[FEATURE] Add collapsible/expandable sections to long model responses in the terminal output.

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

When an agent returns a long response, the terminal scrolls far down and the user loses their place in the conversation. There is no way to collapse verbose output after reading it, or to skip past it without losing context above.

Proposed Solution

Let users collapse individual model responses (or sections within them) to a single summary line, with an option to expand on demand. The pattern already exists in the Claude Code UI: long user inputs show a truncated preview with a "more" toggle. Applying the same behavior to model output would make the experience consistent and easier to navigate.

Alternative Solutions

_No response_

Priority

High - Significant impact on productivity

Feature Category

CLI commands and flags

Use Case Example

When running multi-step agent workflows, responses stack quickly. Collapsing completed steps keeps the active context visible without scrolling. This is especially useful when an agent returns structured output (step-by-step plans, file diffs, research summaries) where most sections only need to be read once.

No new concept required. The toggle pattern is already in the product. This extends it to the other side of the conversation.

Additional Context

Specifically: when a user sends a long message to Claude Code, the interface already collapses it to a summary box showing roughly three lines of text, with a "more" button to expand the full input. That is the exact pattern being requested here, mirrored onto model responses. The implementation already exists for one side of the conversation. This request applies it to the other.