[BUG] Skills upload returns "Internal server error" on Claude Web and Desktop
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?
Uploading a custom skill via Settings > Capabilities > Skills returns "Internal server error" immediately on both Claude Web (browser) and Claude Desktop app. The error appears before any processing occurs.
What I've Tried
- Hard refresh (Ctrl+Shift+R)
- Different browsers
- Drag and drop vs. file picker
- Both .zip and .md formats
- Claude Desktop app and browser (claude.ai)
Environment
OS: Windows
Browser: Microsoft Edge Version 145.0.3800.58
Additional Context
Skill file structure is valid — folder containing SKILL.md with correct YAML frontmatter (name + description). Skill was generated by Claude itself.
What Should Happen?
Skill uploads successfully and appears in the Skills list.
Actual Behavior
"Internal server error" is shown immediately. Upload fails every time.
Error Messages/Logs
Steps to Reproduce
- Go to Settings > Capabilities > Skills
- Click "Upload skill"
- Select a valid .zip file containing a skill folder with SKILL.md
- Observe error
Claude Model
Sonnet (default)
Is this a regression?
Yes, this worked in a previous version
Last Working Version
_No response_
Claude Code Version
Version 1.1.3189 (1b7b58)
Platform
Anthropic API
Operating System
Windows
Terminal/Shell
Windows Terminal
Additional Information
<img width="2224" height="1520" alt="Image" src="https://github.com/user-attachments/assets/8e3b95c7-04f3-4ea3-a2dc-fa00bed6403b" />
29 Comments
Found 3 possible duplicate issues:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
Same issue. It's part of an overall problem with skills.
All in all, seems like something broken in the Claude Skills db or something.
Adding to this after experiencing the same issue for the last few days (was this part of the "claude can no longer write to the skill directory" update in the last ~3-5 days?) -> I can have someone in my organisation upload the exact same skill file, and it works for them. Any way of interacting with skills (add, edit, delete, replace) no longer works on my account.
Via desktop, any browser, Mac app, iPad, iPhone, incognito mode -> all apps return internal server error on attempt. I can create a skill via the 'write skill instructions' option, but neither creating via Claude nor uploading manually works.
On top of this ->
I highly suspect that, because I cannot replicate this with any other account in my organisation, this is an account-specific error.
Same issue on my end. There was a brief period where the skills were not visible in the settings (MacOS and Web), but Claude was still able to call them and see them.
Then another issue where I couldn't delete a skill. And now I get the same "Internal Error" when trying to upload one I created with Claude.
!Image
Same here when uploading a new skill. I am using windows 11 pro
Can confirm this is also affecting existing uploaded skills, not just new uploads.
I have 46 custom skills on my account. 26 load fine and their files are accessible during conversations. The other 20 appear to be registered (names and descriptions show up in the system), but the actual skill files never mount to the filesystem — so Claude can see they exist but can't read any of the instructions or reference files.
This seems related to the same infrastructure issue causing upload failures. The skills are "in the system" but not being served to the runtime environment.
Pro plan, claude.ai browser on Mac.
Same issue here, could not update a skill or add a new one.
I can now add skills but cannot delete or replace old ones. So progress.
Same. Lots of issues all of a sudden with Skills on Mac.
<img width="468" height="77" alt="Image" src="https://github.com/user-attachments/assets/adf5f7d7-4945-42de-b1db-8d9f56dbb1ab" />
<img width="543" height="418" alt="Image" src="https://github.com/user-attachments/assets/e1bd2578-f7db-41c6-932b-0db6a95539dd" />
Same issue. Have a bugged skill that I cannot delete or replace. I can also not see it's contents and get the warning "try re-uploading the skill" when I try.
It now seems to be intermittent to me -> i temporarily could add a skill, then the next one failed. I could delete one, then the next one failed. Still not 100%
It's definitely not intermittent for me -- I uploaded a skill, it uploaded but then didn't work when I tried using it, and now every skill I upload after that is broken in the same way -- none of them work and none of them can be deleted ("internal server error".)
Still have the same problems as above, can't deleate, replace or add new skills. Just get the same error message "internal server error"
Same issue for me as well. Can't delete, replace, or add new skill...tried chrome, edge, desktop app (both on Mac and Windows), same across all sessions.
(generated by claude.ai)
Additional reproduction data — persistent issue since late January 2026
I've been hitting this consistently since ~Jan 26, 2026 across 40+ custom skills in a Claude Pro project. Extensive debugging over multiple sessions. Here's what I've found that may help triage:
Environment
Failure modes observed
| Operation | Result |
|-----------|--------|
| Upload new .skill (zip) | Internal server error |
| Upload new .skill via drag-and-drop | Internal server error |
| Delete existing skill | Internal server error |
| Overwrite existing skill (re-upload same name) | Internal server error |
Both save and delete are broken, suggesting the issue is in the shared skills CRUD backend, not upload parsing.
What I've ruled out (content is NOT the cause)
Through systematic binary-search debugging (splitting skills in half, uploading minimal stubs), I confirmed:
grep -rP '[^\x00-\x7F]'returns 0 lines (no Unicode/emoji)\"sequences (previously caused save errors)None of these workarounds resolve the issue. The same skill files that previously uploaded successfully (before late January) now fail on re-upload.
Timeline
Workaround (partial)
Skills that were uploaded before the breakage still function correctly within project sessions. Claude can read and execute them. The issue is exclusively with the CRUD operations in the Settings > Capabilities > Skills UI.
Suggestion for devs
Since delete fails with the same error as upload, the bug is likely in a shared API handler or auth/validation layer for skill operations, not in file parsing or content validation. The transition from intermittent to 100% failure suggests a backend change shipped between late January and early February 2026.
This is getting a bit frustrating that the issue is dragging out for so long. Have anyone heard claude make any statement about the issue?
There's an official incident on their status page, which seems to be on way to be resolved => https://status.claude.com/incidents/5pr1d63fdjml
However the issue is ongoing since Feb 17...
Thanks for the response! That's a good sign at least!
Last I checked, I still had some skills which are bugged, BUT new skills I upload are NOT bugged anymore. This means I can just reupload the affected skills, and the problem is effectively solved. Maybe worth a try for others here. (I believe this will only work for new conversations; I think existing conversations can't properly pick up new skill uploads.)
On my side nope, still encountering problems like "skill not found":
<img width="395" height="68" alt="Image" src="https://github.com/user-attachments/assets/2e4c1cb3-2f6c-44ef-90ee-7e7db86005d5" />
Can't delete them either
What do you mean "only work for new converstaions"? How are you uploading the skill when it works? Do you drop it into a fresh conversation or upload it in the capabilites tab in settings?
My experience, even before this issue, was that -- when I upload a new skill into the conversations tab in settings -- Claude can immediately see the skill description in any conversation, but the actual skill files are missing if it tries to read them in conversations that were already started before the skill was uploaded. But new conversations started after the upload will have the skill.
Aah, that’s not my experience. I can’t even upload it. I just get internal server error
FYSA: DELETE is working for me now here in case helpful - got this from Anthropic who responded to a different/old request i sent a while back. i haven't tried uploading or saving new skills on this link yet.
https://claude.ai/customize/skills
Issue still persist. anyone found a workaround?
Still having this, any fixes?
Experiencing this issue this morning
Having this exact issue right now as well. Tried loading the skill through the skill creator, tried uploading a .zip file, a .skill file, and the .md file as well. All 4 result in internal server error.
We're hitting this same issue, and it's blocking us at the organization level. We have a skill we'd like to roll out across the team, but with the upload broken, there's no way to make it available to everyone through the official channel.
The usual fallbacks don't really work for an org-wide deployment:
Is there any workaround in the meantime? For example, a way to provision the skill server-side via support, an API endpoint that bypasses the UI, or any temporary path for organization admins? An ETA on the fix would also help us plan.
Thanks.