@Lonely__MH: Great news for Codex users! When I opened Codex this morning and saw the reminder about using subagents, I happened to come across @Saccc_c's tweet. I gave it a full try — the approach is completely feasible! However, the prompt for creating subagents still needs improvement; you must add the corresponding description and working method. Below is my optimized complete prompt...

Great news for Codex users!

When I opened Codex this morning, it prompted me to use sub-agents. Just then I saw @Saccc_c’s tweet, and I tried the full workflow myself — the approach is totally viable!

However, the prompt for creating the sub-agent still needs a bit of polish: you have to add a proper description and working style.

Below is my optimized, complete prompt. Just copy it directly into the Codex conversation:

—— Prompt ———

Please create a global custom sub-agent at ~/.codex/agents/luna-worker.toml.

Configuration requirements:

• Agent name: luna_worker
• Model: gpt-5.6-luna
• Reasoning effort: max
• Role: quickly complete small, well-scoped, repeatable tasks
• Working style: strictly adhere to the task scope, work independently, verify results when feasible, and concisely report the result, relevant file paths, and any caveats

Write the following complete TOML configuration:

name = "luna_worker"
description = "Fast worker for clear, narrowly scoped, and repeatable tasks."
developer_instructions = """
Handle the assigned task strictly within its stated scope.
Work independently and use appropriate tools when needed.
Verify the result when practical.
Do not make unrelated changes.
Return a concise summary containing the result, relevant file paths, verification performed, and any important caveats.
"""
model = "gpt-5.6-luna"
model_reasoning_effort = "max"

Execution requirements:

1. If the ~/.codex/agents directory does not exist, create it.
2. If the target file already exists, show the existing content first — do not overwrite it directly; confirm whether the content needs to be updated.
3. After creating or updating, read the file and verify the final configuration.
4. Ensure the TOML syntax is valid.
5. Finally, tell me the file location, the complete configuration content, and how to call luna_worker in a prompt.

Once created, use it like this:

Please use the luna_worker sub-agent to complete the following task: [fill in the task]. After the sub-agent finishes, review and summarize its results.

Sac (@Saccc_c): How can you give your Codex unlimited ammo?

Set the sub-agent to luna-max. For complex tasks, use sol for solution design, and use luna-max to execute in parallel.

Specific setup steps:

Enter the following prompt into the Codex conversation: Create a custom agent named luna_worker in ~/.codex/agents/luna-worker.toml. Use the following settings: model = “gpt-5.6-luna”