codex-multi-subscription-auth-fallbacks_skill

Overview

This skill sets up multi-provider authentication for OpenClaw with OpenAI Codex OAuth fallback profiles and automatic model switching. It helps import multiple Codex OAuth sessions via device-flow, configure provider failover order, and optionally deploy a cron job that auto-switches models when a provider is on cooldown. Use it to achieve seamless rate-limit failover across multiple OpenAI Codex accounts.

How this skill works

The supplied scripts perform a safe device-flow login for each OpenAI account, extract access and refresh tokens from the Codex CLI auth file, and write timestamped backups before updating OpenClaw's auth-profiles.json. OpenClaw is then configured with profile order and model fallbacks so it will retry other profiles when one hits a rate limit. An optional cron job polls model cooldown state and updates the active model automatically when needed.

When to use it

• You run multiple OpenAI Codex accounts and want automatic failover when one hits rate limits.
• You need to add new Codex OAuth profiles reliably using device-flow (browser-based) login.
• You want OpenClaw to prefer one model/provider but transparently switch when it hits cooldown.
• You plan to run a scheduled job that monitors cooldowns and flips models without manual intervention.
• You want local-only token handling with safe backups during profile import.

Best practices

• Back up ~/.codex/auth.json and OpenClaw config files manually before the first run, even though the script creates timestamped backups.
• Test the add-profile script with a non-production account first to verify the flow and restores.
• Give short, clear profile names (e.g., usernames) so fallback order and logs remain readable.
• Verify openclaw models status manually before enabling the cron job to confirm profiles are working.
• Limit cron auto-switching to environments where automatic model changes are acceptable, and monitor its logs for unexpected switches.

Example use cases

• Import three OpenAI Codex accounts to provide continuous capacity for a high-throughput agent and let OpenClaw fail over across them.
• Add a newly provisioned Codex OAuth profile via device-flow when an account is rotated or newly purchased.
• Deploy a 10-minute cron job that monitors cooldowns and promotes the best available model to minimize latency interruptions.
• Set provider priority so Anthropic or a preferred Codex session is tried first, then fall back to other Codex profiles when needed.
• Maintain local-only token handling while consolidating multiple provider sessions into OpenClaw's auth-profiles structure.