[Question]: It seems that the TodoWrite tool is disabled for GPT-series models?

[CODZR]

Prerequisites

• I have searched existing issues and discussions
• I have read the documentation
• This is a question (not a bug report or feature request)

Question

When using Oh My OpenCode, the Codex agent reported: "There's a conflict about using TODO tools—git commit instructions forbid TodoWrite, and no TodoWrite tool is available."

It seems that the TodoWrite tool is disabled for GPT-series models?

• What could we do?
• Or was this caused by my mistake?

Context

You could use any GPT-5 series model and specify that it uses TodoWrite to run a task to reproduce the result.

Doctor Output (Optional)

Question Category

Agent Usage

Additional Information

No response

[code-yeongyu]

This is a known limitation with OpenAI's Codex models and OpenCode.

Root Cause: OpenAI Codex models have a different tool schema format than standard chat models. The TodoWrite tool may not be properly formatted for Codex's expected input schema.

Workarounds:

1. Use GPT-5.2 (non-codex) for main agent:

{
  "agents": {
    "Sisyphus": {
      "model": "openai/gpt-5.2"
    }
  },
  "categories": {
    "ultrabrain": {
      "model": "openai/gpt-5.2-codex"
    }
  }
}

2. Disable todo-continuation-enforcer (not recommended):

{
  "disabled_hooks": ["todo-continuation-enforcer"]
}

3. Use Claude or Gemini for orchestration, delegate coding tasks to Codex:

{
  "agents": {
    "Sisyphus": {
      "model": "anthropic/claude-sonnet-4-5"
    }
  },
  "categories": {
    "ultrabrain": {
      "model": "openai/gpt-5.2-codex"
    }
  }
}

This is an upstream compatibility issue between OpenCode's tool definitions and Codex's expected format.