Little Snitch Control for Alfred

Switch Little Snitch 6 operation modes and its network filter from Alfred — without UI scripting, without a passwordless sudo rule, and without a background helper. Every change is authorized by macOS and read back from Little Snitch before the workflow claims it worked.

Why

Little Snitch's own controls live in a menu bar item and a settings window. This puts the four states you actually switch between — Alert, Silent Allow, Silent Deny, and filter on/off — one keystroke away, and shows you what the firewall reported the last time it was asked.

The two states that weaken protection, plus turning the filter off, are only reachable with ⌘↩. A stray Return cannot disable your firewall.

Warning

Disabling the Network Filter makes every Little Snitch rule ineffective and allows all connections until you turn it back on. Nothing re-enables it automatically.

Requirements

• macOS with Little Snitch 6.2 or newer installed in /Applications
• Alfred 5 with the Powerpack
• An administrator password each time the workflow reads or changes state
• Little Snitch's Allow access via Terminal setting turned on

That last one is the real cost of using this workflow, and it is worth understanding before you install. Objective Development ships it off specifically to stop scripts from manipulating your firewall settings. Turning it on lifts that restriction for every process running as you, not just this workflow — anything that can already run code as your user and obtain administrator rights can then drive littlesnitch too. This workflow cannot work without it, because littlesnitch refuses to run at all otherwise, even to read a value.

If that trade is not one you want to make, use Little Snitch's own menu bar item instead. See SECURITY.md for the full threat model.

Verified against Little Snitch 6.4.1 on macOS 26 (Apple silicon). Newer 6.x releases are accepted and labelled "untested" in the status row rather than refused. Little Snitch 5 and earlier are refused.

Install

1. Download the latest .alfredworkflow from the releases page.

2. Verify the checksum published with that release:

   shasum -a 256 -c Little-Snitch-Control-*.alfredworkflow.sha256

   Run it in the download folder with both files present. tests/package.zsh fails the release if this command ever names a stale version.

3. Open the file to install it in Alfred.

Or build it yourself with make build; the archive is byte-reproducible, so a local build of a release tag produces the same checksum as the published asset.

Setup

1. In Little Snitch, open Settings → Security.
2. Unlock the settings and enable Allow access via Terminal.
3. Type snitch in Alfred and choose Refresh Status.
4. Approve the macOS administrator dialog.

Important

Step 2 is the largest security cost this workflow asks of you. Objective Development ships that switch off deliberately — in their words, "in order to prevent malicious scripts from manipulating your firewall settings". Turning it on lets any process that can obtain root reconfigure Little Snitch, not only this workflow. Decide that you want that before installing.

The keyword is configurable in the workflow's configuration in Alfred.

Usage

Type snitch (default keyword), then Tab or Space to enter the workflow, and choose an action. Without that keystroke Alfred is still in its normal search, so it may rank its own results — including Little Snitch itself — above the workflow's rows.

The list shows the last verified state:

Row	What it does
Filter: … · Mode: …	Last verified state. ↩ refreshes it.
Refresh Status	Reads and verifies the current state.
Use Alert Mode	Ask when no existing rule matches.
⌘↩ Use Silent Allow	Allow unmatched connections without asking.
⌘↩ Use Silent Deny	Deny unmatched connections without asking.
Enable Network Filter	Apply Little Snitch rules to network traffic.
⌘↩ Disable Network Filter	Stop applying every Little Snitch rule.
Open Little Snitch	Open Network Monitor and manage rules.
Open Setup Help	Open this page.

Rows prefixed ⌘↩ ignore a plain Return by design.

State is labelled Last verified rather than "current", because Little Snitch can also be changed from its own interface, by an active profile, or by another administrator. The workflow never polls a privileged firewall setting in the background.

Uninstall

1. In Alfred → Workflows, right-click Little Snitch Control → Delete.

2. Remove the cached state:

   rm -rf ~/Library/Caches/com.runningwithcrayons.Alfred/Workflow\ Data/com.hashkode.alfred.little-snitch-control

3. Optionally turn Allow access via Terminal back off in Little Snitch.

Nothing else is installed. There is no daemon, login item, privileged helper, or sudoers entry to remove — you can confirm with:

ls /Library/LaunchDaemons /Library/LaunchAgents ~/Library/LaunchAgents /Library/PrivilegedHelperTools
sudo ls /etc/sudoers.d

How it works

Little Snitch ships an official command-line utility. The workflow sets these preferences:

Action	Preference and value
Alert Mode	activeSilentMode 0
Silent Allow	activeSilentMode 1
Silent Deny	activeSilentMode 2
Network Filter On	networkFilterEnabled true
Network Filter Off	networkFilterEnabled false

Note

read-preference and write-preference are documented commands, but these two preference keys are not a documented public API. Objective Development's documentation shows activeSilentMode only in passing, with the value 0; the values 1 and 2, and networkFilterEnabled entirely, were established by observation. They can change in any Little Snitch update without notice. This is why every action reads the state back and reports what Little Snitch actually says rather than what was requested. The mapping above was confirmed against Little Snitch 6.4.1 — see docs/VERIFIED-MODES.md — and is re-checked before each release with scripts/verify-modes.zsh.

Every privileged action follows the same sequence:

1. Alfred passes one opaque identifier from a closed list of six.
2. A fixed AppleScript maps that identifier to one predefined Little Snitch command. No text from Alfred reaches the shell.
3. macOS requests administrator authorization.
4. Both preferences are read back and accepted only if they are literally 0, 1, 2 and true, false.
5. Only a confirmed result is written to a user-owned cache and displayed.

Opening the result list never requests administrator access.

Security model

What this design does defend against:

• Argument and environment injection from Alfred input: a closed six-action allowlist, fixed argument vectors built with quoted form of rather than string interpolation, and a sanitized privileged environment.
• Elevating the wrong executable: the path is fixed, each component is checked for root ownership and non-root writability, and the binary must satisfy codesign --verify --strict against Objective Development's Team ID.
• A forged or stale cache: it is refused unless it is a regular file owned by you with mode 0600, and every field is validated. Nothing privileged ever reads it.
• Concurrent runs: actions are serialized, and results are read back before success is reported.

What it does not defend against — and cannot:

• Code already running as your user. An Alfred workflow bundle is user-writable by design, so anything running as you can rewrite this workflow and obtain root at the next authorization prompt you approve. The macOS dialog is titled osascript and does not show what will run, so you cannot detect this by reading the prompt.
• Little Snitch state changed after the last refresh, by anything at all.

No password is stored or piped anywhere, and no passwordless sudo rule or persistent root helper is installed. See SECURITY.md for the full threat model.

Troubleshooting

"Little Snitch rejected the request"

Confirm Little Snitch → Settings → Security → Allow access via Terminal is enabled. This consent is deliberately not automated.

The mode changed back immediately

An active Little Snitch profile can select its own operation mode. Change the profile or its mode setting, then refresh.

Signature verification failed

Reinstall Little Snitch from Objective Development. The workflow will not elevate an executable that fails verification.

The status looks stale

Choose Refresh Status. The workflow never polls in the background.

Development

No third-party runtime dependencies.

make test          # pure: no Little Snitch, no license, no password required
make lint          # MegaLinter (needs Docker) -- a subset of CI, see below
make build         # writes dist/
make package-test  # validates the built archive

Linting covers Markdown, YAML, Python, GitHub Actions, spelling and .editorconfig compliance. It does not cover zsh: shellcheck and shfmt reject zsh rather than degrading, so every zsh file is syntax-checked by make test instead.

make lint is a strict subset of what CI runs. Secret scanning (gitleaks) and the pull-request title check happen only in CI, and on Apple silicon make lint disables actionlint's shellcheck integration. A green make lint is necessary, not sufficient.

make test runs entirely against mocks and writes nothing outside a temporary directory. Privileged behaviour cannot be automated; follow the manual test checklist on a non-critical Mac before tagging a release, and re-run scripts/verify-modes.zsh.

workflow/            Alfred bundle source (info.plist, icons, bin/)
scripts/build.zsh    reproducible release packaging
scripts/make-icons.py  generates the workflow artwork from source
scripts/verify-modes.zsh  read-only check of the operation-mode mapping
tests/run.zsh        unit and behaviour tests
tests/package.zsh    release-archive validation
docs/                developer documentation (not shipped in the workflow)

Roadmap

No fixed schedule. Under consideration, in rough order of likelihood:

• Profile switching, if it can be done without letting free-form text cross the privileged boundary.
• Rule group and blocklist toggling.

A timed disable with automatic re-enabling, and a signed helper for promptless operation, have both been considered and set aside: each requires background persistence or a paid signing pipeline that this design deliberately avoids. See SECURITY.md.

Profiles are deliberately not exposed through free-form Alfred input. Privileged arguments remain a closed list.

License and trademarks

MIT.

This project is unofficial and is not affiliated with, endorsed by, or sponsored by Objective Development Software GmbH or Running with Crayons Ltd. Little Snitch is a product of Objective Development Software GmbH. "Alfred" is a registered trademark of Running with Crayons Ltd. The workflow's artwork is original; Little Snitch's own icon is only ever read from your installed copy at runtime and is never redistributed here.

References

• Little Snitch command-line overview
• Little Snitch operation modes
• Little Snitch security settings
• Alfred Script Filter JSON