agents · review · first seen Oct 24, 01:28 UTC

OpenAI adds Codex presets and inline settings

Codex 0.164.0 shipped reusable prompt slots, inline model and permission settings, image previews, and a default-enabled Amazon Bedrock setup wizard. Builders should care if these remove repeated setup from a terminal workflow they already use.

L0 · primary1 source4 claims checked4 min read

What shipped

Codex 0.164.0 shipped ten numbered /presets slots for reusable prompts and slash commands. The release notes also describe a keyboard menu for quick access. The concrete change is a place to keep commands you would otherwise enter again; the supplied notes do not specify how presets are shared or stored. [1]

The release shipped inline arguments for /model and /permissions. They let users set model, reasoning effort, speed, and permission profiles directly. That makes these settings available at the point of command entry, rather than requiring a builder to infer the active choice from a separate description of the workflow. The excerpt does not give argument syntax or list the available profiles. [1]

The fullscreen transcript now previews viewed and generated images on terminals supporting Kitty or Sixel graphics. This is a terminal-dependent feature, not a claim that every terminal displays images. The Amazon Bedrock setup wizard is also enabled by default. The release notes say API key entry is grouped with other AWS sign-in methods when credentials are detected; they do not establish how the wizard behaves when credentials are not detected. [1][2]

Why it matters if you build

If you repeat a prompt or slash command while working, /presets gives you a direct candidate for reducing re-entry. Ten slots set a clear capacity for what you can keep readily available. A useful trial is to put one repeated instruction in a slot and compare the resulting workflow with entering it manually. The notes establish the slots and menu, not any measured time saving or a guarantee that a preset suits every task. [1]

Inline settings matter when you deliberately change how a task runs. You can choose model, reasoning effort, speed, and permission profiles through the shipped commands, then check whether that choice fits the work at hand. Treat the controls as explicit choices, not as evidence of a particular cost, latency, or access outcome. None of those outcomes is quantified in the supplied material. [1]

Image previews may help when the transcript contains images you viewed or generated and your terminal supports Kitty or Sixel graphics. Check that condition before making the preview part of a review process. The default-enabled Bedrock wizard is relevant if you are setting up that route; the excerpt’s detail about grouped API key entry is conditional on detected credentials. Neither change, on its own, establishes broader terminal support or a completed Bedrock connection. [1][2]

How to try it today

1. Open Codex 0.164.0 and save a prompt or slash command in one of the ten numbered /presets slots.
2. Use the /presets keyboard menu to retrieve that entry, then decide whether it is worth keeping.
3. Use inline /model and /permissions arguments to set the model, reasoning effort, speed, or permission profile you intend for a task.
4. If your terminal supports Kitty or Sixel graphics, view or generate an image and inspect its preview in the fullscreen transcript.
5. If you use Amazon Bedrock, inspect the default-enabled setup wizard and the sign-in choices it presents in your situation.

What we checked

Every claim below was walked down the source ladder. The rung is how close the source is to the thing itself; the tense is the source's own.

1. shipped L0 · primary

   Codex 0.164.0 adds ten numbered /presets slots for reusable prompts and slash commands.

   per GitHub releases · openai/codex (L0)
2. shipped L0 · primary

   Codex 0.164.0 lets users set model, reasoning effort, speed, and permission profiles through inline /model and /permissions arguments.

   per GitHub releases · openai/codex (L0)
3. shipped L0 · primary

   Codex 0.164.0 previews viewed and generated images in the fullscreen transcript on terminals supporting Kitty or Sixel graphics.

   per GitHub releases · openai/codex (L0)
4. shipped L0 · primary

   The Amazon Bedrock setup wizard is enabled by default in Codex 0.164.0.

   per GitHub releases · openai/codex (L0)

What is not confirmed

The supplied release excerpt does not specify preset storage, sharing, or behavior beyond ten numbered slots and a keyboard menu. It also does not provide the exact inline argument syntax, the available models or profiles, or measured effects on cost and speed. Do not turn the presence of controls into a claim about those outcomes. [1][2]

The sources do not establish image preview support outside terminals with Kitty or Sixel graphics. They also do not say that enabling the Bedrock wizard by default completes authentication, or that API key entry appears when credentials are not detected. Those are boundaries to check in your own setup, not shipped guarantees stated here. [1][2]

Who should act, who should wait

Move now if you already use Codex in a terminal and have a repeated prompt, a setting you need to choose explicitly, or images to inspect on a supported terminal. Each gives you a specific shipped behavior to test today. Wait before changing a team workflow around shared presets, assumed cost or speed effects, unsupported terminal graphics, or automatic Bedrock authentication: the supplied notes do not establish those properties. [1][2][3][4]

Sources

• 0.164.0 — GitHub releases · openai/codex L0 · primary