preset-skill-codex-pool

Documents the codex-pool provider template and codex-auth-pool.json manual-edit protocol.

692|60|Updated Mar 15, 2026
One-click install
npx skills add https://github.com/Lingtai-AI/lingtai --skill preset-skill-codex-pool
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: preset-skill-codex-pool
Source: https://github.com/Lingtai-AI/lingtai/tree/main/tui/internal/preset/skills/lingtai-preset-skill/reference/codex-pool
Command: npx skills add https://github.com/Lingtai-AI/lingtai --skill preset-skill-codex-pool

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Managing multiple ChatGPT OAuth accounts for the Codex provider requires a load-balancing pool file, but editing codex-auth-pool.json by hand risks corrupting account classification, leaking token material, or silently breaking authentication. This Skill provides the official-source-led manual for the TUI codex-pool template and a safe manual-edit protocol.

Core Features & Use Cases

  • Pool File Format Reference: Documents the v1 flat and v2 exact-model-classified codex-auth-pool.json schemas, including refs, integer weights, and weight-0 disabled-account semantics.
  • Selection Behavior Explained: Explains deterministic, sticky weighted account selection, session-anchored seeds, and why edits do not reselect accounts for running sessions.
  • Safe Manual Edit Protocol: Defines a six-step destructive-edit discipline covering authorization, timestamped backups, hash gates, atomic rename, live validation, and rollback.
  • Use Case: When you need to add a per-model account classification to your Codex OAuth pool, follow this Skill to edit the pool file without clobbering concurrent writes or exposing token contents.

Quick Start

Ask the agent to explain the codex-auth-pool.json v2 models format and walk you through safely adding a per-model account classification with a backup and validation step.

Frequently Asked Questions about preset-skill-codex-pool

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I load-balance multiple ChatGPT OAuth accounts for Codex?

Create a codex-auth-pool.json file listing account token-file refs with integer weights. The kernel's codex-pool provider reads this file and makes a deterministic weighted selection among positive-weight accounts, falling back to the legacy single token if the pool is invalid.

What is the difference between codex-auth-pool.json v1 and v2?

v1 is a flat accounts list applied to every model, while v2 adds a top-level models dict mapping exact model strings to account lists. The presence of the models key classifies the pool; the version field is ignored by the kernel.

Does editing the pool file change the account for a running session?

No. Account selection happens at adapter construction, not per turn, and the seed is sticky within a session. An explicit refresh, relaunch, or new service construction is required to pick up pool-file changes.

Why does the TUI refuse to edit my codex-auth-pool.json?

The TUI refuses flat weight edits on classified pools containing a models key, raising errCodexPoolModelClassified. This protects hand-authored per-model classifications; edit classified pools manually following the documented protocol.

What does weight 0 mean in the Codex auth pool file?

Weight 0 keeps the account visible in the pool file and TUI but excludes it from kernel selection, since only strictly positive integer weights are eligible. A missing weight defaults to 1, and enabled: false drops the entry entirely.