Assistant text block silently not rendered when followed by tool calls after a mid-turn user message
Environment
- Claude Code 2.1.170 (CLI, Linux, bash)
- Model: claude-fable-5
- Interactive session, default permission prompts on Edit
What happened
During a long interactive session, an assistant text block was never displayed in the terminal, even though the assistant produced it and acted on its content. The turn shape that triggered it:
- Assistant turn in progress: short text preamble, then a Bash tool call (a lookup).
- While that turn was still running, the user sent a new message (delivered to the model as "The user sent a new message while you were working").
- The assistant continued the same turn: it emitted a multi-line text block (an answer to the user's question, including a small table and a closing "applying now" sentence), followed immediately by 3 Edit tool calls.
- The terminal displayed the Edit permission prompts, but the text block from step 3 was never rendered.
From the user's perspective, their question was followed by three unexplained edit prompts — they repeatedly concluded the assistant was ignoring their questions and "just making changes." We verified message-by-message: the user scrolled back and confirmed the text block is absent from their terminal, while the assistant-side transcript contains it.
Expected
Any assistant text emitted in a turn is rendered, including text produced after a mid-turn user interjection and immediately before tool calls.
Actual
The text block sharing a turn with subsequent tool calls (after a mid-turn user message) was dropped from the display entirely. Tool permission prompts for the same turn rendered normally.
Notes
- Turn-final text messages (no trailing tool calls) rendered fine throughout the same session.
- We did not capture whether the block reappears after a /resume or in the session transcript file; happy to provide session details privately if useful.
This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗