---
title: "claude-in-codex"
description: "Run Claude Code from Codex for adversarial review, feedback triage, and in-scope fixes."
canonical: https://agentpluginsdirectory.com/plugins/claude-in-codex
last-updated: 2026-09-18
---

# claude-in-codex
Run Claude Code from Codex for adversarial review, feedback triage, and in-scope fixes.
- Slug: claude-in-codex
- Publisher: Alex Baker
- Repository: https://github.com/0xA3B/skills
- Manifest: plugins/claude-in-codex/plugin.json
- Version: 1.3.1
- Category (editorial): other
- Skills: 2 (adversarial-review, using-claude-cli)
- MCP servers: 0
- Stars: 0
- Repository created: 2026-05-02
- Repository last pushed: 2026-09-17
- Publisher type: User
- Listing: https://agentpluginsdirectory.com/plugins/claude-in-codex
- Schema: https://agent-plugins.org/schemas/1.0.0/plugin.schema.json

## What claude-in-codex does, in the publisher's words

Run Claude Code from Codex for adversarial code review and task delegation.

- claude-in-codex:adversarial-review: Invokes Claude Code as a review-scoped adversarial reviewer, has Codex triage the feedback, applies accepted in-scope fixes for current working-tree changes, validates the result, and summarizes the outcome.
- claude-in-codex:using-claude-cli: Internal contract for non-interactive Claude Code CLI use: command shape, permission posture, sessions, structured output, and prompting guidance. The review skill and the Claude proxy subagent use it; it is not meant for direct invocation.

Codex plugins cannot package subagent definitions, so this plugin ships a copyable one instead. Copy skills/using-claude-cli/references/claude-agent.toml to.codex/agents/claude.toml in a project, or to ~/.codex/agents/claude.toml for all projects. The copied definition lets Codex spawn Claude Code as a general-purpose proxy subagent for delegated tasks.

The definition is a work in progress.

From the project README, punctuation lightly normalized. Full text: https://raw.githubusercontent.com/0xA3B/skills/HEAD/plugins/claude-in-codex/README.md

## Skills

- adversarial-review: Use when the user explicitly asks Codex to use Claude or Claude Code to review repository code changes, the working tree, a branch, or a PR/MR, including requests that name a specific Claude model, or when another loaded skill directs a Claude review pass that the user requested. Do not use for gen…
- using-claude-cli: Internal contract for invoking the Claude Code CLI non-interactively. Use only when another claude-in-codex skill or a Claude proxy subagent needs to run claude -p, including to support a review workflow, delegate work, produce structured output, or resume or continue a prior Claude CLI session wit…

Descriptions come from the frontmatter of each SKILL.md, punctuation lightly normalized.