disk-cleanup

Disk Cleanup

Reclaim disk space on a Linux VM. The skill runs a triage — survey every space consumer, split targets into safe vs ask, execute the approved set, then verify the delta.

When to Use

• User says "disk cleanup", "free up space", "clean up the VM"
• Disk usage above 80%
• After large builds, git operations, or Docker image accumulation

Procedure

1. Baseline

df -h

Done: used and available numbers recorded for all mounts for the before/after comparison.

2. Survey

Requires root for apt, journalctl, and /var/log operations. Non-root users can survey but cannot clean system paths.

Build a ranked list of every consumer over ~50M. Survey all mounts, not just /.

# All mounts (not just /)
df -h
df -i                           # inode exhaustion — "No space left on device" with free blocks

# Deleted-but-open files (canonical "df says full, du says empty" case)
lsof +L1 2>/dev/null | head -20

# Hermes data + user caches + tmp (stay on one filesystem, don't cross mounts)
du -xsh ~/.hermes/*/ 2>/dev/null | sort -rh | head -20
du -xsh ~/.hermes/projects/*/ 2>/dev/null | sort -rh | head -20
du -xsh ~/.cache/*/ 2>/dev/null | sort -rh | head -15
du -xsh /tmp/* 2>/dev/null | sort -rh | head -15

# Memory store (if Mnemosyne is installed)
du -xsh ~/.hermes/mnemosyne/data/* 2>/dev/null | sort -rh

# Session checkpoints (if present)
du -xsh ~/.hermes/checkpoints/store/ 2>/dev/null

# Docker
docker system df                # shows build cache separately from images
docker images --format "{{.Size}}\t{{.Repository}}:{{.Tag}}" | sort -rh | head -25
docker ps -a --filter "status=exited" --format "{{.Image}}" | sort -u

# System
du -xsh /var/cache/apt /var/log ~/.cache/pip ~/.cache/uv 2>/dev/null
journalctl --disk-usage
find /var/log \( -name "*.gz" -o -name "*.1" -o -name "*.old" \) 2>/dev/null

# Bloated git repos (orphaned tmp_pack files from interrupted operations)
find ~/.hermes/projects -name ".git" -type d -exec sh -c 'du -sh "$1" 2>/dev/null' _ {} \; | sort -rh | head -10