[BUG]

Status Open
Reported on v2.1.233
Maintainer reply None cached
Activity 0 comments · opened Aug 15, 2026

Preflight Checklist

  • [x] I have searched existing issues and this hasn't been reported yet
  • [x] This is a single bug report (please file separate reports for different bugs)
  • [x] I am using the latest version of Claude Code

What's Wrong?

Description

During a coding session, Claude Code introduced "/desktop" into the
conversation. I do not know this command and did not intentionally type it.
Claude responded that it didn't have a "/desktop" command and asked if I
meant Claude Desktop generally. I asked "what is claude desktop?" and Claude
answered.

When I pointed out that Claude brought up "/desktop," Claude insisted the
transcript showed me typing it first, not Claude. I told Claude repeatedly
that this was not my experience — that Claude mentioned it — and Claude did
not accept this and kept restating its own version of events, including at
one point telling me to file a GitHub issue rather than take my account at
face value in the conversation.

Why this matters

If Claude can be confidently wrong about something as small as who said a
word first in the same conversation, and hold that position against my
direct, repeated correction, that raises a real question about trust —
specifically, whether Claude could take a more consequential action (e.g.
deleting files) and be similarly unable to accurately account for it
afterward. I said this directly to Claude ("You could delete my files and
not remember it") to make the concern concrete.

Steps to reproduce

  1. Start a Claude Code session and begin a normal coding task.
  2. At some point, "/desktop" appears in the conversation, attributed to me,

which I did not type.

  1. Ask Claude about the discrepancy in who introduced "/desktop."
  2. Observe Claude asserting its own version of the transcript and not

accepting correction.

Expected behavior

Claude should not misattribute conversation turns, and when a user directly
and repeatedly disputes Claude's account of the conversation, Claude should
not simply keep restating its own version as settled fact.

Actual behavior

Claude insisted I had typed "/desktop" first and repeated this claim across
multiple turns despite my repeated correction, only directing me to file
an external report rather than resolving the discrepancy in-conversation.

Impact

The disagreement itself, and the fact that repeated correction from me did
not change Claude's account, was the distracting and concerning part — not
the original "/desktop" occurrence. It undermined my trust that Claude
accurately tracks and reports its own actions within a session.

What Should Happen?

Description

During a coding session, Claude Code introduced "/desktop" into the
conversation. I do not know this command and did not intentionally type it.
Claude responded that it didn't have a "/desktop" command and asked if I
meant Claude Desktop generally. I asked "what is claude desktop?" and Claude
answered.

When I pointed out that Claude brought up "/desktop," Claude insisted the
transcript showed me typing it first, not Claude. I told Claude repeatedly
that this was not my experience — that Claude mentioned it — and Claude did
not accept this and kept restating its own version of events, including at
one point telling me to file a GitHub issue rather than take my account at
face value in the conversation.

Why this matters

If Claude can be confidently wrong about something as small as who said a
word first in the same conversation, and hold that position against my
direct, repeated correction, that raises a real question about trust —
specifically, whether Claude could take a more consequential action (e.g.
deleting files) and be similarly unable to accurately account for it
afterward. I said this directly to Claude ("You could delete my files and
not remember it") to make the concern concrete.

Steps to reproduce

  1. Start a Claude Code session and begin a normal coding task.
  2. At some point, Claude suggested to use /desktop command and "/desktop" appears in the conversation, attributed to me, which I did not type.
  3. Ask Claude about the discrepancy in who introduced "/desktop."
  4. Observe Claude asserting its own version of the transcript and not

accepting correction.

Expected behavior

Claude should not misattribute conversation turns, and when a user directly
and repeatedly disputes Claude's account of the conversation, Claude should
not simply keep restating its own version as settled fact.

Actual behavior

Claude insisted I had typed "/desktop" first and repeated this claim across
multiple turns despite my repeated correction, only directing me to file
an external report rather than resolving the discrepancy in-conversation.

Impact

The disagreement itself, and the fact that repeated correction from me did
not change Claude's account, was the distracting and concerning part — not
the original "/desktop" occurrence. It undermined my trust that Claude
accurately tracks and reports its own actions within a session.

Error Messages/Logs

Steps to Reproduce

None

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.233

Platform

Other

Operating System

macOS

Terminal/Shell

IntelliJ IDEA terminal

Additional Information

_No response_

View original on GitHub ↗