claude-code - 💡(How to fix) Fix [FEATURE] Visual conventions for blocking questions and end-of-session summaries [1 participants]

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

I'm a non-developer running ~5 personal projects in parallel through Claude Code, switching between sessions multiple times an hour. Two specific kinds of content are hard to spot when scanning a long response:

1. Blocking questions — when Claude needs my input to continue coding, the question is often a single sentence buried in a paragraph. I scroll past it and Claude ends up waiting for an answer I never noticed it asked for. I only realize after coming back from another session, which costs me real time.

2. End-of-task summaries — after Claude finishes a code change, the recap of what was actually done is mixed in with verification logs, file paths, and explanations. I can't always re-read everything Claude produced when juggling multiple conversations, so I miss state changes I should retain.

The issue isn't verbosity — Claude is already good at being concise when asked. It's that the visual format gives equal weight to every line, whether it's a status update, a code reference, a question I need to answer, or a summary I should remember. Nothing visually separates "this is information" from "this is something you must act on or remember".

I've worked around this by asking Claude (via memory files) to color-code blocking questions in bold yellow ANSI and end-of-session summaries in green ANSI. It works, but it's fragile — it depends on Claude remembering to apply the convention every turn, and I had to invent it myself.

Proposed Solution

Native, opt-in visual conventions for content types that demand user attention or retention. At minimum:

• A standardized way to flag "I need an answer to continue" so the user can spot it instantly when scrolling — could be a colored prefix, an icon in the gutter, a side panel, or a dedicated render block. The form doesn't matter as long as it's visually distinct from regular prose.

• A standardized end-of-turn recap slot when code was modified, rendered differently from explanation/log output. Could appear as a styled "Summary" block at the end of the turn, or as a persistent panel that updates after each code change.

The exact UX is your call. The underlying need: let users scan a response and instantly find the question they need to answer and the summary they need to retain.

This would also reduce the load on user-side memory tricks (which are currently fragile) and make Claude Code easier for non-developers like me, who switch contexts often and rely on the chat itself as their working memory.

Alternative Solutions

My current workaround: I've added memory files that ask Claude to color-code blocking questions in bold yellow ANSI and end-of-session code summaries in green ANSI. It works most of the time, but it's fragile because it depends on Claude remembering to apply the convention every turn, and on me having thought to set it up — most users wouldn't.

I've also tried being more disciplined about reading every response in full before switching sessions. That defeats the purpose of having an agent: I end up spending the time I was supposed to save just reviewing output.

I haven't found a satisfying alternative inside Claude Code itself — the issue is that the chat is a single stream of equally-weighted text, so any workaround has to live inside that stream.

Priority

Medium - Would be very helpful

Feature Category

Other

Use Case Example

A typical day:

1. Morning — I'm working on Project A. Claude starts a code change and, halfway through, asks me a clarifying question buried in a paragraph of explanation.
2. I switch to Project B in another window without noticing the question. I work on B for an hour.
3. Mid-afternoon — I come back to Project A. I see ~600 lines of output from Claude. I need to figure out two things, fast:
   • Is Claude waiting on me for an answer?
   • What did Claude actually code while I was away (so I can decide what to ask next)?
4. Today I have to scroll back and skim the whole conversation. Sometimes I miss the question entirely and only realize Claude has been idle for an hour when I notice nothing has happened. Sometimes I miss a file Claude modified and only catch it later via git status.

With the proposed visual conventions, opening the conversation would surface the blocking question and the latest code summary at a glance — same way unread badges or status pills work in chat apps. The information is already in the response; it just needs to be visually distinguishable from the prose around it.

Additional Context

Parallels in other tools that already do something similar:

• Browser DevTools console uses different colors and icons per log level (info / warn / error) so developers can scan a long log instantly.
• Slack and Discord visually distinguish mentions, threads, and unread messages from regular messages.
• Linear and GitHub themselves render code blocks, callouts, and quotes with distinct backgrounds so they stand out from prose.

Claude Code already does this for code blocks and tool calls — they're rendered differently from prose. The proposal is just to extend the same idea to two more content types that demand user attention: blocking questions and code-change summaries.

Note on context: I'm a non-developer who codes through Claude Code by description, not by writing code directly. People in my situation rely even more heavily on the chat itself as the source of truth (we don't always read the diff afterwards), so any visual differentiation in the response itself has an outsized impact on usability.