Proposal: Thread Reincarnation workflow for long-running Codex tasks

Hi Codex team,

I’m a heavy Codex user running long-lived, multi-thread agent workflows. One recurring problem I’ve observed is that long Codex tasks can become slow, context-heavy, and prone to forgetting critical rules, external-call methods, past failure fixes, and project gates.

To address this, I built an open-source Codex skill called thread-reincarnation-skill:

https://github.com/pplong-15/thread-reincarnation-skill

The idea is simple: when a Codex thread becomes overloaded, repeatedly blocked, or starts forgetting P0 rules, the skill creates a compact successor handoff instead of continuing the old thread indefinitely.

The workflow preserves:

• verified critical facts
• P0 safety rules
• external API/key handling methods
• known failure fixes
• output paths and project gates
• an on-demand old-thread index, without loading the full old history

It also requires the successor thread to pass a memory check before continuing, so the new thread stays fast while avoiding dangerous forgetting.

I’m sharing this because it may be useful as a reference pattern for long-running Codex workflows, especially for users managing multiple agent tasks, product builds, research pipelines, or local automation systems.

Thanks for building Codex. This skill came directly from real-world friction while using it heavily.

Best,
pplong-15