SafeDisk AI

Claude Code Says ENOSPC Or 0MB Free, But Your Mac Has Space

If Claude Code loses command output under a task capture path while df -h still shows plenty of free space, do not start by deleting Claude state. First prove whether this is real disk pressure, inode exhaustion, or a capture-layer failure.

No mail app or GitHub login? Send this directly from any inbox.

liuminsheng3@gmail.com - SafeDisk Claude Code ENOSPC Review Payment Link
Claude Code incident review

Turn one Claude Code ENOSPC incident into a safe cleanup order.

Use the read-only checks to separate real disk pressure from a capture-layer failure. The $29 review covers one public-safe summary or scan and a safe/review/do-not-touch cleanup order.

task_dir="/tmp/claude-503"; df -h "$task_dir" "$HOME"; df -i "$task_dir" "$HOME"; du -sh "$task_dir" 2>/dev/null
Email $29 payment link Request $29 private payment link Request $99 team policy Send incident details Read-only evidence AI CLI runbook Open backup request
Before deleting Claude state

Request the $29 Claude ENOSPC incident review.

Email is enough to request the payment link. Add the exact Claude warning only if you have it handy; after payment we reply with the next read-only check and cleanup boundary before you delete any Claude state.

Use private checkout

Do not paste secrets, transcripts, private logs, or full project paths.

Evidence Before Cleanup

The key split is whether normal writes still work outside Claude's task output capture path. If $HOME writes succeed while the task directory reports 0MB free, broad cache deletion may not address the real failure mode.

task_dir="/tmp/claude-503"
df -h "$task_dir" "$HOME" 2>/dev/null
df -i "$task_dir" "$HOME" 2>/dev/null
du -sh "$task_dir" 2>/dev/null
find "$task_dir" -maxdepth 3 -type f -name "*.output" -print 2>/dev/null | head -40
printf "capture-test\n" > "$HOME/claude-output-workaround.txt"
ls -lh "$HOME/claude-output-workaround.txt"

Do Not Delete These First

Copy-ready GitHub reply

Use this when Claude reports 0MB free but macOS has space.

This keeps the thread focused on proof instead of risky reset advice.

This looks like a task-output capture ENOSPC, not necessarily real disk exhaustion. Before deleting Claude state, I would collect the exact task-dir evidence and prove writes outside the harness still work:

task_dir="/tmp/claude-503"
df -h "$task_dir" "$HOME" 2>/dev/null
df -i "$task_dir" "$HOME" 2>/dev/null
du -sh "$task_dir" 2>/dev/null
find "$task_dir" -maxdepth 3 -type f -name "*.output" -print 2>/dev/null | head -40
printf "capture-test\n" > "$HOME/claude-output-workaround.txt"
ls -lh "$HOME/claude-output-workaround.txt"

If $HOME writes succeed while Claude says the task capture path has 0MB free, keep redirecting important command output outside the task dir and avoid broad deletion of Claude app support, ~/.claude, VM bundles, or active session state.

Paid Review Scope

The $29 review is intentionally narrow: one Claude Code ENOSPC incident, one public-safe summary or read-only scan, and one safe/review/do-not-touch cleanup order. If the same incident affects a team, the $99 pilot turns it into a reusable policy.