Complex turns: important text output gets buried between tool calls

Problem

During complex multi-step sessions (3+ tool calls in a single turn), Claude's text output gets scattered between tool calls and system reminders. Users who scroll to the bottom of the terminal only see the last chunk of output and miss critical recommendations, analysis, or summaries that appeared earlier in the turn.

Example

Claude outputs a recommendation → runs 5 file edits → outputs another block of text. The user only sees the final text block and misses the recommendation entirely.

This is especially common in sessions involving:

• Multi-file refactors
• Skills/workflows that trigger many sequential edits
• Analysis followed by implementation in the same turn

Suggested Solutions (any would help)

1. "Text only" view toggle — Show just Claude's text output for a turn, hiding tool call details
2. Collapsible tool output sections — Let users collapse edit/read/bash blocks to see the text flow
3. Turn summary panel — A sidebar or footer that shows a condensed view of Claude's text output for the current turn
4. Searchable session transcript — Ability to search through all output in the current session (Ctrl+F equivalent)
5. Auto-summary at turn end — Built-in behavior where Claude appends a summary after complex turns (currently achievable via user instructions, but a native feature would be more reliable)

Current Workarounds

• Adding a user instruction to CLAUDE.md asking Claude to recap at end of complex turns (works ~80% of the time)
• Using /export after sessions to review what was missed
• Splitting analysis and execution into separate prompts

Impact

This is a daily friction point for power users running complex workflows. The more capable Claude Code becomes (more tool calls, longer turns), the worse this problem gets.

---

🤖 Generated with Claude Code