Codex Account Switcher

Query OpenAI Codex OAuth account quotas in OpenClaw, switch the preferred account by email/profile id, and optionally auto-switch when 5h quota drops below a...

Security Scan

The skill's files and instructions coherently implement an OpenClaw Codex account quota checker and automatic profile switcher; it reads/writes local OpenClaw auth files and queries ChatGPT's WHAM usage endpoint using stored OAuth access tokens — behavior matches the description, but it performs network calls and writes tokens into OpenClaw profiles so review before use.

Purpose & Capability

Name/description claim: manage openai-codex OAuth profiles and switch preferred account based on quota. The scripts read OpenClaw auth/profile files, query quota endpoints, and reorder auth-state.json. The declared required binary ('openclaw') and provided scripts match the stated purpose.

Instruction Scope

Runtime instructions and bundled scripts read local OpenClaw agent files (auth-profiles.json, auth-state.json, agent dir), optionally read ~/.codex/auth.json, call an external quota endpoint (https://chatgpt.com/backend-api/wham/usage) using each profile's access token, and may rewrite auth-state.json (to switch accounts). The SKILL.md promises tokens won't be printed; the code appears careful not to print token values. This is in-scope but is sensitive: the scripts do transmit local access tokens to the chatgpt.com usage endpoint (necessary for quota checks).

Install Mechanism

No install spec; the skill is instruction- and script-based and runs local Python scripts. Nothing is downloaded from external URLs or written to non-standard system locations beyond the OpenClaw agent directory and logs, so install risk is low.

Credentials

Registry metadata declares no required env vars; SKILL.md lists optional env vars the scripts use (OPENCLAW_HOME, OPENCLAW_AGENT_ID, etc.). The scripts access local auth/profile files (and may import ~/.codex/auth.json). They copy access/refresh tokens into OpenClaw profiles (codex-cli-sync) and use tokens to call the WHAM endpoint — this is proportional to the task but is sensitive access to local credentials and results in tokens being present in OpenClaw profile files.

Persistence & Privilege

Skill does not request always:true and does not attempt to modify other skills or system-wide settings beyond the OpenClaw auth state and profiles it legitimately manages. It writes backups before changing files. It can be run manually or as cron; autonomy is the platform default but not elevated here.

Assessment

This skill appears to do what it says: it reads your OpenClaw auth profiles, uses each profile's OAuth access token to query the ChatGPT WHAM usage endpoint, and can reorder auth-state.json to make another profile primary. Before installing/running: 1) Inspect the scripts locally (they are bundled) and confirm the WHAM_URL (chatgpt.com) is acceptable for your environment. 2) Understand codex-cli-sync will import tokens from ~/.codex/auth.json into OpenClaw profiles (tokens end up on disk in auth-profiles.json) — ensure you trust the machine and repository. 3) Run with --dry-run first to confirm behavior and check backups created (profiles/state backups are made). 4) Confirm OpenClaw and the scripts run under an account with appropriate permissions; consider limiting runtime to a dedicated user or container if you are nervous. 5) If you use notifications, verify NapCat/sendUrl configuration so messages do not leak account details to unintended recipients. If you want higher assurance, run the scripts manually from a secure shell and audit network traffic during a dry-run.

Use the bundled scripts to manage openai-codex OAuth profiles without exposing tokens.

Commands

From this skill directory:

python3 scripts/openclaw-accounts-query.py
python3 scripts/openclaw-account-switch.py [e-mail] --dry-run
python3 scripts/openclaw-account-switch.py [e-mail]
python3 scripts/codex-cli-sync.py --dry-run
python3 scripts/codex-cli-sync.py
python3 scripts/openai-codex-auto-switch.py --dry-run
python3 scripts/openai-codex-quota-query.py --json

Behavior

• Reads OpenClaw auth profile metadata from the selected agent directory.
• Queries quota directly from ChatGPT WHAM usage using each profile's OAuth access token.
• Never prints access tokens, refresh tokens, API keys, or credential file contents.
• Account switching only rewrites auth-state.json provider order for openai-codex.
• codex-cli-sync.py is advanced/explicit: it imports the current Codex CLI ~/.codex/auth.json login into OpenClaw and writes backups first.
• Auto-switch defaults to switching only when the active account's 5h remaining quota is below 20%.

Useful environment variables

export OPENCLAW_HOME="$HOME/.openclaw"
export OPENCLAW_AGENT_ID="main"
export OPENCLAW_CODEX_SWITCH_THRESHOLD=20
export OPENCLAW_CODEX_SWITCH_SUFFICIENT_THRESHOLD=20

For cron, use an absolute path to scripts/openai-codex-auto-switch.py and redirect output to logs.

Safety

Do not paste script output containing local paths into public places without review. Do not commit auth-state.json, auth-profiles.json, .env, logs, or screenshots containing account emails unless you intend to share them.