Pair: Make Claude and Codex Check Each Other's Work
A free Claude Code skill that splits one job between two models: Claude plans and Codex audits, then they swap and Codex builds while Claude audits. What it is, setup in four steps, the catch, and the full skill
Pair is a Claude Code skill that stops one model from grading its own work. Claude writes a plan, a second model (Codex) audits it until it holds up, then they trade places: Codex builds and Claude audits the build.
Here's how it works, how to set it up, what it cost to run, and the full skill file to copy.
What happens in a run
- Claude writes PLAN.md: goal, steps, files, how to test.
- Codex audits it read-only and ends with VERDICT: PASS or VERDICT: BLOCK. On BLOCK, Claude fixes the plan and Codex audits again, up to 3 times.
- Codex builds from the passed plan and writes BUILD-NOTES.md.
- Claude checks every changed file against the plan and runs the tests. Failures go to FIXES.md and back to Codex.
In the test run, Claude's plan said to replace every non-letter with a dash. Codex pointed out that "Hello, World!" would become "hello--world", before a line of code existed. Two more audits added edge-case tests, then Codex built it and Claude confirmed the tests passed.
Free, for your email
Get the full setup steps + the copy-paste setup prompt
The rest of this guide (4 more sections, 1 copy-paste prompt) opens right here. The setup prompt, one paste into Claude that sets it all up, comes by email only. Use your real address.
Unlocks: Setup · The catch · The skill (copy into SKILL.md) · Sources
50,000+ follow where these guides come from. No spam, reply “unsubscribe” any time.