[BUG] SANDBOX ERROR CODE ENOSPC: no space left on device,

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?

Here's a public-safe version formatted for a GitHub issue. Drop it straight into the bug report.

Title: Cowork sandbox returns ENOSPC on every operation, 100% reproducible, persists across new sessions and projects for 10+ days
Body:
Summary
The Cowork sandbox environment returns an ENOSPC: no space left on device error on every operation, including trivial commands. The error is 100% reproducible and has persisted for over 10 days across multiple new sessions and multiple new projects. All code execution, file creation, and document-generation skills (docx, pptx, xlsx, pdf) are blocked.
Error
Every sandbox operation - including a minimal echo or date - returns:
ENOSPC: no space left on device, mkdir '/sessions/[REDACTED]/tmp/[REDACTED]'
The session path is identical on every retry, including in freshly created sessions, suggesting bad state on the account-level sandbox allocation rather than a transient per-session issue.
Steps to reproduce

Open Cowork in the Claude desktop app
Start a new session in any project
Ask Claude to run any command, create any file, or invoke any document skill
Sandbox returns the ENOSPC error immediately
Repeat across new sessions and new projects - identical result every time

Expected behaviour
Sandbox should provision a working temp directory and execute basic commands.
Actual behaviour
Sandbox fails to create its temp directory on initialisation, returning ENOSPC before any user command can run.
Environment

Product: Cowork (Claude desktop app)
Plan tier: Max 20x
Usage limits: confirmed not the cause - account usage across all categories is effectively zero
Duration: 10+ days continuous

Troubleshooting already attempted

Started multiple new Cowork sessions across the 10+ day window - identical failure every time
Created multiple new projects - identical failure every time
Submitted in-app feedback on multiple occasions
Opened a support ticket - was told the issue was fixed; re-tested immediately afterwards and the error is unchanged with the same session path, suggesting the fix did not apply to the affected account

Impact

All document and code-generation work via Cowork is blocked
Document skills (docx, pptx, xlsx, pdf) cannot run
File outputs to the workspace folder cannot be produced
Live compliance work has had to be moved off Cowork to continue

Likely cause (best guess from the outside)
The identical session path returned on every retry, including in fresh sessions, points to stuck state in the account's sandbox allocation rather than a globally exhausted resource. A general infrastructure fix may not clear it - the affected account likely needs a manual sandbox reset or storage reallocation.
Additional notes

Error reproduces on demand, every single time, so triage from the error string and account allocation should be straightforward
Happy to provide further logs, screenshots, or session detail through a private channel - this issue is being tracked in a support ticket in parallel

What Should Happen?

fix my issue

Error Messages/Logs

Steps to Reproduce

ENOSPC: no space left on device, mkdir '/sessions/vibrant-stoic-euler/tmp/claude-1036'

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

latest version

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_