Allow intentional empty/whitespace-only assistant responses without the "no visible output" nudge

Resolved 💬 0 comments Opened Jul 12, 2026 by yuriver Closed Jul 12, 2026

Summary

When a user asks the assistant to stay silent (produce no reply), emitting an empty or whitespace-only response causes the harness to inject a nudge on the next turn:

[Your previous response had no visible output. Please continue and produce a user-visible response.]

This makes an intentional "no output" turn effectively impossible from the user's perspective — every attempt is followed by a re-prompt pushing the model to speak again.

Request

Provide a way to allow an intentional empty/whitespace-only response without the re-prompt, e.g.:

  • a setting/flag to suppress the "no visible output" nudge, or
  • treating a deliberate silent turn as valid completion.

Environment

  • Claude Code CLI
  • Observed consistently across consecutive turns when the assistant emits only whitespace.

View original on GitHub ↗