# Space Hog

> Find what fills a drive. Paste a storage listing - du -ah or du -k, ls -l or ls -lR, find -printf "%s %p", or a CSV
> or tab-separated export with a name and a size column (a Google Drive storage list) - and the page totals the space by
> folder and by kind of file, lists the biggest files and the likely duplicates, and draws a squarified treemap. A
> metered run plans each candidate - archive it off the drive, remove the extra copies of a duplicate, keep it, or ask
> first - and the page adds up what the plan frees against the amount you want back. Nothing in the plan deletes a file's
> only copy. Built on the SkillSafe skill @googleworkspace/recipe-find-large-files.

https://space-hog.skillsafe.ai/

## What is free
- Reading listings: du (size then path; folder totals set aside when their files are listed; bare numbers as 1 KB blocks), ls -l and ls -lR (the size column under each folder heading), find -printf (bytes then path), and tables with a name (or path) and a size column, optionally a folder column. Sizes with units count in powers of 1024.
- Windows: dir /s output (Directory of headings, byte sizes with separators) and a PowerShell Get-ChildItem export with FullName and Length columns. The page shows a copyable command for Mac, Linux and Windows. Up to 20,000 lines are read; a longer listing is cut and the cut is announced.
- Before any extra copy is removed, a copyable checksum command (shasum on Mac/Linux, Get-FileHash on Windows) for each group of likely copies, since the same size and name is only a guess.
- Totals by folder (rolled up) and by kind of file (video, raw photo, image, audio, archive, disk image, pdf, document, spreadsheet, presentation, app or code, other).
- Likely duplicates: the same size and the same name once Copy of, (1) and copy are set aside.
- The treemap (PNG, SVG) and a CSV of every file with its size, kind and plan.
- Three samples replay their saved plans at no cost.

## The metered task
One task, `space`, on gpt-terra, metered per run. Input string fields: `task`, `goal`, `target`, `totals`, `folders` and `items` (one line per candidate, `i1.` ...: a group of copies with its extra size, or a single file with its size and kind). The reply is one JSON object: `verdict` (`reaches` or `short`), `plan` (each `item`, `action` - `remove_copies`, `archive`, `ask` or `keep` - `why`, `question`), `frees` and `note`.

The page refuses an action that does not fit its item (removing copies of a single file, archiving a group of copies), adds up what the plan frees from its own byte counts, checks the verdict against the target, and warns when the reply talks about deleting a file.

## Links
- API tutorial: https://space-hog.skillsafe.ai/api.html
- Source skill: https://skillsafe.ai/skill/@googleworkspace/recipe-find-large-files