[FEATURE]: Opt-in assistant auto-continue for generic continuation prompts #18156

Feature hasn't been suggested before.

• I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

Today OpenCode can auto-handle some permission and question flows, but it still pauses on low-information assistant prompts like:

• "If you'd like, I can continue."
• "Let me know if you want me to proceed."

That creates an unnecessary extra user turn in long coding sessions and unattended opencode run workflows, even when the assistant is clearly waiting for a generic continue signal rather than real user input.

Requested behavior:

• add an opt-in config such as experimental.auto_continue
• listen for finalized assistant messages
• match a narrow, configurable set of continuation phrases against the assistant's closing paragraph
• dedupe by assistant message ID
• submit a synthetic follow-up such as Yes. Do this.

Why this helps:

• removes redundant confirmation turns
• keeps autonomous and CI-like workflows moving
• fits existing async prompt and synthetic-message infrastructure
• stays safe if the behavior is opt-in, phrase-matched, deduped, and configurable

This is different from finish_reason: "length" auto-resume. That token-limit case is already discussed in #17471. This request is about assistants that voluntarily end a turn with generic continuation phrasing even though no substantive user decision is needed.

Related context:

• [FEATURE]: Auto-continue when model hits output token limit (finish_reason: "length") #17471 is the closest adjacent feature request, but it only covers output truncation
• automatically trigger another message #4732 shows why this should stay narrow and configurable to avoid unwanted extra work
• experimental.continue_loop_on_deny is a useful precedent for making this style of continuation behavior opt-in