System snapshot

• Uptime: 7 days, 9 hours
• Main filesystem: /dev/nvme0n1p3
• Capacity: 930G total, 770G used, 157G free
• Utilization: 84%

Biggest top-level items in /var/home/sasha

296G
/var/home/sasha/para

228G
/var/home/sasha/Downloads

37G
/var/home/sasha/PROJECTOR

2.5G
/var/home/sasha/actions-runner

1.7G
/var/home/sasha/models

Downloads findings

• Downloads is a major storage consumer and should be reviewed separately if cleanup is desired.

Hidden directories in ~

38G
~/.cache

22G
~/.cache/rattler

6.6G
~/.cache/uv

4.2G
~/.cache/Homebrew

1.9G
~/.cache/ms-playwright

36G
~/.local

36G
~/.local/share

18G
~/.julia

5.6G
compiled

5.3G
artifacts

2.6G
juliaup

16G
~/.var

16G
~/.var/app

4.8G
~/.npm

4.0G
~/.npm/_cacache

~/para breakdown

211G
~/para/areas

211G
~/para/areas/dev

85G
~/para/projects

85G
~/para/projects/work

~/para/areas/dev breakdown

211G
~/para/areas/dev/gh

Largest subtrees:

176G
~/para/areas/dev/gh/charly

30G
~/para/areas/dev/gh/phorma

3.2G
~/para/areas/dev/gh/sk

2.1G
~/para/areas/dev/gh/ak

gh/charly breakdown

75G
dont

57G
fotos

38G
wai

3.4G
pretender

1.3G
espectacular

gh/phorma breakdown

29G
PAPRIVA

758M
pap-fixes

Why dont is large

63G
~para/areas/dev/gh/charly/dont.claude

13G
~/para/areas/dev/gh/charly/dont/target

60M
~para/areas/dev/gh/charly/dont.beads

7.6M
~para/areas/dev/gh/charly/dont.git

dont/.claude breakdown

63G
~para/areas/dev/gh/charly/dont.claude/worktrees

4K
settings.local.json

The large files under .claude/worktrees are repeated Rust build outputs, for example:

~168M
target/debug/dont and target/debug/deps/dont-*

~175M
target/debug/deps/libcozo-*.rlib

Per-agent worktree sizes:

6.1G
agent-a8b7be7b

5.9G
agent-aa96cafe

5.8G
agent-a9ec9965

5.7G
agent-a252e688

5.7G
agent-a497a44a

5.7G
agent-a617d702

5.7G
agent-ae276a1d

5.7G
agent-ae2a2032

5.7G
agent-aedbbd7a

5.4G
agent-aab5bca7

5.0G
agent-a27af15c

3.6M
agent-aeb12e5f

Conclusion: dont is large primarily because Claude agent worktrees each contain duplicated Rust target directories.

Why fotos is large

32G
~/para/areas/dev/gh/charly/fotos/target

25G
~para/areas/dev/gh/charly/fotos.flatpak-builder

295M
~para/areas/dev/gh/charly/fotos.flatpak-repo

41M
~para/areas/dev/gh/charly/fotos.flatpak-build

Conclusion: fotos is large primarily because of Rust/Tauri build outputs and Flatpak build cache/artifacts.

Work project breakdown

66G
~para/projects/work.local

66G
~/.local/share

7.0G
~/.cache

3.2G
uv

2.1G
ms-playwright

1.5G
Homebrew

4.1G
navari

3.6G
navari-thinking

3.6G
navari-ak-mcp-sop

Inside the code directories, size is mostly from .venv, common, frontend, mcp, and .git.

.claude/worktrees across ~/para/areas/dev

Found 4 worktrees directories under ~/para/areas/dev.

By repo:

63G
~para/areas/dev/gh/charly/dont.claude/worktrees

2.6M
~para/areas/dev/gh/charly/nayra.claude/worktrees

0
~para/areas/dev/gh/charly/wai.claude/worktrees

0
~para/areas/dev/gh/sk/elegua.claude/worktrees

Total across all discovered worktrees directories:

• 61.97 GiB

Worktree counts by repo:

12
~/para/areas/dev/gh/charly/dont

1
~/para/areas/dev/gh/charly/nayra

0
~/para/areas/dev/gh/charly/wai

0
~/para/areas/dev/gh/sk/elegua

Main cleanup candidates

Safest obvious candidates:

• Large stale files in ~/Downloads
• ~/.cache/rattler
• ~/.cache/uv
• ~/.cache/Homebrew
• ~/.npm/_cacache
• ~/para/areas/dev/gh/charly/fotos/target
• ~para/areas/dev/gh/charly/fotos.flatpak-builder
• ~/para/areas/dev/gh/charly/dont/target

Potentially high-impact candidate requiring intent check:

• ~para/areas/dev/gh/charly/dont.claude/worktrees
  • Large because of duplicated per-agent build outputs
  • Should confirm whether any active Claude agent sessions still depend on these worktrees before deletion