codex-claude-loop
Use when driving a gated build loop where Claude orchestrates and Codex CLI implements — Claude plans and judges, Codex plans/implements from a persistent…
Run one gated Codex wave — plan, you approve, Codex implements, you review
$ npx -y skills add ozzaii/codex-claude-loop --skill wave --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/waveContext preview
The summary Claude sees to decide when to auto-load this skill.
Run one gated Codex wave — plan, you approve, Codex implements, you review
name: wave description: Run one gated Codex wave — plan, you approve, Codex implements, you review argument-hint: <slug> <brief.md> allowed-tools: Bash, Read, Write
Load the `codex-claude-loop` skill and run one wave for `$ARGUMENTS`, following the skill's cycle and its rails exactly. You are the gate, so do not hand both decisions to `cl_wave` and walk away.
Two things this command adds on top of the skill:
1. **If no brief file was given, write one first** and show it to the user: goal, constraints, acceptance criteria, out of scope. The loop can only be as good as its brief, and a vague brief produces a plan you cannot judge. 2. **Report at the end**: what landed, what you sent back and why, what is still open.
A gated build loop between Claude Code and Codex CLI. Codex writes the plan, Claude approves it, Codex implements that same plan from the same thread, Claude reviews the diff against it.
Repo: ozzaii/codex-claude-loop
Use when driving a gated build loop where Claude orchestrates and Codex CLI implements — Claude plans and judges, Codex plans/implements from a persistent…
Check the codex-claude-loop substrate (codex CLI, jq, git, repo, verdict schema)
Send an additional request to the exact Codex session for an implemented codex-claude-loop lane, sequenced behind its writer lock