Protected Paths

Software project roots

A directory containing any of these markers is treated as a software project root and protected outright — organize (recursive or not) never descends into it or reclassifies its contents:

.git
Cargo.toml
package.json
pyproject.toml
pubspec.yaml

sift folders reflects the same rule: a child folder that’s a software project is reported under Protected, never moved or even suggested.

Traversal boundaries

Recursive traversal (organize --recursive, recursive Watch) also stops at known build/dependency output directories, without needing a project marker at that exact level:

node_modules
target
.venv

Sift’s own category directories

Recursive traversal also avoids descending into the categories Sift itself creates (Documents, Images, Audio, Video, Archives, 3D, Code, Data, Other), which prevents repeated nesting like:

Documents/Documents/file.pdf

A custom rule destination (e.g. destination = "Invoices") is protected from re-entry the same way — Sift never reclassifies a file it (or a nested .sift.toml) already placed there.

Hidden entries and symlinks

Hidden entries (dotfiles/dotdirs) are left alone entirely. Symlinks are never followed for organization decisions — see Safety Model for the full no-follow guarantee.

Checking what’s protected without applying anything

sift doctor ~/Downloads --recursive

reports software projects, symlinks, and other findings without touching the filesystem. See Doctor.