github.com web signal

Claude Code 2.1.198 auto-skips AskUserQuestion after 60s

TL;DR

  • In Claude Code 2.1.198, AskUserQuestion auto-returned a 'No response after 60s' message and told Claude to proceed on its own judgment.
  • The behavior was undocumented, missing from the changelog, and a regression from 2.1.196, which worked correctly per the report.
  • Maintainer ThariqS said a release will expose the setting under /config with the timeout configurable and defaulting to off.

The whole point of AskUserQuestion is that Claude stops and waits for you. Issue #73125 in anthropics/claude-code reports that in version 2.1.198, the tool started auto-returning a synthetic message reading 'No response after 60s — the user may be away from keyboard. Proceed using your best judgment based on the context so far; you can re-ask this question later if it's still relevant' whenever a user had not answered within a minute.

That quietly reframes AskUserQuestion from a hard human-in-the-loop stop into a soft nudge. The concerns laid out in the issue are that no timeout parameter appears in the tool's schema, the change was not documented in the changelog, and version 2.1.196 worked correctly per the report. So it landed as an undocumented behavior change on what people were treating as a safety primitive, on the Bedrock, Linux, VS Code, Opus stack the reporter was running.

Maintainer ThariqS replied in the thread that a release cut 'tomorrow morning PT' would expose the setting, writing that it 'will appear in /config and the timeout will be configurable, but will default to not being on'. That is roughly the right shape of fix for anyone using AskUserQuestion to gate destructive actions, since the safety contract is restored by default and the escape hatch is opt-in.

The honest caveat is that the reporting doesn't tell you whether the 60 second auto-continue also fired on stacks other than the one in the bug report, or how many 'proceed with best judgment' runs happened before the fix landed. If you ran 2.1.198 against anything irreversible recently, it's worth scanning your transcripts for that synthetic response. The upside, once the toggle ships, is that people running truly autonomous loops that can't hang forever get an opt-in, and everyone else keeps the stop-and-wait behavior they were relying on.

Shared on Bluesky by 2 AI experts