agentleFS
htlin222htlin222/dotfiles/.claude/skills/housekeeping/SKILL.md

housekeeping

Skill78 starsChanged 4 months agoMITView on GitHub

Use when macOS disk space is low, "disk full", 硬碟空間不夠, startup disk almost full, or before installing something large — diagnoses what is eating space and reclaims it safely

Files
SKILL.md

macOS Disk Housekeeping

Overview

Diagnose → reclaim safe targets → verify. Known-safe cleanup only; never touch personal data. reference/prune.sh automates the whole flow (dry-run by default).

Diagnosis

df -h /System/Volumes/Data        # the REAL usage (df / shows the sealed system snapshot)
du -sh ~/Library/* | sort -rh | head        # usual suspects
du -sh ~/* ~/.[a-z]* 2>/dev/null | sort -rh | head
tmutil listlocalsnapshots /       # MSUPrepareUpdate = staged macOS update

Known hogs on this machine (2026-07 findings)

TargetTypical sizeAction
~/Library/Application Support/Claude/vm_bundles~10Gdelete if VM not running (pgrep -fl claudevm); rebuilds on demand
~/Library/pnpm store7G+pnpm store prune
~/Library/Caches/{com.spotify.client,BraveSoftware,Google,com.brave.Browser,pip,go-build,node-gyp}~4Gdelete outright
~/.cache/uv2G+uv cache prune (remainder is referenced by live venvs — leave it)
/opt/homebrew20G+brew cleanup --prune=all && brew autoremove
MSUPrepareUpdate snapshotfew GBfinish the pending macOS update (reboot) — softwareupdate -l to see it

Do NOT delete

  • ~/Library/Caches/ms-playwright* — breaks E2E until playwright install
  • ~/.cache/codex-runtimes — Codex plugin runtime
  • ~/Pictures, ~/Library/Group Containers, Mail — personal data
  • leftover ~/.cache/uv after prune — hardlinked into venvs

Gotchas

  • rip does not free space. It moves files to /tmp/graveyard-$USER on the same volume. After ripping large targets run yes | rip --decompose and re-check df.
  • brew cleanup frees nothing when packages are outdated — it only deletes versions older than the installed-latest. Upgrade first if you want real savings.
  • uv cache prune vs clean: prune keeps in-use entries (correct); clean wipes all (pointless — hardlinked files free nothing).

Automation

reference/prune.sh          # dry-run report of known-safe targets
reference/prune.sh --apply  # actually clean them

# Find NEW candidates: large + old + cold, ranked by "storage rent"
# (score = size_GB × idle_days; gates on created>180d AND atime>90d;
#  top hits cross-checked with Spotlight kMDItemLastUsedDate)
reference/clean-large-and-olds.sh                       # scan $HOME, top 20
reference/clean-large-and-olds.sh --root DIR --min-size 50 --top 30

clean-large-and-olds.sh is report-only — review, then rip <paths> + yes | rip --decompose.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.