Imported from reason-machines/codex-skills (skills/codex-auto-register/SKILL.md). Install upstream with npx skills add reason-machines/codex-skills --skill codex-auto-register. Copyright stays with the author.

Codex Auto Register

Skill by ara.so — Codex Skills collection.

Overview

codex_auto_register is a Python tool for automated ChatGPT/Codex account registration and OAuth token generation using the DuckMail temporary email API. It supports two workflows:

1. Root script (chatgpt_register.py): Register ChatGPT accounts with optional OAuth
2. Codex protocol script (codex/protocol_keygen.py): Full Codex OAuth flow with CLIProxyAPI v6 compatible token output

Installation

Dependencies

For root registration script:

pip install curl_cffi

For Codex protocol script:

pip install requests urllib3

Configuration Setup

1. Copy example configs to active configs:

cp config.example.json config.json
cp codex/config.example.json codex/config.json

2. Edit config.json and codex/config.json with your credentials (see Configuration section)

Root Script: ChatGPT Registration

Configuration (config.json)

{
  "total_accounts": 5,
  "duckmail_api_base": "https://duckmail.example.com",
  "duckmail_bearer": "${DUCKMAIL_TOKEN}",
  "proxy": "http://username:password@proxy.example.com:8080",
  "output_file": "registered_accounts.txt",
  "enable_oauth": true,
  "oauth_required": false,
  "upload_api_url": "https://cpa.example.com/api/accounts",
  "upload_api_token": "${CPA_API_TOKEN}"
}

Key Fields:

• total_accounts: Number of accounts to register
• duckmail_api_base: DuckMail API endpoint
• duckmail_bearer: DuckMail authentication token (use env var)
• proxy: HTTP/HTTPS proxy (required for registration)
• enable_oauth: Whether to perform OAuth flow
• oauth_required: Fail if OAuth fails
• upload_api_url/upload_api_token: Optional CPA upload

Usage

python chatgpt_register.py

The script will:

1. Create temporary email via DuckMail
2. Register ChatGPT account
3. Optionally perform OAuth
4. Save results to registered_accounts.txt
5. Optionally upload to CPA

Output Format

registered_accounts.txt contains:

email@example.com:password123:access_token:refresh_token
email2@example.com:password456:access_token2:refresh_token2

Codex Protocol Script

Configuration (codex/config.json)

{
  "total_accounts": 10,
  "duckmail_api_base": "https://duckmail.example.com",
  "duckmail_bearer": "${DUCKMAIL_TOKEN}",
  "proxy": "http://user:pass@proxy.example.com:8080",
  "output_dir": "codex_tokens",
  "accounts_file": "accounts.txt",
  "csv_file": "registered_accounts.csv",
  "upload_api_url": "https://cpa.example.com/api/codex",
  "upload_api_token": "${CPA_API_TOKEN}",
  "save_access_token": true,
  "save_refresh_token": true,
  "ak_file": "ak.txt",
  "rk_file": "rk.txt"
}

Usage

python codex/protocol_keygen.py

The script performs:

1. DuckMail temporary email creation
2. ChatGPT account registration
3. Codex OAuth login flow
4. Token extraction and storage
5. CLIProxyAPI v6 compatible file generation

Output Files

Token JSON files (CLIProxyAPI v6 format):

codex_tokens/
├── fk-xxxxxx.json
├── fk-yyyyyy.json
└── ...

Each JSON contains:

{
  "access_token": "ey...",
  "refresh_token": "ey...",
  "expires_at": 1234567890
}

Account files:

• accounts.txt: email:password:access_token:refresh_token
• ak.txt: Access tokens (one per line)
• rk.txt: Refresh tokens (one per line)
• registered_accounts.csv: CSV format with all fields

Common Patterns

Environment Variables for Secrets

Never hardcode credentials. Use environment variables:

export DUCKMAIL_TOKEN="your_duckmail_token"
export CPA_API_TOKEN="your_cpa_token"

Reference in config:

{
  "duckmail_bearer": "${DUCKMAIL_TOKEN}",
  "upload_api_token": "${CPA_API_TOKEN}"
}

Proxy Configuration

Both scripts require proxies to bypass rate limits:

{
  "proxy": "http://username:password@proxy-host:port"
}

For SOCKS5:

{
  "proxy": "socks5://username:password@proxy-host:port"
}

Batch Registration

Register multiple accounts in sequence:

# In your automation script
import subprocess
import json

config = {
    "total_accounts": 50,
    "duckmail_api_base": "...",
    # ... other config
}

with open("codex/config.json", "w") as f:
    json.dump(config, f)

subprocess.run(["python", "codex/protocol_keygen.py"])

Token File Naming Convention

Codex tokens follow CLIProxyAPI v6 naming:

• Format: fk-{unique_id}.json
• Compatible with CLIProxyAPI token directory structure
• Can be directly placed in CPA token folders

DuckMail API Integration

The project uses DuckMail for temporary email:

Create email:

POST {duckmail_api_base}/api/email/create
Authorization: Bearer {duckmail_bearer}

Check inbox:

GET {duckmail_api_base}/api/email/{email_id}/messages
Authorization: Bearer {duckmail_bearer}

See duckmaildoc.md for full API reference.

Troubleshooting

Registration Fails

Problem: ChatGPT registration returns errors

Solutions:

1. Verify proxy is working and not banned
2. Check DuckMail API is accessible
3. Ensure duckmail_bearer token is valid
4. Try different proxy IP range

OAuth Token Not Generated

Problem: enable_oauth: true but no tokens

Solutions:

1. Set oauth_required: false to continue without OAuth
2. Verify proxy supports OAuth endpoints
3. Check account was successfully verified
4. Inspect network logs for OAuth failures

DuckMail Email Not Received

Problem: Verification email not arriving

Solutions:

1. Wait 30-60 seconds (DuckMail polling interval)
2. Check DuckMail API rate limits
3. Verify email was successfully created
4. Try different email domain if available

CPA Upload Fails

Problem: upload_api_url returns errors

Solutions:

1. Verify upload_api_token is correct
2. Check CPA endpoint is accessible through proxy
3. Ensure token format matches CPA expectations
4. Review CPA API documentation for required fields

Token Files Not Compatible

Problem: CLIProxyAPI doesn't recognize tokens

Solutions:

1. Verify filename format: fk-*.json
2. Check JSON structure matches expected schema
3. Ensure expires_at is Unix timestamp
4. Validate tokens are not expired

Proxy Authentication Errors

Problem: 407 Proxy Authentication Required

Solutions:

1. Verify proxy username/password are correct
2. Use URL-encoded credentials in proxy string
3. Test proxy with curl: curl -x proxy http://api.openai.com
4. Check proxy supports CONNECT method for HTTPS

File Structure

Generated files (all in .gitignore):

project_root/
├── config.json                    # Root script config
├── registered_accounts.txt        # Root script output
└── codex/
    ├── config.json               # Codex script config
    ├── accounts.txt              # email:password:tokens
    ├── ak.txt                    # Access tokens only
    ├── rk.txt                    # Refresh tokens only
    ├── registered_accounts.csv   # CSV format
    └── codex_tokens/             # CLIProxyAPI compatible
        ├── fk-xxxxx.json
        └── fk-yyyyy.json

Security Notes

• Never commit config.json files (use .example.json templates)
• Store tokens securely (files are gitignored by default)
• Use environment variables for all secrets
• Rotate DuckMail tokens regularly
• Keep proxy credentials separate from code
• Review registered_accounts.txt for sensitive data before sharing logs

Integration with CLIProxyAPI

Place generated token files directly in CPA token directory:

cp codex/codex_tokens/*.json /path/to/CLIProxyAPI/tokens/codex/

CPA will automatically detect and use tokens with fk-*.json naming pattern.