OPINIONATED OMARCHY
TROUBLESHOOTING CORPUS
790
RECORDS
790
AUDITED
0
UNCHECKED
2388
SOURCES
LIVE
security low occasional DISPUTE

omarchy-debug bundles unredacted LAN MAC and IP addresses, and the bug template asks you to attach it

omarchy
CORRECTED An auditor found the fix wrong or incomplete and rewrote it. · checked against Omarchy 4.0.4-1 on 2026-09-19
AUDIT NOTE

Gate 1 holds: issue #10255 (wbnns, 2026-09-04) and PR #10257 (open, unmerged as of 2026-09-18) are by a third party and predate anything here. Gate 2 held on the cause but not on the fix: the record is scoped to MAC addresses and [UFW BLOCK] lines while its own sources go further. The issue names the hostname on the Hostname: line and the PR review measured it 41 times in one run through the journal prefix, the PR review found a real IPv6 address in a systemd-resolved warning outside the firewall lines, and a reviewer found the lowercase-only MAC pattern the record uses misses the uppercase form BlueZ prints, so the record's verify reports a clean file that is not. The record's danger argues against masking dotted quads because of version strings, which the PR author retracted in the PR itself: it is true of the whole file and false of the two sections the leak lives in, so the fix is rewritten to scope every substitution to the DMESG and JOURNALCTL sections with a sed range, mask the hostname where the script writes it, keep loopback, link-local and unspecified addresses, and surface IPv6 candidates for review rather than pretend a paste-sized pattern handles them. The rewritten recipe was tested on a synthetic log in a scratch directory, never on this machine's output. Confirmed on this machine: /usr/bin/omarchy-debug (omarchy-settings 4.0.2-1) is byte-identical to the pinned blob and to current quattro, so the fix applies to the installed version, /etc/ufw/ufw.conf carries LOGLEVEL=low and man ufw says low is the default, and the [UFW BLOCK] LOG rules in /etc/ufw/user.rules set no --log-level, which is consistent with the warning level the issue reports but was not checked against kernel source. The script also leaves the unscrubbed file at /tmp/omarchy-debug.log with mode 0644, which the fix now removes. Gate 3 holds. Severity lowered from high to low: the brief's low level is information disclosure with no credential value, which is exactly this. Nothing on the machine is exposed to anyone by the default, no attacker acts against the machine, and the only path to a reader is the user attaching the file, which is the step the record intervenes in. The BSSID-to-location consequence is why the record exists, and it is still disclosure without credential value. Not exercised: omarchy-debug was not run here. Re-checked against 4.0.4-1 2026-09-19: Re-checked against 4.0.4-1 on 2026-09-19. /usr/bin/omarchy-debug (omarchy-settings 4.0.4-1) is byte-identical to the v4.0.4 tag (c668141e9c42b13c80c9ca4ea108e11708c5e8a5) and to the pinned blob at 9174fbf, and the quattro history for bin/omarchy-debug has no commit after 9174fbf, so every line the cause and fix rely on still holds: the fixed /tmp/omarchy-debug.log path, unfiltered sudo dmesg and journalctl -b -p 4..1, the Hostname: line, filtered inxi -Farz, and the section headers the sed ranges key on. The bug template at the v4.0.4 tag is byte-identical to the pinned 14f8038 blob and still asks for omarchy-debug output on line 22. On this machine /etc/ufw/ufw.conf still carries LOGLEVEL=low and the [UFW BLOCK] LOG rules in /etc/ufw/user.rules still set no --log-level. Issue #10255 and PR #10257 are both still open, and neither changed after 2026-09-08. None of the 145 files in the v4.0.2...v4.0.4 compare touches the script, the template or ufw. The only text change is the dated currency claim in fix, now 2026-09-19 with the version range named. PR #7995 (open, 2026-08-24, third party) independently describes the fixed /tmp path and its 0644 mode, so it is added as a source for that sentence. Other open PRs touch the script without redacting anything (#10548 swaps hostname for uname -n, #9477 and #8828 change collectors). If #10548 merges, the fix still works because uname -n and hostname return the same nodename. Not exercised: omarchy-debug was not run and nothing was uploaded.

The cause above was rewritten on 2026-09-18 to match this note; the fix was corrected by the audit itself.
RISK

Do not scrub the whole file with a broad address pattern. Outside the DMESG and JOURNALCTL sections every dotted-quad string is a version number, inxi 3.3.41.1 and every package in the list, and masking those corrupts the parts of the report that are actually diagnostic, so scope the substitutions to the two sections as above. Do not treat a zero count from a lowercase-only MAC pattern as clean: the kernel prints MAC addresses in lowercase and BlueZ prints them in uppercase, and a lowercase-only check reported the first version of PR #10257 clean while a paired Bluetooth device's address remained. The script writes the unscrubbed original to /tmp/omarchy-debug.log on every run, so remove it each time.

Symptom

You are about to attach the log from omarchy-debug to a bug report, because the issue template asks you to "attach the output of omarchy-debug if possible," and you want to know what is actually in it before pasting it somewhere public.

Cause

bin/omarchy-debug writes sudo dmesg and journalctl -b -p 4..1 into the log with no filtering. ufw's default is LOGLEVEL=low (/etc/ufw/ufw.conf, and man ufw says ufw defaults to a loglevel of 'low'), Omarchy's installer enables ufw, and the [UFW BLOCK] LOG rules set no log level, so every blocked packet is logged by the kernel at warning level as a [UFW BLOCK] line carrying the source and destination MAC and IP addresses of devices on the local network. The journal section reproduces the same lines because -p 4..1 includes warnings. On a laptop that has associated to Wi-Fi that boot, the kernel's own association messages in both sections carry the access point's BSSID, which public wardriving databases can resolve to a physical location. Outside the firewall lines the same sections carry the addresses of paired Bluetooth devices, which BlueZ prints in uppercase where the kernel prints lowercase, and any address a service put in a warning, such as the DNS server in a systemd-resolved message. The hostname is written twice over: once on the Hostname: line the script adds itself, and again as the prefix of every journal line, because the script uses journalctl's default output format. The inxi -Farz call in the same script is filtered, since -z masks IP addresses, serial numbers and MAC addresses in its own section, so the smallest section of the log is redacted and the two largest are not.

Fix

There is no flag that redacts the kernel and journal sections, and the upstream change that would do it, PR #10257, is open and unmerged as of 2026-09-19, and bin/omarchy-debug is unchanged from 4.0.2-1 through 4.0.4-1. --no-sudo skips dmesg entirely but leaves the journal untouched, and the bug template does not mention it. Before attaching a debug log anywhere public, generate it, remove the unscrubbed copy the script leaves at /tmp/omarchy-debug.log (mode 0644, readable by every local account), and scrub the two unfiltered sections. The script asks for your sudo password for dmesg.


omarchy-debug --print > ~/omarchy-debug-review.log
rm -f /tmp/omarchy-debug.log

Mask the hostname where the script writes it, on the Hostname: line and as the prefix of every journal line:


sed -i -E -e 's/^Hostname: .*/Hostname: <host>/' \
  -e "/^JOURNALCTL/,/^INSTALLED PACKAGES\$/ s/^([A-Z][a-z]{2} [ 0-9][0-9] [0-9:]{8}) $(hostname) /\1 <host> /" \
  ~/omarchy-debug-review.log

Then, only between the DMESG and INSTALLED PACKAGES headers so the inxi report and the package list are never touched, drop the firewall lines and mask MAC addresses in either case and IPv4 addresses. Loopback, link-local and the unspecified address are kept because they identify nobody and a failed DHCP lease shows as 169.254.x.x:


sed -i -E '/^DMESG$/,/^INSTALLED PACKAGES$/{
/\[UFW BLOCK\]/d
s/(^|[^0-9A-Za-z:])([0-9A-Fa-f]{2}:){5}[0-9A-Fa-f]{2}($|[^0-9A-Za-z:]|:($|[^0-9A-Fa-f]))/\1<mac>\3/g
s/(^|[^0-9A-Za-z:])([0-9A-Fa-f]{2}:){5}[0-9A-Fa-f]{2}($|[^0-9A-Za-z:]|:($|[^0-9A-Fa-f]))/\1<mac>\3/g
s/(^|[^0-9.])(127|0|169)\.([0-9]{1,3})\.([0-9]{1,3})\.([0-9]{1,3})/\1\2_\3_\4_\5/g
s/(^|[^0-9.])([0-9]{1,3}\.){3}[0-9]{1,3}($|[^0-9.])/\1<ip>\3/g
s/(^|[^0-9.])([0-9]{1,3}\.){3}[0-9]{1,3}($|[^0-9.])/\1<ip>\3/g
s/(^|[^0-9.])(127|0|169)_([0-9]{1,3})_([0-9]{1,3})_([0-9]{1,3})/\1\2.\3.\4.\5/g
}' ~/omarchy-debug-review.log

The MAC and IPv4 rules each run twice so two addresses one character apart are both caught. A four-number version string inside those two sections, such as a firmware version, can come out as <ip>, which costs nothing. IPv6 addresses are not masked automatically: a pattern short enough to paste here also matches timestamps and C++ names such as Foo::bar, so list the candidates and mask any real address by hand:


grep -nE '[0-9A-Fa-f]{1,4}(:[0-9A-Fa-f]{1,4}){3,}|[0-9A-Fa-f]{1,4}::|::[0-9A-Fa-f]' ~/omarchy-debug-review.log

Attach ~/omarchy-debug-review.log, not the original.

Verify

Confirm the firewall lines and the hostname are gone from the review copy, and review what remains rather than assuming it is clean:


grep -c '\[UFW BLOCK\]' ~/omarchy-debug-review.log
grep -Eic '([0-9a-f]{2}:){5}[0-9a-f]{2}' ~/omarchy-debug-review.log
grep -nF "$(hostname)" ~/omarchy-debug-review.log

The first reads 0. The second counts case-insensitively, which is what catches the uppercase form BlueZ prints, and reads 0 unless a colon-separated fingerprint such as SHA256:aa:bb:... remains, which is not an address. The third prints nothing, and any line it does print, for example an Avahi Host name is ... message, is masked by hand before you attach the file.

Sources

omarchy-debug-exposes-lan-macs-and-ips