code needs a second opinion, verification, or deeper research on technical matters. This includes researching how a library or API works, confirming implementation approaches, verifying technical assumptions, understanding complex code patterns, or getting alternative perspectives on architectural decisions. The agent leverages the Codex CLI to provide independent analysis and validation, and can implement changes when the user explicitly asks Codex to edit or fix code.
Informations de source
- Dépôt
- cathrynlavery/codex-skill
- Dernière activité de la source
- 30 octobre 2026 à 20:33
- Langue détectée de SKILL.md
- anglais
- Étoiles
- 227
- Forks
- 24
codex: technical second opinions
Ask Codex CLI for a technical second opinion in Claude Code, returning findings and proposed changes without editing by default.
Uses
Use it to check implementation approaches, investigate unfamiliar libraries or APIs, examine code patterns, or compare architecture trade-offs. Consultations produce evidence, risks, recommended checks and open questions rather than a guarantee that an approach is correct.
Prerequisites
The documented integration uses Claude Code with an installed, authenticated and configured Codex CLI. The README calls for an OpenAI API key. Automatic plan review is an optional plugin/hook integration; its shell hook uses Bash and jq. No minimum host or CLI version is established by this fixed source.
How to use
In Claude Code, provide the question or plan, project context and constraints. The README suggests: "Can you verify this approach with Codex?" Consultation runs codex exec --sandbox read-only from the current project root and returns findings and suggested changes. Commands inherit your configured Codex model unless you select one. With the optional hook configured, PostToolUse:ExitPlanMode triggers a read-only plan review. It takes the plan string or reads fallback plan files, including a depth-two project search for PLAN.md or plan.md. Only an explicit manual request to edit, implement or review and fix code uses workspace-write, with edits and checks kept within the requested scope. A plan describing edits does not authorize the hook to implement them.
Limitations
Read-only checks do not switch to write access when blocked. The source forbids sandbox bypass, danger-full-access and --full-auto, but sets no fixed approval-policy flag; it does not establish human confirmation for every command. The hook configuration has a 120-second timeout. If no plan is found, the hook runs exit 0; a Codex failure reports a skipped review and also runs exit 0. This fail-open behavior and PostToolUse registration do not prove the post/README claim of an effective review before every approval or a mandatory safety gate. Plans are placed in command arguments, and plan/code context sent to the configured online Codex service can leave the host. Read-only is not redaction or a local-only privacy guarantee; the source advises avoiding secrets but supplies no verified full-context scrubber. Model availability, response-time claims and runtime compatibility were not verified. The MIT license requires retaining copyright and permission notices and provides the software AS IS.
Options d'installation
Le prompt qui vérifie d'abord la source est sélectionné par défaut. Vous pouvez passer à une commande directe ou télécharger une copie locale.
Vérifiez les fichiers source
Lisez SKILL.md et les fichiers associés affichés par SkillsMP avant de décider de l'installer.
Affichage de SKILL.md
| name | codex |
|---|---|
| description | Use when Claude Code needs a second opinion, verification, or deeper research on technical matters. This includes researching how a library or API works, confirming implementation approaches, verifying technical assumptions, understanding complex code patterns, or getting alternative perspectives on architectural decisions. The agent leverages the Codex CLI to provide independent analysis and validation, and can implement changes when the user explicitly asks Codex to edit or fix code. |