Hey, it's Cindy 🌱 You commented route, so here is the official OpenAI plugin setup and a routing prompt that makes Claude the planner and reviewer while Codex handles a bounded implementation job.
btw the waitlist for AI System Club is now open! If you're keen on using AI to automate your content, video editing, job search, business or build AI agents, feel free to sign up (only if you want to hehe). Ps. the waitlist gets first dibs:)
the outcome
What you get
The official Codex plugin installed in Claude Code, a verified first call, and a reusable routing skill with clear review and stop conditions.
Use the guide in order. Gather the real inputs, run the prompts against those inputs, inspect the evidence, then make one reviewable change at a time.
start here
Set it up step by step
Confirm Node.js 18.18 or newer and either a ChatGPT subscription with Codex access or an OpenAI API key.
Add OpenAI's plugin marketplace, install codex@openai-codex, reload plugins, and run /codex:setup.
Run a read-only review first. Use /codex:review --background, then /codex:status and /codex:result.
Create the route skill below. Keep one system as planner and reviewer, and give Codex a bounded task with acceptance tests.
Stop the loop after a defined number of revisions or when a test fails for an external reason. Escalate the blocker instead of burning usage.
Create a reusable skill called route. When I invoke /route with a task, first interview me only for missing requirements that would materially change the implementation. Then write a short plan, acceptance criteria, protected files, and a maximum of two implementation-review cycles. Delegate the bounded implementation to Codex through the official plugin. Require Codex to report changed files, tests run, failures, and unresolved risks. Review its diff against the acceptance criteria and existing repository instructions. If the work fails review, send one precise correction. Never widen scope, expose secrets, run destructive commands, or claim success while tests are failing. Return the final diff summary and evidence.
check the work
Quality gate before you trust it
1 /codex:setup reports that Codex is ready.
2 The first delegated job is read-only and its result can be retrieved.
3 The route skill defines scope, acceptance criteria, protected files, and a loop limit.
4 Usage is tracked under the connected Codex account; the guide makes no promise of unlimited or free usage.
5 Claude reviews the actual diff and test output, not just the executor's summary.
6 Save the source links, as-of date, and final evidence beside the output so another person can reproduce the check.
Guardrail. Running two agents can improve separation of duties, but it also multiplies context, tool access, and failure modes. Keep tasks bounded, protect secrets and production systems, and require human approval for destructive, financial, legal, privacy, or deployment actions.
keep it useful
Save the evidence with the workflow
Save the prompt with the source links and the date you verified them. When the tool, policy, price, interface, or documentation changes, update the guide before you reuse the workflow. Keep the inputs and outputs beside the project so the next run can be checked instead of reconstructed from memory.
the waitlist is now open
AI System Club is launching soon
Build AI systems with me through recorded lessons you can follow at your own pace. Join the waitlist for first access and the founding price.
btw this list gets first dibs the moment doors open :)