disk-cleanup
Skill6 starsChanged 45 days agoMITView on GitHub
Audit disk usage on this PC and clean it up safely. Shows terminal box charts of what is fat (home dirs, Steam games, caches, rebuildable build output), then a verdict panel and a pick-what-to-delete question. Use when the user asks to tidy files and dirs, free disk space, find what is safe to delete, prune caches, or "what is eating my disk".
22 other projects use a version of this.
Adopted by
Most recently taken up by
- kdy1/kdy1-scriptstheir own version, first seen Nov 15, 2026
- addisonk/skillstheir own version, first seen Nov 8, 2026
- DaraBoth/skilljmskilltheir own version, first seen Nov 8, 2026
- jars-demo/jars-skillstheir own version, first seen Nov 1, 2026
- huihuisang/ajigu-skillstheir own version, first seen Nov 1, 2026
- pabloleonalcaide/development-skillstheir own version, first seen Nov 1, 2026
Used with
Projects with it also use api-design, tdd and 1 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
MicrosoftSkillapi-design
Guide for designing and documenting RESTful APIs. Use when asked to design an API, create endpoints, or document an API.
Also in 3 projects that use this
GoogleSkilltdd
Test-driven development. Use when the user wants to build features or fix bugs test-first, mentions "red-green-refactor", or wants integration tests.
Also in 4 projects that use this
MicrosoftSkillrelease
Use ONLY when the user explicitly asks to cut/publish a Huabu desktop release or tag a version. Performs an outward action — bumps the version in apps/desktop/package.json, commits + annotated-tags vX.Y.Z, and pushes the branch + tag. That tag push auto-builds the macOS/Windows installers and publishes a GitHub Release. Provides `status`, `bump`, `tag-push`, and a one-command `run` (bump → tag-push).
Also in 3 projects that use this
Other versions22 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 22
- glebis/claude-skills★ 414same name
- mtarcure/claude-vibe-squad★ 220same name
- kdy1/kdy1-scripts★ 34same name
- DaraBoth/skilljmskill★ 8same name
- MakiDevelop/claude-skills★ 3same name
- vndee/engineering-skills★ 3same name
- TsinHzl/claude-code-config★ 2same name
- addisonk/skills★ 1same name
- CharlesLadera/claude-custom-skills★ 1same name
- foogunlana/skills★ 1same name
12 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository12 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- kaizen-conceptSkill
- kaizen-exploreSkill
- kaizen-helpSkill
- kaizenSkill
- kaizen-bookSkill
- kaizen-deckSkill
- kaizen-gameSkill
- kaizen-videoSkill
- kaizen-computerSkill
- lemo-opuscarSkill
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.
0
Common
Most of the 15 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
Verify before saying "safe": Look inside (ls, du, diff -rq, read project.json) before calling anything cache or a copy.
- Distinctive: line 15
Delete only on a pick. Prefer move-to-Archive for anything not obviously regenerable.
- Very specific: line 54
Every box line must be 40 cells wide by w() (emoji = 2).
“Common” is measured: kept by at least 6 in 10 of the 15 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Disk cleanup
Audit first, delete only what the user picks. Never delete without an explicit pick.
Rules learned the hard way
- Verify before saying "safe". Look inside (
ls,du,diff -rq, readproject.json) before calling anything cache or a copy. A dir named.cachecan hold a venv, a backup, or project history. - Never total unlike things. A sum of cache dirs is not "reclaimable". Report each group on its own and state what is regenerable and what is not.
- Use real numbers. Say what a number is (apparent vs on-disk). The disk is btrfs with zstd, so use
du -xanddf -h /home. - Delete only on a pick. Prefer move-to-Archive for anything not obviously regenerable.
- Report outcomes faithfully: freed space comes from
dfbefore/after, not from the plan.
Procedure
- Overview.
df -h /home, thendu -xsh ~/.[!.]* ~/* | sort -rh | head -25. Expect~/.local/share/Steamto dominate; drill into it separately. - Drill the fat ones with
du -xh --max-depth=1 <dir> | sort -rh | head:- Steam:
steamapps/common,shadercache,compatdata. Game name, size and last played come fromsteamapps/appmanifest_*.acf(name,SizeOnDisk,LastPlayed; missing or 0 = never played). - Projects: rebuildable output =
node_modules,target,.next,dist(find Projects -name node_modules -type d -prune). ~/.cache,~/.configbrowser profiles,~/.hermes,~/.local/share/mise/installs(old tool versions).
- Steam:
- Round 1: show charts, then ask where to go. Print the box charts as chat text (see Format), then
AskUserQuestionwith: explore outside Steam / explore Steam / continue to cleanup. Do the exploration the user picks, show new boxes, ask again. - Round 2: verdict, then ask what to run. Print the verdict panel as chat text, then
AskUserQuestion(multiSelect, max 4 options; overflow items go via "Other") whose options are the verdict rows. - Run only the picks. Close Steam before touching
steamapps. Re-rundf -h /homeand report freed space.
Verdict tiers
| Tier | Meaning | Examples |
| ✓ SAFE NOW | regenerates, no work lost | Steam shadercache, node_modules/target/.next, .zcompdump, empty dirs, stale .claude.json.tmp.* |
| ✓ RE-DOWNLOAD | recoverable, costs time | never-played Steam games, model caches under ~/.cache/huggingface |
| ? YOUR CALL | depends on use | idle games (2+ months), old mise versions, VMs, headless browser profiles |
| ✗ KEEP | never offer | Projects source, Documents, .ssh, .gnupg, .kaizen, Games/roms |
Watch for: whisperx-venv is a venv, not cache; kaizen-video needs it. Desktop shortcuts in ~ for a "never played" game mean the user may be about to play it.
Format
Boxes are 38 wide, two per row. No scale text in the title: the title is bold, the top-right shows the box's used total and a 5-cell gauge. Sizes show as TB, GB, MB or KB (never bare G); a blank line separates the title from the rows. Row bars fill the width up to the right-aligned size column and scale to the box's biggest row (▌ half, ▏ under one half-cell). Mark never-played games ✗.
Gauge = box total ÷ the limit (the disk size for HOME and STEAM boxes; the biggest dir's parent for others). Each filled cell is an emoji square coloured by that ratio: 🟩 <50%, 🟨 <75%, 🟧 <90%, 🟥 ≥90%; empty cells are ⬜. Both are 2 cells wide, so box counts them double. Bold titles use Unicode bold letters (𝗛𝗢𝗠𝗘) because markdown bold does not work inside a code block. Never hand-draw boxes: import the shared component tui.py (same dir as this file) for every report that uses a box.
import sys, os; sys.path.insert(0, os.path.expanduser("~/.claude/skills/disk-cleanup"))
from tui import box, panel, render, storage, GB, MB, KB
render(box("HOME", [("Steam", GB(1500)), ("Projects", GB(52))], GB(1552), GB(1900)), box("CACHE", [("huggingface", GB(12)), ("tmp", MB(300), "✗")], GB(12), GB(1900)))box(title, [(name, bytes, flag?)], used, limit) makes a bold-title box: sizes in bytes (GB(), MB(), KB() build them; shown as GB, MB, KB), bars fill the row up to the size column and scale to the box's biggest row, the top-right gauge is box total vs limit; panel(title, [(line, note)], foot) the verdict panel; render prints two per row. storage([(label, bytes)], total, free) returns the full-width DISK box: one bar split by pattern (█ ▓ ▒, then ░ other, blank free; chat text cannot colour per section), real disk size top-right, and a one-line legend underneath (short lowercase labels, sizes without a space, so it fits one line). Print it above every chart set, using disjoint sections (e.g. steam games, steam other, home rest) and the real df total and free. A box total must be the real du figure, not the sum of the rows shown.
Every box line must be 40 cells wide by w() (emoji = 2); check before printing.
Verdict panel: panel(), one numbered row per option ([1] ✓ SAFE NOW <what> <size> with a └ note line), a ✗ KEEP footer.
The user sees chat text and the question dialog, but a question dialog shows only its own text, so charts go in chat before the call, never inside the question.
Bash output is collapsed in the user's terminal ("Ran 2 shell commands"), so a chart printed by render is invisible. After running tui.py, copy its output verbatim into your own reply inside a ``` code block, before the AskUserQuestion call. Never ask "next step?" until the chart is in reply text. Same for the verdict panel.
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
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.

