codex - 💡(How to fix) Fix Codex Desktop regression: slash/local skill chip submission hangs, errors, and duplicates messages on Windows after GPT-5.5 update [1 comments, 2 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
openai/codex#19249Fetched 2026-04-24 05:58:02
View on GitHub
Comments
1
Participants
2
Timeline
7
Reactions
0
Author
Timeline (top)
labeled ×5commented ×1unlabeled ×1

After the recent Codex Desktop update that introduced/updated GPT-5.5, messages containing slash commands or local skill chips became unreliable in Codex Desktop on Windows.

Before the update, slash/local skill commands like openspec-review submitted and started seamlessly. Now they often show Error submitting message, sit on Thinking for a long time, and sometimes duplicate the same user message bubble before eventually starting execution.

This appears to happen before the agent actually begins work, so it looks like a desktop app submission/context-packaging/retry regression rather than a model or repo-task issue.

Error Message

Before the update, slash/local skill commands like openspec-review submitted and started seamlessly. Now they often show Error submitting message, sit on Thinking for a long time, and sometimes duplicate the same user message bubble before eventually starting execution.

  • Error submitting message The UI sometimes reports Error submitting message, delays for a long time, duplicates the same submitted bubble, and then may eventually execute.
  • The failure text is Error submitting message, before useful agent work starts.
  • Error submitting message

Root Cause

After the recent Codex Desktop update that introduced/updated GPT-5.5, messages containing slash commands or local skill chips became unreliable in Codex Desktop on Windows.

Before the update, slash/local skill commands like openspec-review submitted and started seamlessly. Now they often show Error submitting message, sit on Thinking for a long time, and sometimes duplicate the same user message bubble before eventually starting execution.

This appears to happen before the agent actually begins work, so it looks like a desktop app submission/context-packaging/retry regression rather than a model or repo-task issue.

RAW_BUFFERClick to expand / collapse

What version of the Codex App are you using (From “About Codex” dialog)?

26.422.20832

What subscription do you have?

Plus

What platform is your computer?

Microsoft Windows NT 10.0.26200.0 x64

What issue are you seeing?

Summary

After the recent Codex Desktop update that introduced/updated GPT-5.5, messages containing slash commands or local skill chips became unreliable in Codex Desktop on Windows.

Before the update, slash/local skill commands like openspec-review submitted and started seamlessly. Now they often show Error submitting message, sit on Thinking for a long time, and sometimes duplicate the same user message bubble before eventually starting execution.

This appears to happen before the agent actually begins work, so it looks like a desktop app submission/context-packaging/retry regression rather than a model or repo-task issue.

Environment

  • App: Codex Desktop
  • OS: Windows
  • Shell: PowerShell
  • Workspace: large local repo with repo-local skills/plugins
  • Model/update context: started after recent Codex Desktop upgrade that came with GPT-5.5
  • Time zone: Africa/Lagos
  • Approx observed date: April 24, 2026

What steps can reproduce the bug?

Repro Steps

  1. Open Codex Desktop in a large local coding workspace.
  2. Submit a message containing a local skill/slash command chip, for example: commit + openspec-review
  3. Observe one or more of:
    • Error submitting message
    • long delay on Thinking
    • same submitted message appears multiple times
    • eventual execution starts after a delay
  4. Plain text messages appear less affected; slash/local skill chips are the consistent trigger.

What is the expected behavior?

Expected Behavior

A message containing a slash/local skill chip should submit once, resolve the skill, and begin model execution promptly, as it did before the recent desktop update.

Actual Behavior

The UI sometimes reports Error submitting message, delays for a long time, duplicates the same submitted bubble, and then may eventually execute.

Why This Seems App-Side

  • The failure text is Error submitting message, before useful agent work starts.
  • The task eventually runs correctly once submission succeeds.
  • The issue is correlated with slash/local skill chip submission.
  • It began immediately after a Codex Desktop upgrade.
  • The same workspace and workflow were previously seamless.

Additional information

Relevant Local Skill

Example skill used:

openspec-review

Path:

C:\Users\Didi\Documents\Projects\VigilAfrica\vigilafrica\.agents\skills\openspec-review\SKILL.md

Skill workflow reads:

  • .agents/workflows/openspec-review.md
  • openspec/proposals/chore-frontend-tests.md
  • openspec/specs/chore-frontend-tests.md
  • Task.md
  • docs/standards/developers-react.md

The repo has many available skills/plugins, so this may be related to large skill/tool context packaging.

Attachments

  • Screenshot showing:
    • Error submitting message
    • repeated commit + openspec-review bubbles
    • delayed Thinking / Working for 11s

Possible Regression Areas

  • Local skill chip resolution
  • Slash command submission path
  • Large context/tool/skill manifest packaging
  • Retry/deduplication after failed submission
  • Windows Codex Desktop app message queue
<img width="1030" height="527" alt="Image" src="https://github.com/user-attachments/assets/ae528885-57ed-4b3d-82d6-c5e6f1fa063b" />

extent analysis

TL;DR

The issue with unreliable message submissions containing slash commands or local skill chips in Codex Desktop on Windows may be resolved by investigating and adjusting the local skill chip resolution, slash command submission path, or large context packaging.

Guidance

  • Verify that the issue is specific to messages with slash commands or local skill chips by testing plain text messages and comparing the behavior.
  • Check the skill workflow and manifest files (e.g., openspec-review\SKILL.md, .agents/workflows/openspec-review.md) for any recent changes or potential issues that could be causing the submission errors.
  • Investigate the Windows Codex Desktop app message queue and retry/deduplication mechanisms to see if there are any configuration issues or bottlenecks that could be contributing to the delays and duplicate submissions.
  • Consider testing the same workflow and skills in a smaller repository or a different environment to isolate the issue and determine if it's related to the large skill/tool context packaging.

Example

No specific code snippet is provided, but reviewing the skill workflow files (e.g., openspec-review\SKILL.md) and the Windows Codex Desktop app configuration may help identify potential issues.

Notes

The issue appears to be specific to the Windows platform and the recent Codex Desktop update that introduced GPT-5.5. The large local repository and numerous skills/plugins may be contributing to the issue, but further investigation is needed to determine the root cause.

Recommendation

Apply a workaround by testing the skill workflows and submissions in a smaller repository or a different environment to isolate the issue and determine the best course of action. This will help identify if the issue is related to the large skill/tool context packaging or another factor.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING