| name | cook |
|---|---|
| description | Code in an Opus session, using Fable subagent to plan and Codex to write code and finally Fable subagent review for every task. |
| user-invokable | true |
Implement all tasks given to you. Use a Fable high subagent to write a plan.md for each task. Require my approval for all plans before beginning implementation. Then write all code by calling out to codex exec for each task using codex --yolo exec -c model="gpt-5.5" -c model_reasoning_effort="high" < plan.md
(allow cmd to run for up to 20 minutes), then review code (using a subagent with Fable high) and smoke testing new features (usage a subagent with Opus medium) by running pnpm dev and doing real requests, before committing and pushing, then moving onto next task. Remember that every subagent or codex session is completely independent and must be given the full context it needs to complete its work.
source & further reading
gist.github.com — original article
A plan-of-plans for a production-grade end-to-end e-commerce website.
Charte du vibe coding — Dernier Cri (gabarit à forker pour votre organisation)
OpenCode AI config to deny read access to .env, node_modules, build artifacts, cache dirs and ask before bash execution