harn disk

harn disk [path] [options]

disk measures regular files under a directory, including hidden folders, ignored files, nested repositories, and initialized submodules. The default path is the current directory. It also accepts one regular file.

The report shows selected totals, a Git-state breakdown, the largest groups, and the largest files. It reads filesystem and Git index metadata. It does not read file contents, hash files, or delete anything.

Git filters

harn disk --git tracked
harn disk --git ignored --group directory --depth 2
harn disk --git untracked,ignored --group repository
harn disk --git git-metadata

tracked means present in the repository’s current index. untracked means untracked and not ignored; ignored covers files excluded by Git’s ignore rules. git-metadata covers .git directories and pointer files. Files outside a working repository are non-repository. Failed Git classification is unknown, makes the report incomplete, and sets a nonzero exit status.

Each nested working repository uses its own index and ignore rules. A submodule’s Git link does not classify the files inside it. Starting from a subdirectory still uses its enclosing repository’s rules. Missing tracked files and uninitialized submodules consume no measured bytes.

The sizes describe files currently on disk. --git tracked does not estimate compressed Git history or clone download size. Git metadata outside the scan directory is not visited.

Other filters and grouping

harn disk --exclude node_modules,.next,.git
harn disk projects --type video --min-size 100MB --group files --top 10
harn disk --type .png,.jpg --group extension
harn disk --group repository --json

Option	Behavior
--git <states>	Select comma-separated Git states. Omit to include all.
--type <types>	Select extensions or video, audio, image, media, archive.
--exclude <paths>	Skip names at any depth or paths relative to the scan root. Repeatable and comma-separated. No glob syntax.
--group <by>	directory (default), repository, extension, git, or files.
--depth <n>	Directory grouping depth, default 1. The scan still visits deeper files.
--top <n>	Maximum displayed groups and largest files, default 20. Totals cover all selected files.
--min-size <size>	Minimum individual file size using the selected measurement. Accepts bytes, decimal units such as MB, and binary units such as MiB.
--apparent	Rank and filter by file lengths instead of allocated disk blocks.
--json	Emit harnery.disk-usage/v1 with byte counts, filters, Git states, groups, scan issues, and largest files.

Measurement limits

Allocated bytes count each hard-linked inode once across selected files. When links span groups, allocation belongs to the first visited path; file lengths still count every path. Largest-file rows show each file’s own allocation, so adding those rows can count a hard link more than once.

Sparse files can have large file lengths while allocating little disk space. The report always includes apparent bytes for comparison. Windows lacks a portable allocated-block count, so the default falls back to file lengths and labels the measurement apparent-fallback; JSON allocated bytes are null.

The scanner skips symlinks, special files, and directory metadata. It never follows symlinks. Files can change during a scan; read failures appear in the report and cause a nonzero exit. A live scan is not an atomic snapshot.

Use storage inventory for registered agent storage and its retention policies. disk inventories the selected filesystem tree without applying storage policies.