--- title: "claude-code-for-codex" description: "Use Claude Code from Codex to review code or delegate tasks." canonical: https://agentpluginsdirectory.com/plugins/claude-code-for-codex last-updated: 2026-09-27 --- # claude-code-for-codex Use Claude Code from Codex to review code or delegate tasks. - Slug: claude-code-for-codex - Publisher: Tien Nguyen - Repository: https://github.com/tiennguyentt/agent-plugins - Manifest: claude-code-for-codex/plugin.json - Version: 1.0.1 - License: Apache-2.0 - Category (editorial): agent-tooling - Skills: 11 (claude-adversarial-review, claude-cancel, claude-cli-runtime, claude-rescue, claude-result-handling, claude-result, claude-review, claude-setup, claude-status, claude-transfer, opus-5-prompting) - MCP servers: 0 - Stars: 0 - Repository created: 2026-07-31 - Repository last pushed: 2026-08-14 - Publisher type: User - Listing: https://agentpluginsdirectory.com/plugins/claude-code-for-codex - Schema: https://agent-plugins.org/schemas/1.0.0/plugin.schema.json ## What claude-code-for-codex does, in the publisher's words Use Claude Code from inside Codex for code reviews or to delegate tasks to Claude Code. This is the inverse of OpenAI's Codex plugin for Claude Code: same architecture, opposite direction of delegation. It is a port of that plugin, not an independent design, see NOTICE.md for attribution and the Apache-2.0 terms it inherits. Codex invokes skills with a $ prefix: - $claude-review: a read-only Claude Code review of your changes - $claude-adversarial-review: a steerable challenge review - $claude-rescue: hand a substantial task to Claude Code - $claude-transfer: hand the current Codex session across - $claude-status, $claude-result, $claude-cancel: manage background jobs - $claude-setup: check the runtime and toggle the stop-time review gate From the project README, punctuation lightly normalized. Full text: https://raw.githubusercontent.com/tiennguyentt/agent-plugins/HEAD/claude-code-for-codex/README.md ## Skills - claude-adversarial-review: Run a steerable Claude Code review that challenges the chosen implementation, design tradeoffs, and assumptions, not just a stricter pass over defects. Use for "pressure-test this design", "challenge whether this was the right approach", "review this for race conditions and question the caching st… - claude-cancel: Cancel an ACTIVE background Claude Code job in this repository, stop a rescue, review, or adversarial-review run that is still in progress. Use for "stop that review", "cancel the rescue job", or "kill the background Claude Code run." Takes an optional job ID. Do NOT use this to check whether a jo… - claude-cli-runtime: Internal helper contract for calling the claude-companion runtime from Codex. Not a user-invocable command: the `claude-rescue` agent is the only caller. Defines which claude-companion subcommand to use, how to strip routing flags before forwarding, and what the agent is and is not allowed to do b… - claude-rescue: Hand a substantial coding task to Claude Code through the claude-companion runtime, an investigation, an explicit fix request, or a follow-up on a task Claude Code already started. Use when Codex is stuck, wants a second implementation or diagnosis pass from a different model family, or should del… - claude-result-handling: Internal guidance for presenting claude-companion helper output back to the user. Not a user-invocable command: applies whenever any claude-* command-skill or the `claude-rescue` agent has helper output to hand back, defining what to preserve, what never to auto-apply, and how to report a failed o… - claude-result: Show the stored final output for a FINISHED Claude Code job in this repository, the completed verdict, summary, findings, and next steps of a rescue, review, or adversarial-review run. Use for "what did Claude Code find", "show me the review results", or "get the output of that rescue task." Takes… - claude-review: Run a read-only Claude Code review against local git state, uncommitted changes or a branch diff against a base ref. Review-only: never fixes issues or applies patches. Use for "review my changes", "have Claude Code review this before I ship", or "review this branch against main". Takes `--wait`/`… - claude-setup: Check whether the local `claude` CLI is installed, on PATH, and authenticated, and optionally toggle the stop-time review gate. Use for "is Claude Code set up", "check whether Claude Code is ready", "install Claude Code", or "turn on/off the review gate". This is a diagnostic and configuration skil… - claude-status: Show active and recent Claude Code jobs for this repository, including review-gate status, running, queued, or recently finished rescue/review/adversarial-review jobs and their phase. Use for "is Claude Code still working", "what jobs are running", "check progress on the background review", or "sh… - claude-transfer: Transfer the current Codex session into a resumable Claude Code session and print the `claude --resume ` command to continue it. Use when a debugging or implementation conversation started in Codex and should continue with the same context directly in Claude Code, "move this conversati… - opus-5-prompting: Internal guidance for composing Claude Code prompts (Opus 5 / Sonnet 5) for coding, review, diagnosis, and research tasks delegated from Codex. Not a user-invocable command: used by the `claude-rescue` agent to tighten a raw request into a Claude-Code-ready prompt before the single `task` call. Descriptions come from the frontmatter of each SKILL.md, punctuation lightly normalized.