Task output temp file grew to 1.4 TB, filling disk — /private/tmp/claude-501/[session]/tasks/*.output
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?
A temporary task output file created by Claude Code under /private/tmp/claude-501/ grew to 1.4 TB, completely filling my disk.
File path pattern:
/private/tmp/claude-501/<session-id>/tasks/<taskname>.output
The specific file was bpnhmxvxm.output, dated Mar 14, 2026 at 2:16 AM, and reported as 1.4 TB in macOS Finder. Other .output files in the same directory are normal sizes (2 KB, 3 KB, 633 KB, 31 KB). This one file appears to have grown unbounded — likely a runaway write loop or stdout/stderr not being capped during parallel task execution.
Steps to reproduce:
- Run Claude Code with an agentic/multi-step task (e.g., using subagents or parallel tasks)
- Monitor
/private/tmp/claude-501/disk usage - Observe one
.outputfile growing unbounded until disk is full
Environment:
- OS: macOS (path: Macintosh HD > private > tmp > claude-501)
- Session temp dir:
/private/tmp/claude-501/ - File in question:
.../cf50a1f0-0feb-4de8-a8a9-a6eb50aae0ca/tasks/bpnhmxvxm.output
What Should Happen?
Temp task output files should be bounded in size. Claude Code should not fill the user's disk. Suggested fixes:
- Cap
.outputfile size or truncate/rotate after a threshold - Add disk space checks before writing task output
- Clean up
/private/tmp/claude-501/on session end
Error Messages/Logs
Steps to Reproduce
This happens as Claude plans and creates taks.
Claude Model
Opus
Is this a regression?
I don't know
Last Working Version
_No response_
Claude Code Version
2.1.77 (Claude Code)
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Other
Additional Information
<img width="1048" height="152" alt="Image" src="https://github.com/user-attachments/assets/0b2e2e84-7415-48dd-95d7-4cd0a3a7a97d" />
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗