Problem

Codex native chat currently has no next-prompt suggestions after an answer. Users must write a follow-up from scratch or copy text from the reply.

Proposed behavior

Show a small set of optional Codex-authored suggestions after the latest completed answer. Clicking a suggestion should fill the editable composer draft and focus it. It must never send a message or run an action automatically. Preserve existing drafts, attachments, IME composition, queued work, and unavailable sessions.

Scope

Codex structured/native chat first; no terminal scraping, extra model request, global configuration edit, or changes to other agents. A minimal implementation can ask newly created Codex threads for a bounded optional trailing JSON block, then render it as draft-only buttons. Existing resumed threads should retain their saved instructions. If the provider cannot read effective instructions or omits suggestions, normal chat continues.

Related: #22316 and #22317 cover Claude-originated suggestions; this request is specifically for Codex.

I am preparing a focused implementation and tests. This issue and the implementation are AI-assisted with OpenAI Codex.