realtime prompt changes (#13376)

# External (non-OpenAI) Pull Request Requirements

Before opening this Pull Request, please read the dedicated
"Contributing" markdown file or your PR may be closed:
https://github.com/openai/codex/blob/main/docs/contributing.md

If your PR conforms to our contribution guidelines, replace this text
with a detailed and high quality description of your changes.

Include a link to a bug report or enhancement request.

codex-rs/protocol/src/prompts/realtime/realtime_end.md

@@ -1,5 +1,3 @@
 Realtime conversation ended.
 
-You are still operating behind an intermediary rather than speaking to the user directly. Use the conversation transcript and current context to decide whether backend help is actually needed, and avoid verbose responses that only add latency.
-
-Subsequent user input may return to typed text rather than transcript-style text. Do not assume recognition errors or missing punctuation once realtime has ended. Resume normal chat behavior.
+Subsequent user input will return to typed text rather than transcript-style text. Do not assume recognition errors or missing punctuation once realtime has ended. Resume normal chat behavior.

codex-rs/protocol/src/prompts/realtime/realtime_start.md

@@ -6,5 +6,4 @@ When invoked, you receive the latest conversation transcript and any relevant mo
 
 When user text is routed from realtime, treat it as a transcript. It may be unpunctuated or contain recognition errors.
 
-- Ask brief clarification questions when needed.
-- Keep responses concise and action-oriented.
+- Keep responses concise and action-oriented. Your updates should help the intermediary respond to the user.