[FEATURE] Suggested-reply/autocomplete in Claude Desktop can leak answers in active-recall learning workflows

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

The Claude Desktop app shows an auto-suggested next message below the input box during a conversation. In active-recall learning workflows — where Claude presents a question, word, or prompt and the user is meant to produce the answer themselves before seeing it (e.g. flashcards, vocabulary drills, Q&A-style learning tools) — this suggestion sometimes contains the actual answer, effectively spoiling it before the user has a chance to recall it on their own.

Proposed Solution

The suggested-reply feature should either avoid suggesting content when the conversation context indicates the user is meant to produce a specific answer themselves (active-recall/learning format), or there should be a setting to disable auto-suggested replies entirely.

Alternative Solutions

_No response_

Priority

Medium - Would be very helpful

Feature Category

Interactive mode (TUI)

Use Case Example

• Build/run any learning-style conversation in Claude Desktop where Claude asks a question or shows a prompt, and the user is supposed to type the answer themselves (e.g. a custom vocabulary quiz, a Q&A drill, flashcard-style practice).
• Observe the auto-suggested reply chip/autocomplete below the input box.
• The suggestion occasionally contains the expected answer, inferred from conversation context (e.g. an example sentence containing the target word/concept).

Additional Context

Defeats the purpose of active-recall learning workflows, since the answer is visible before the user attempts to recall it themselves. This isn't specific to one app built on top of Claude — any conversation where the user is meant to self-test is affected.