Issue Station
Sign In
Tools hereToolbox

claude-code - 💡(How to fix) Fix Feedback rating prompt collides with assistant's numbered questions in the same input context [2 comments, 3 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
anthropics/claude-code#50741•Fetched 2026-04-20 12:14:20
View on GitHub
Comments
2
Participants
3
Timeline
7
Reactions
0
Author
AiRaxAiRax
Participants
0xbrainkid0xbrainkidAiRaxAiRaxgithub-actions[bot]github-actions[bot]
Timeline (top)
labeled ×4commented ×2unlabeled ×1

Root Cause

Because both flows consume single-digit input in the same place, a user intending to answer question #1 from the assistant can instead submit a "1 = Bad" rating. The feedback gets recorded incorrectly, and the intended answer never reaches the assistant.

RAW_BUFFERClick to expand / collapse

Preflight Checklist

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

Problem Statement

Description:

Claude Code occasionally shows a feedback prompt like "1-Bad, 2-OK, 3-Good" (or similar rating scale) in the main chat input. Meanwhile, the
assistant frequently asks numbered questions in its replies — e.g., "1. Do X? 2. Should we Y? 3. Do Z?" — and expects short numeric answers like "1" or "1, 3".

Because both flows consume single-digit input in the same place, a user intending to answer question #1 from the assistant can instead submit a "1 = Bad" rating. The feedback gets recorded incorrectly, and the intended answer never reaches the assistant.

Reproduction:

  1. Have the assistant ask 2–3 numbered questions in a single message.
  2. Shortly after, a feedback rating prompt appears in the input.
  3. Type "1" (or any single digit) intending to answer question #1.
  4. Claude Code captures it as a feedback rating.

Impact:

  • Low-signal telemetry (users rate "Bad" without meaning to).
  • Lost intent in the conversation (assistant never sees the intended answer).
  • Frustration — happened to me at least twice so far.

Suggestions (any one would help):

  • Show the feedback prompt in a modal / separate UI surface, not in the same input that accepts chat.
  • Require an explicit verb or prefix for feedback (e.g., /rate 1).
  • Defer feedback until a pause in active conversation (idle timeout).
  • Make it obvious in the input when a digit will go to feedback vs to the chat (distinct styling / border color / placeholder text).

Proposed Solution

Make feedback rating options clickable buttons instead of accepting numeric input via the chat input field.

  • Render the rating scale (e.g., "Bad / OK / Good" or 1–5) as a row of clickable buttons in a dedicated UI surface (inline panel, modal, or a
    fixed feedback bar), separate from the chat input.
  • Clicking a button submits the rating directly; the chat input stays exclusively for conversation.
  • This eliminates any ambiguity between "digit I type = feedback rating" and "digit I type = answer to assistant's numbered question". The two
    input surfaces are physically separate, and accidental miss-rating becomes structurally impossible.

Optionally, a small "Skip" or "Dismiss" action on the same panel for users who don't want to rate — so closing it doesn't require typing
anything into the chat input.

Alternative Solutions

No response

Priority

Critical - Blocking my work

Feature Category

CLI commands and flags

Use Case Example

No response

Additional Context

No response

extent analysis

TL;DR

Implementing clickable buttons for feedback ratings can resolve the issue of accidental rating submissions.

Guidance

  • Consider rendering the rating scale as a row of clickable buttons in a dedicated UI surface, such as an inline panel or modal, to separate it from the chat input.
  • Ensure that clicking a button submits the rating directly without affecting the chat input, eliminating ambiguity between feedback ratings and answers to numbered questions.
  • Adding a "Skip" or "Dismiss" action to the rating panel can provide users with an easy way to opt out of rating without interacting with the chat input.
  • Verify that the new implementation does not introduce any new issues with the conversation flow or feedback collection.

Example

<!-- Example of a simple rating panel with clickable buttons -->
<div class="rating-panel">
  <button class="rating-button" value="1">Bad</button>
  <button class="rating-button" value="2">OK</button>
  <button class="rating-button" value="3">Good</button>
  <button class="skip-button">Skip</button>
</div>

Notes

The proposed solution focuses on changing the UI to prevent accidental submissions. However, it's essential to test the new implementation thoroughly to ensure it works as expected and does not introduce new issues.

Recommendation

Apply the workaround of using clickable buttons for feedback ratings, as it directly addresses the issue of accidental rating submissions and provides a clear separation between feedback and chat inputs.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#memory optimization#batch processing#GPU compatibility#latency issue#model loading

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING

claude-code - 💡(How to fix) Fix Feedback rating prompt collides with assistant's numbered questions in the same input context [2 comments, 3 participants]