chore(todo): archive completed tasks to Resolved, age out to task-archive

2 files changed, 391 insertions, 424 deletions

diff --git a/archive/task-archive.org b/archive/task-archive.org
index 1c8f170..5abdbba 100644
--- a/archive/task-archive.org
+++ b/archive/task-archive.org
@@ -460,3 +460,123 @@ Resolved 2026-06-14: the runnable script already existed — =scripts/package-in
 ** DONE [#C] paru vs yay — evaluated, staying with yay
 CLOSED: [2026-06-10 Wed]
 Research done 2026-06-10: [[file:docs/2026-06-10-paru-vs-yay-evaluation.org][docs/2026-06-10-paru-vs-yay-evaluation.org]]. The maintenance picture inverted since the task was filed: yay released v12.6.0 on 2026-06-07 with active triage, while paru has had no release in 11 months, no commit in 5, and a stable that fails to build against current libalpm (issue #1468 open 6 months). For an installer that bootstraps the AUR helper unattended, paru is the riskier choice on every axis that matters. No decision needed — the evidence closes this one; revisit only if paru's maintenance resumes.
+** DONE [#B] Idle-inhibitor keybind + synced waybar indicator :hyprland:waybar:
+CLOSED: [2026-06-23 Tue]
+Shipped 2026-06-23 as dotfiles commit =a004201=. Super+I toggles the hypridle daemon (kill = inhibit, relaunch = restore). The built-in waybar =idle_inhibitor= module was replaced with a =custom/idle= module backed by a =waybar-idle= script, so the keybind, the bar click, and the icon share one source of truth (whether hypridle is running) and stay in sync. Icons 󰒳 inhibited / 󰒲 active, with a 5s poll safety net. Freed =Super+I= by pruning the unused ai-term pyprland scratchpad from both host configs. TDD'd (=waybar-idle= + =hypridle-toggle= suites); dupre/hudson theme CSS updated. From a home-project handoff 2026-06-23; Craig confirmed it works live.
+** DONE [#B] Verify package signature verification not bypassed by --noconfirm
+CLOSED: [2026-06-23 Tue]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-05-21
+:END:
+Audited 2026-06-23. =--noconfirm= does not bypass signature verification — it only auto-answers interactive prompts. Signature checking is governed by =SigLevel= in =/etc/pacman.conf=, which archsetup leaves at the Arch default (=Required DatabaseOptional=): its only pacman.conf edits are ParallelDownloads, Color, and enabling multilib (=archsetup:913,917=), none of which touch =SigLevel=. So every repo package stays signature-verified regardless of =--noconfirm=.
+
+One real integrity bypass exists, and it is not =--noconfirm=: =archsetup:2403= runs =yay -S --noconfirm --mflags --skipinteg python-lyricsgenius=, where =--skipinteg= skips makepkg's checksum and PGP-signature checks for that one AUR package (a documented workaround for an expired-signature issue upstream). It's scoped to a single package, not global. Tracked for periodic re-check below.
+** DONE [#C] Harden sshd in the installer (explicit prohibit-password) :solo:
+CLOSED: [2026-06-24 Wed]
+Done 2026-06-24: the openssh block (=archsetup:1271-1277=) now writes =/etc/ssh/sshd_config.d/10-hardening.conf= with =PermitRootLogin prohibit-password= and reloads sshd, right after starting the service. =PasswordAuthentication= left untouched so ssh-copy-id to the user still works. Makes the posture intentional rather than dependent on the upstream default. Velox and ratio (which carried an explicit =PermitRootLogin yes= at =sshd_config:33= from earlier provisioning) were already fixed by hand 2026-06-23. Verified =bash -n= + =shellcheck -S error= clean; full drop-in-on-fresh-install confirmation is VM-deferred (the unit harness covers helpers, not inline install steps).
+** DONE [#C] Build security dashboard command :solo:
+CLOSED: [2026-06-23 Tue]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-05-21
+:END:
+Shipped 2026-06-23 as dotfiles commit =1b9b205=: =security-status= (=common/.local/bin=, on PATH). Read-only dashboard showing disk encryption (LUKS *and* ZFS native — the fleet runs ZFS, so a LUKS-only check would have falsely reported "no encryption"), ufw state, externally-reachable ports (counts all listening, lists only the non-loopback exposures), and running/failed service counts. Command lookups are env-overridable; parsing covered by unit tests against canned output. New file, so ratio needs =git pull && make stow hyprland= to link it.
+** DONE [#C] Teach archsetup to stow the host tier :solo:
+CLOSED: [2026-06-23 Tue]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-06-11
+:END:
+Already implemented in =user_customizations()= (=archsetup:1049-1058=): after stowing =common= + the DE package, it derives =host_tier="$(cat /etc/hostname 2>/dev/null || uname -n)"= and stows that package when =$dotfiles_dir/$host_tier= exists, else prints "no host tier for '<host>' — skipping". The =/etc/hostname=-first detection is the right call for install time (=uname -n= still reports the ISO's name until reboot), and it's the same skip-if-absent semantics as the dotfiles Makefile. Verified by reading the installer 2026-06-23; no code change needed.
+** DONE [#C] Waybar indicators unevenly spaced :quick:solo:waybar:
+CLOSED: [2026-06-24 Wed]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-06-24
+:END:
+The right-side module icons don't sit at even intervals — spacing reads as inconsistent across the group. Noticed 2026-05-21 after adding the airplane indicator.
+
+Done 2026-06-24: a screenshot showed the standalone module icons were already even — the unevenness was the tray, whose icons clustered tight (tray =spacing: 4= vs the ~0.3rem margins on every other module). Bumped tray =spacing= 4 → 10 in the waybar =config=; restarting waybar and re-screenshotting confirmed the row reads even. The lever was the tray spacing, not the per-module CSS the original body guessed at.
+** DONE [#B] Separate mpd playlist_directory from music_directory :mpd:music:quick:
+CLOSED: [2026-06-24 Wed]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-06-24
+:END:
+Done 2026-06-24 (dotfiles a9bfdf3): set =playlist_directory= to =~/.local/share/mpd/playlists= (separate from =music_directory= ~/music). git-moved the 73 radio-stream playlists from =common/music/= into =common/.local/share/mpd/playlists/= (history preserved); dropped the empty =60s Sounds.m3u= (Craig's call); git rm'd the stray =Black Flamingos - Space Bar.m4a= and moved the real track into the music library. Curated playlists left flat in ~/music (Craig's call — avoids rewriting the 7 relative-path ones). The ~/music/radio orphan was already gone. Relinked surgically (a pre-existing =whereami= stow conflict blocked a full =stow common=). mpd restarted clean: 73 radio playlists load from playlist_directory (verified SomaFM stream URLs), 24 curated browsable from the music tree. ratio needs the same restow + mpd restart on its next pull (reminder filed). Decisions answered: 60s dropped, curated flat.
+Spec written and approved (option 1), pinned before execution on 2026-06-03. Root issue: mpd.conf has =playlist_directory= == =music_directory= == ~/music, so the whole audio library is the playlist store and radio streams mix with curated playlists. Option 1: radio stream playlists (portable, 73 in the dotfiles repo) move to a dedicated =playlist_directory= (=~/.local/share/mpd/playlists=) via stow; the 22 curated local playlists (machine-specific track refs) live in the music tree. Also removes the broken ~/music/radio/ orphan (73 dead symlinks).
+
+Full step-by-step spec (mpd.conf edit, repo restructure of =common/music/= → =common/.local/share/mpd/playlists/=, curated relocation, restow, verification incl. the 7 relative-path curated playlists, ratio propagation) is in the 2026-06-03 session record under .ai/sessions/. Two open decisions before executing: (1) drop the empty =60s Sounds.m3u= or refill with the SomaFM 60s URL; (2) curated playlists into =~/music/playlists/= subdir vs leave flat in ~/music/. Side cleanup surfaced: a stray audio file =Black Flamingos - Space Bar.m4a= is wrongly committed in the dotfiles repo's =common/music/= — git rm it and move to the synced library.
+** DONE [#C] Install adopted modern CLI tools :tooling:solo:
+CLOSED: [2026-06-24 Wed]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-06-24
+:END:
+Done 2026-06-24: added bat/dust/hyperfine/doggo to archsetup General Utilities (tealdeer was already declared), installed all five on velox, set =BAT_THEME=ansi= in =common/.profile.d/tools.sh= (tracks the dupre terminal palette), seeded the tldr cache. ratio still needs the =pacman -S= (additive; lands on its next archsetup run).
+Decision (Craig, 2026-06-24): adopt all five recommended tools — =bat=, =dust=, =hyperfine=, =tealdeer=, =doggo= (all in extra). Add them to archsetup's package list and install on both machines. Optional candidates (=xh=/=jless=/=sd=/=ouch=) declined for now. Full evaluation: [[file:docs/2026-06-10-modern-cli-tools-evaluation.org][docs/2026-06-10-modern-cli-tools-evaluation.org]].
+
+- Add the five to the appropriate pacman package section in =archsetup=.
+- =pacman -S bat dust hyperfine tealdeer doggo= on velox + ratio.
+- =bat=: set =BAT_THEME= to match the dupre palette once installed.
+- =tealdeer=: run =tldr --update= to seed the cache after install.
+** DONE [#C] Review file manager options for Wayland
+CLOSED: [2026-06-24 Wed]
+Decision (Craig, 2026-06-24): keep nautilus only; skip yazi. File management lives in Emacs dired plus the Super+F dirvish popup, so a TUI file manager has no daily user here. ranger was already ruled out (frozen upstream). Full evaluation: [[file:docs/2026-06-10-file-manager-evaluation.org][docs/2026-06-10-file-manager-evaluation.org]]. Follow-on surfaced: nautilus needs dark theming (filed as its own task).
+** DONE [#B] Theme nautilus to a dark theme :bug:solo:
+CLOSED: [2026-06-24 Wed]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-06-24
+:END:
+nautilus rendered blindingly white (Craig, 2026-06-24). As a GTK4/libadwaita app it follows the appearance portal's =org.freedesktop.appearance color-scheme=, which mirrors =org.gnome.desktop.interface color-scheme=. Two stacked causes:
+
+1. velox had no system-wide dconf db at all — no =/etc/dconf/profile/user=, no =/etc/dconf/db/site.d/00-archsetup-defaults=, no compiled =site= db — so archsetup's declared default (=color-scheme='prefer-dark'=, =archsetup:1109-1119=) never reached the machine (velox predates that block). Created the profile + site defaults as archsetup writes them and ran =dconf update=. =gsettings get= then returned =prefer-dark=.
+
+2. That alone did NOT fix the running session: a system-db default emits no GSettings change signal, so the appearance portal kept reporting =0= (no-preference → light), and libadwaita reads the portal, not =GTK_THEME=. (An early screenshot looked dark only because the shell env carries =GTK_THEME=Adwaita:dark=, which Hyprland-launched apps don't inherit — masking the real state.) Fix: a user-level =gsettings set org.gnome.desktop.interface color-scheme prefer-dark=, which signals the portal live. It now reports =1=, and a portal-driven nautilus (GTK_THEME unset) renders dark — screenshot-verified.
+
+Durable: the user value persists in =~/.config/dconf/user=; archsetup's system-db handles fresh installs (the portal reads the default fresh at login, so no signal is needed there). No archsetup change. ratio may need the same one-two — see the Active Reminder.
+** CANCELLED [#D] Test wlogout menu on laptop
+CLOSED: [2026-06-24 Wed]
+Merged into the "Wlogout exit-menu buttons are rectangular, not square" task ([#C]) — same effort (per-host wlogout button sizing across velox/ratio). The fixed-pixel-margins hint was folded into that task's body.
+** DONE [#B] Enlarge org-capture popup to scratchpad size :hyprland:
+CLOSED: [2026-06-24 Wed]
+From a .emacs.d inbox handoff (2026-06-15, captured via roam): the quick-capture / org-protocol popup is too small to be effective — it should be about the size of a terminal scratchpad.
+
+*** 2026-06-24 Wed @ 17:21:11 -0400 Sized the popup to the scratchpad, per-host in pixels
+The 06-15 read was wrong: the real size lever is the Hyprland window rule, not the quick-capture char-cell count. The =size 900 500= rule on the org-capture window pinned it to 900x500 regardless of the frame's requested geometry (demoing 120x24 vs 180x32 looked identical because both clamped to 900x500). Tried a percentage rule (=size 75% 70%=) to auto-adapt per host like the pyprland scratchpad — native window rules do NOT honor percentages (only pyprland does), so the frame fell back to char-cell geometry and overflowed the screen. Fix: absolute pixels matching each host's terminal scratchpad, placed in the host tier (=<host>/conf.d/local.conf=) since pixels don't adapt across monitors. velox = 1078x671 (75%x70% of its 1437x958 logical desktop) — verified on-screen. ratio = 1892x936 (55%x65% of 3440x1440) — set but not yet eyeballed on ratio (tracked as an Active Reminder in notes.org). The shared hyprland.conf keeps float/center/stay_focused and a comment pointing at the per-host size. dotfiles change — needs commit in =~/.dotfiles=.
+
+*** 2026-06-15 Mon @ 19:19:55 -0500 AI Response: popup size is the frame's char-cell count, not the Hyprland rule
+Triaged under auto inbox-zero. The popup is the emacsclient frame named "org-capture", created by =~/.dotfiles/hyprland/.local/bin/quick-capture= with =(width . 90) (height . 22)= — 90 columns by 22 lines. Emacs sizes by character cells and overrides the Hyprland rule =windowrule = match:title ^(org-capture)$, size 900 500= (hyprland.conf:182). The live frame measured ~889x860 px; the width tracks the 90-column count, not the window rule. Setting the Hyprland rule to =size 55% 65%= (the scratchpad's pyprland spec) did not change the frame width, so I reverted it — dotfiles left clean.
+
+Real lever: the column/line count in the quick-capture script. Scratchpad reference on ratio (DP-4, 3440x1440) is 55% 65% ~= 1892x936 px ~= 190 cols by 24 lines. Why this isn't a solo auto-fix — it needs a tradeoff decision:
+- The script lives in the shared =hyprland/= stow tier, so a fixed ~190 columns overflows velox's 1920-wide laptop, and 24+ lines overflows velox's 1080 height (22 lines ~= 860 px is already near the safe max there).
+- Emacs char-cell sizing doesn't adapt to the monitor the way pyprland's percentage does, so "scratchpad-size on both machines" needs one of: a fixed compromise count, a per-host override via the ratio/velox tiers, or a script that computes columns from the active monitor.
+Options to weigh: (a) a safe-on-both compromise like width 120-130 / height 24; (b) per-host width through the ratio/velox tiers; (c) dynamic sizing in quick-capture from =hyprctl monitors=. Pick the tradeoff and I'll implement.
+** DONE [#C] Highlight current month and year in the calendar hover :feature:waybar:quick:solo:
+CLOSED: [2026-06-24 Wed]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-06-24
+:END:
+From the roam inbox (2026-06-24): the waybar clock's calendar tooltip highlights today's date in goldenrod; the current month and year header should be goldenrod too.
+
+Done 2026-06-24: the date module is the custom =waybar-date= script (not the built-in clock), so the highlight lives in its tooltip markup. Added a sed wrapping line 1 of the current-month =cal= output (the centered "Month Year") in the same =#daa520= goldenrod the day highlight uses. Verified the tooltip JSON carries =<span color='#daa520'><b>June 2026</b></span>= with today's highlight intact and waybar live; the on-hover look is Craig's spot-check.
+** DONE [#C] Wallpaper-set from dirvish doesn't work on Wayland :hyprland:
+CLOSED: [2026-06-24 Wed]
+From the roam inbox (2026-06-24, claimed for archsetup by Craig): typing =bg= in the dirvish popup doesn't change the wallpaper — Craig's read is it may still be wired to feh/X11 instead of a Wayland utility.
+
+Findings (2026-06-24): the Wayland wallpaper utility on this setup is =awww= (waypaper's configured =backend = awww=; =set-theme= sets the default via =awww img <file>=). There was no shared wallpaper script (=bg= on PATH is just the shell builtin), and the dirvish =bg= command lives in the Emacs config, so it was calling the wrong (or no Wayland) setter.
+
+Done 2026-06-24 (dotfiles 8be2484): added =set-wallpaper <image>= to the hyprland tier — sets live via =awww img= and persists the choice into =waypaper/config.ini=, the single Wayland-correct entry point. Resolves relative paths, validates the file, exits non-zero without persisting if awww fails. 8 Normal/Boundary/Error tests green; live-verified (awww set it, config rewrote). Notified =.emacs.d= to point the dirvish =bg= command at =set-wallpaper <file>= — that wiring is its piece (dependency cleared, =:blocker:= dropped).
+
+Follow-up (separate, small): the login restore =exec-once= in =hyprland.conf= is hardcoded to =trondheim-norway.jpg=, so a wallpaper set via =set-wallpaper= shows live but won't survive a relogin until the exec-once becomes =waypaper --restore= (which reads the now-persisted config). Filed below.
+** DONE [#C] Proton Mail Bridge font size :chore:quick:
+CLOSED: [2026-06-24 Wed]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-06-24
+:END:
+From the roam inbox (2026-06-22): adjust the Proton Mail Bridge UI font to a comfortable size. The bridge is a Qt app, so it likely keys off Qt scaling or the qt5ct/qt6ct config like the other Qt apps (QT_SCALE_FACTOR or a font setting).
+
+Done 2026-06-24 (dotfiles =hyprland.conf:47=): the bridge is a Qt6 *QML* app, so it ignores the qt6ct General font — bumped the UI font via =QT_FONT_DPI= on the autostart instead. Changed the exec-once to =env QT_FONT_DPI=108 protonmail-bridge --no-window= (default DPI is 96; 108 = 1.125x). Iterated live with Craig: 120 too big, 108 comfortable. hyprland.conf is a stow symlink so the change is already live; applies at every login. The =~/.config/autostart/Proton Mail Bridge.desktop= entry is dormant under Hyprland (no XDG-autostart), so it was left as-is.
+** DONE [#C] Wallpaper login-restore is hardcoded, not waypaper --restore :hyprland:quick:solo:
+CLOSED: [2026-06-24 Wed]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-06-24
+:END:
+The Hyprland =exec-once= (=hyprland.conf:26=) restores the wallpaper with a hardcoded =awww img ~/pictures/wallpaper/trondheim-norway.jpg=, so any wallpaper set later (via =set-wallpaper=, waypaper, or the dirvish =bg=) reverts on relogin. =set-wallpaper= now persists the choice to =waypaper/config.ini=, so switch the exec-once to =waypaper --restore= (after =awww-daemon= is up) to make set wallpapers survive a relogin. Small, dotfiles-only; verify by setting a different wallpaper, relogging, and confirming it sticks.
+
+Done 2026-06-24 (dotfiles): swapped the line-26 exec-once from the hardcoded =awww img …/trondheim-norway.jpg= to =awww-daemon & sleep 1 && waypaper --restore=. waypaper has a real =awww= backend (in its =--backend= list), the stowed =waypaper/config.ini= carries =backend = awww= plus a default =wallpaper == line, so =--restore= works on a fresh install too. Mechanism verified live: =waypaper --restore= reapplied the persisted wallpaper via awww, exit 0. Relogin confirmation filed under "Manual testing and validation". Follow-up filed: =set-wallpaper='s =mv= detached the live =waypaper/config.ini= from its stow symlink, so set-wallpaper changes no longer flow back to dotfiles.
diff --git a/todo.org b/todo.org
index 19e5400..70e058b 100644
--- a/todo.org
+++ b/todo.org
@@ -24,156 +24,17 @@ The vocabulary is open — topic tags are coined as needed — so these are conv
 ** TODO [#B] Audio panel spec :feature:waybar:audio:
 Work Craig's ask (roam inbox, 2026-07-02) into a spec, net/bt-panel kin: an audio panel replacing the pypr audio scratchpad (Super+A) with the same functionality — change the default/active output (speaker) and input (mic), volume control for both. The one new capability: a push-to-talk mic mode for meetings — mic stays muted except while the space bar is held, releasing re-mutes. (Hold-to-talk under Wayland needs a global key grab — likely a hyprland bind pair on press/release or an evdev listener; feasibility research belongs in the spec.) Related current bindings: Super+M audio-cycle ring, Super+Shift+A mic-toggle.
 
-** DONE [#B] Network panel UI — review findings :feature:waybar:network:
-CLOSED: [2026-07-01 Wed]
-Full UI review 2026-07-01 (visual walk of every state + code pass over the view logic), 30 findings: 21 from the agent review + color audit, 9 from Craig. All fixed the same night in a no-approvals speedrun, five commits, each test-gated (33 suites) and the whole panel re-verified via AT-SPI smoke + screenshots.
-
-*** 2026-07-02 Wed @ 00:00 -0400 Theme pass: contrast, hierarchy, focus, hover, destructive, dialogs (dotfiles 82aad0b, 998829b)
-Selected-row captions turned cream (the dim gray measured 2.2:1 on the slate fill, a WCAG fail on the auto-selected active row). Names gained weight and captions dropped a size (type hierarchy beyond color). Focus rings gold, scrollbars slim slate, rows got a hover wash. Disconnect and Forget became terracotta destructive-action (red = off). The Join/Add dialogs carry the dupre contract now (ground, mono, styled entries with a gold focus border, no stock blue anywhere), the Add dialog disables its action button until the SSID is non-empty, its password field gained the peek icon, and long SSIDs ellipsize.
-
-*** 2026-07-02 Wed @ 00:00 -0400 Connections logic: lies, gaps, races (dotfiles 693f820)
-Saved profiles stopped claiming "open" (security shows only when the scan knows it; subtitles are view-aware — Available adds signal %, Saved says just "out of range"). An active wired connection pins above the wifi scan in Available. Add connects immediately (Craig's decision), both add dialogs' action reads Connect. Rescan went through the op state machine (the dead guard let double-clicks double-scan). A failed initial load shows in the boxes with a retry instead of stranding "Loading…". Available's first message is "Scanning networks…". Live-info ages humanize (7m, not 445s). A held portal replaces the internet line (stable row height under the poll). The poll pauses on other tabs. VPN profiles got a VPN glyph. Forgetting the active network warns it will disconnect you.
-
-*** 2026-07-02 Wed @ 00:00 -0400 Diagnostics restructure: selector, live speed test, streamed verdicts, leaner chrome (dotfiles 787b475)
-The six-button wall became a dropdown tool selector (full tool names, a description of the selection, one Run button) revealed under Advanced. Speed Test became Network Performance: a reveal with Run Speedtest + Stop (no button-flips-to-Cancel), a once-a-second elapsed ticker while running, and Download / Upload / Ping (high-latency warn) / Server / Tip rows in the diagnose aesthetic. Get Me Online streams the diagnose rows first, announces each attempt ("Trying Reset Connection…"), then its result, and closes with a bold verdict row — as do diagnose, the single tools, and the portal login; verdicts left the status line. Get Me Online at a held captive portal opens the login flow (doctor runs the safe, reversible portal-login when fix is requested). Connecting to a network that turns out captive sends a desktop heads-up, waits a beat, then opens the login page. Diagnose evidence humanized ("open internet (HTTP 204)", "names resolve (captive.apple.com)"). The title row and Close button are gone (Esc + focus-loss auto-hide cover a transient popup) and the status line became a self-clearing toast. The AT-SPI driver anchors on the Diagnostics tab now.
-
-** DONE [#B] Advanced repair buttons: half width, two per row :feature:waybar:network:quick:
-CLOSED: [2026-07-01 Wed]
-The wide Advanced buttons shrink the panel and leave the diagnostics output impossible to read. Make each half width, two to a row, and rename where needed to fit. Origin: roam inbox capture.
-
-Done 2026-07-01 (dotfiles aca6827): the Advanced reveal became a 3-column, 2-row grid (Diagnose, Unblock WiFi, Reset / Restart, Test DNS, Force Portal) per Craig's follow-up; labels shortened, tooltips carry the full descriptions. Verified live + AT-SPI smoke.
-
-** DONE [#B] Panel action-button rows fill the panel width :feature:waybar:network:quick:
-CLOSED: [2026-07-01 Wed]
-Disconnect / Rescan / Add / Add Hidden, and the Saved row, should be as wide as the panel and the buttons above them. Apply the same homogeneous full-width treatment used on the Available / Saved sub-tabs. Origin: roam inbox capture.
-
-Done 2026-07-01 (dotfiles aca6827): both Connections action rows are homogeneous full-width, which also fixed the panel resizing when Connect flips to Disconnect. Verified live.
-
-** DONE [#B] Live connection info in the row subtitle :feature:waybar:network:
-CLOSED: [2026-07-01 Wed]
-The live connection information shown in the row hover should also appear in the small print under the connection name, updated in realtime like the hover. Origin: roam inbox capture.
-
-Done 2026-07-01 (dotfiles aca6827): a 1.5s poll fills the active row's subtitle with the bar-tooltip fields minus the SSID (signal, interface, internet + age, portal note, throughput), sharing the bar's per-field formatters. Verified live.
-
 ** TODO [#B] Network panel: other network interfaces (tailscale, VPNs, wireguard) :feature:waybar:network:
 Consider displaying other relevant network info in the panel: tailscale, installed VPNs (can we turn them on/off here?), wireguard — and whatever else is in this category. Pretty big; probably deserves its own spec. Origin: roam inbox capture 2026-07-02.
 
 Initial spec written 2026-07-02: [[file:docs/design/2026-07-02-net-panel-other-interfaces-spec.org]] (DRAFT — four decisions await Craig's review before build).
 
-** DONE [#B] Right-click date/time: ntp sync + timezone update :feature:waybar:
-CLOSED: [2026-07-02 Thu]
-Right-click on the date updates the clock from ntpd (or whatever keeps the clock in sync); right-click on the time runs the update-timezone script. Neither opens a terminal — both run transparently in the background. Errors surface as notifications. The module should detect whether we're online and, if not, show a message instead of running the script. Origin: roam inbox capture 2026-07-02.
-
-Shipped 2026-07-02 (dotfiles 2f7993d). Date right-click = clock-sync (chronyc makestep behind a connectivity gate; offline/captive get explanatory notifications). Time right-click = timezone-set, rewritten to prefer WiFi geolocation (whereami → timeapi.io) over IP lookup — the hotel IP geolocated two timezones off, and ipapi.co is paywalled now (ipinfo.io is the fallback). Both promptless (sudo -n), all outcomes notify, worldclock gained signal 16 for an instant refresh. 13 new tests across clock-sync + timezone-set; live-verified on velox (WiFi path resolved Rhode Island → America/New_York correctly).
-
-** DONE [#B] Screenshot "view image" option :feature:hyprland:
-CLOSED: [2026-07-02 Thu]
-The screenshot flow should also offer a "view image" selection: saves the shot, opens it in a viewer, and puts the path on the clipboard. Origin: roam inbox capture 2026-07-02.
-
-Shipped 2026-07-02 (dotfiles 10e5961). View Image entry in the post-capture fuzzel menu: opens the shot via xdg-open (default viewer, currently feh) and puts the path on the clipboard. Script gained env seams + a nine-test suite (menu dispatch, both capture modes, cancel, failure). Live check pending: take a shot and pick View Image once.
-
-** DONE [#C] Collapse-triangle buttons: dimmer, inlaid styling :waybar:
-CLOSED: [2026-07-02 Thu]
-The triangle collapse buttons should look embedded (inlaid) and be slightly dimmed so they don't compete for eye attention with the other components. Origin: roam inbox capture 2026-07-02.
-
-Shipped 2026-07-02 (dotfiles 15cb93c): muted color + dark inset well + smaller glyph; hover still brightens. Hudson theme carries the same shape. Screenshot-verified.
-
-** DONE [#C] Contrast button ignores the white=on / red=off paradigm :bug:waybar:
-CLOSED: [2026-07-02 Thu]
-The contrast button doesn't respect the white=on, red=off color paradigm the other waybar modules follow. Cosmetic × every time = P3. Origin: roam inbox capture 2026-07-02.
-
-Shipped 2026-07-02 (dotfiles 15cb93c). The "contrast button" is the auto-dim module — its ON icon (nf-fa-adjust) is the classic contrast glyph. It showed gold when on and nothing when off; now on = default silver, off = terracotta, matching every other toggle. Screenshot-verified both states.
-
-** DONE [#C] Off-state red inconsistent across waybar modules :bug:waybar:
-CLOSED: [2026-07-02 Thu]
-Terracotta red isn't applied uniformly for "off": the sleep icon is a different shade than the mouse/trackpad when off, and the dim indicator doesn't show red when off at all. Cosmetic × every time = P3. Origin: roam inbox capture 2026-07-02.
-
-Shipped 2026-07-02 (dotfiles 15cb93c + 7f1f334). The touchpad script's Pango-markup red predated the terracotta theme pass (#d47c59 vs the CSS's #cb6b4d) — unified on #cb6b4d. Dim-off now red (see the contrast task). Bonus find: waybar hands every pulseaudio instance the sink's .muted class, so a muted speaker also painted the mic red — scoped with :not(.mic) so each glyph keys on its own device.
-
-** DONE [#B] Waybar volume/mic toggle like the touchpad module :feature:waybar:
-CLOSED: [2026-07-02 Thu]
-Make the volume/mic waybar component look and behave like the touchpad/mouse toggle.
-- Move the mic to the other side of the volume so the percentage isn't in the way. The mic and speaker icons sit the same distance apart as the hand and mouse.
-- One keybinding cycles the four states: volume on / mic on, volume on / mic off (red), volume off (red) / mic on, volume off (red) / mic off (red).
-- Move the trackpad/mouse toggle to another keybinding (discuss an open mnemonic, e.g. =d= for disable) and assign Super+M to this module (for mute).
-Origin: roam inbox capture.
-
-Shipped 2026-07-02 (dotfiles 7f1f334). New audio-cycle script walks the four-state ring in the exact order above (wpctl-backed, explicit set-mute so the pair can't desync, 6 tests) on Super+M; live-verified the full ring on velox. Mic moved left of the speaker and hugs it via paired margins, percentage on the outside. Touchpad toggle moved to Super+Shift+I ("input devices" — d-for-disable was taken at both levels by removemaster and dim-toggle); its tooltip and tests follow.
-
-** DONE [#C] Timer end sends no notification :bug:waybar:
-CLOSED: [2026-07-02 Thu]
-The end of a wtimer timer didn't fire a desktop notification. Needs reproduction to confirm frequency; priority follows the severity-by-frequency matrix once known (a reliably-missing timer-end alert would rate higher). Origin: roam inbox capture.
-
-Root-caused 2026-07-02 (dotfiles ca35642). The pipeline works (a live 3s timer fired and persisted), but notify sent everything --urgency=normal and dunstrc delays normal-urgency popups while a fullscreen window has focus — a timer ending mid-video sat invisible until fullscreen exit. Alarms now go critical urgency, which rides the fullscreen_show_critical rule; verified CRITICAL in dunst history. The alarm sound (paplay, separate from dunst) was never affected.
-
-** DONE [#B] Bake captive-portal login into the net panel :feature:network:
-CLOSED: [2026-07-01 Wed]
-Make the captive-portal login a first-class net-panel feature instead of the one-off =~/.local/bin/hotel-wifi= script. When the engine sees a held portal, offer "Log in to this network" that runs the plain-DNS + clean-browser flow reversibly (disable DoT -> recover the portal URL from the redirect -> open a clean Chrome profile -> restore DoT when online). Reconcile with the existing =net portal= / =captive= helper, whose DNS-hijack-to-gateway model did NOT match the real Hyatt portal.
-
-Full mechanism writeup, the working script, and the integration plan: [[file:docs/design/2026-06-30-captive-portal-login.org]]. From the 2026-06-30 Hyatt saga.
-
-*** 2026-06-30 Tue @ 11:40 -0400 Engine core landed (dotfiles a7d7559)
-Replaced =net portal='s old captive-helper hand-off with a =portal-login= repair tier: drop DoT to plain DNS, probe the portal URL (302 / meta-refresh), open a throwaway browser profile, spawn a detached watcher that restores DoT once online (or on timeout). =net portal --restore= is the manual fallback. 7 tests. So =net doctor= / the bar's =net portal= hookups already run the real flow now. Remaining: (1) name the DoT-blocking cause in =net diagnose=; (2) a dedicated "Log in to this network" button in the panel's Diagnose/Repair tab (today it rides the generic =net portal=); (3) live validation against a real captive portal (unit-tested only — didn't run it live to avoid disrupting a meeting).
-
-*** 2026-07-01 Wed @ 22:41:51 -0400 Live-validated end to end against a local captive simulator (dotfiles c1401db)
-The last remainder. tests/net/captive_sim.py is a local redirect portal (302s to a login page until "logged in", then a clean 204). NET_PROBE_URL and NET_PORTAL_TRIGGERS point the whole flow at it (an overridden probe skips the interface binding, which can't reach loopback). Ran live on velox, both restore paths verified: online-detect (login click, watcher saw the 204, DoT drop-in restored within ~2s, clean exit) and the timeout fallback (a watcher that never saw online restored DoT at its 300s deadline). Real sudo mv, real resolved restarts, real redirect URL recovery, real clean-profile Chrome — against a temp drop-in dir, so live DNS was untouched. All three remainders are done; the task is closed. The remaining what-if is a real venue's walled-garden quirks, which only an actual portal exercises.
-
-*** 2026-07-01 Wed @ 21:44:05 -0400 Diagnose names the DoT block; panel gained Log in to This Network (dotfiles 51e0e2d)
-Remainders 1 and 2 landed. The dns-resolve step names the DoT pin when resolution is dead and the drop-in exists (sysio.dot_forced), and routes next_action to the portal login. The panel's hidden Open Portal button became a first-class suggested-action "Log in to This Network", shown whenever the report holds a portal signal (portal step with or without a URL, or the DoT-blocked resolution) via the unit-tested viewmodel.wants_portal_login. TDD, 33 suites green. Remainder 3 (live validation against a real portal) still open.
-
-*** 2026-06-30 Tue @ 14:59:53 -0400 Live test on velox surfaced two fixed bugs + a deeper follow-up
-Force portal (panel Repair tab) = =net-popup net portal= = the same portal-login tier. Tested live on @Hyatt_WiFi (already authorized, so no real intercept). Two bugs fixed in dotfiles (TDD, full suite green):
-- Chrome first-run wizard fired on every launch — =_open_portal= made a fresh tempfile profile but passed no first-run flags. Added =--no-first-run --no-default-browser-check= + a unit test.
-- Flashing sudo prompt for the DoT drop + pointless resolved restart on velox, where the DoT drop-in the code looks for (=/etc/systemd/resolved.conf.d/dns-over-tls.conf=) doesn't exist. Guarded =_disable_dot=/=_restore_dot= to be true no-ops (no sudo, no restart) when there's no DoT drop-in to move; tests assert no systemctl call fires.
-
-** DONE [#B] Consistent red=off across waybar toggle modules :waybar:
-CLOSED: [2026-07-01 Wed]
-Extend the red=off convention (just added to the touchpad/mouse indicator) to the other toggles — sound volume, microphone mute, and caffeine — so a disabled / muted / off state reads red across the board. Skip the "cross"/slash; the color alone carries it. Origin: roam inbox capture.
-
-Already implemented (verified 2026-07-01): =style.css= gives =#pulseaudio.muted=, =#pulseaudio.mic.source-muted=, and =#custom-caffeine.inhibited= the off-state color =#d47c59=, matching =#custom-touchpad.disabled=. Note: caffeine's red fires on =.inhibited= (caffeine ON / staying awake), which is arguably the inverse of "off" — leave as-is unless you want strict off=red semantics there.
-
-** DONE [#B] Microphone-mute keybind :feature:waybar:quick:
-CLOSED: [2026-07-01 Wed]
-A keyboard shortcut to toggle the mic mute. The pulseaudio#mic module shows the state but there's no hotkey to flip it. Wire a hyprland bind to a mic-mute toggle. Origin: roam inbox capture.
-
-Already implemented (verified 2026-07-01): hyprland.conf binds both =XF86AudioMicMute= and =Super+Shift+A= to =mic-toggle= (no conflict — airplane is Super+Shift+X).
-
 ** TODO [#B] File-manager swallow pattern :feature:hyprland:
 When the file manager launches another app, it should hide to a special workspace (the "swallow" pattern) and return when that process ends, rather than vanishing. Today it disappears with no signal of whether it's coming back, so the user can't tell success from failure — they should quit explicitly instead. Origin: roam inbox capture.
 
-** DONE [#C] Keybind hints in waybar module tooltips :waybar:
-CLOSED: [2026-07-02 Thu]
-Every module's hover tooltip should list its keyboard shortcut(s), for discoverability. Audit the modules and add the bindings to each tooltip. Origin: roam inbox capture.
-
-Shipped 2026-07-02 (dotfiles 4c32aec). Audited every module: arrows (Super+[ / Super+]), sysmon (Super+R), net (Super+Shift+N), layout (Super+Shift+←/→), menu (Super+Space / Super+Shift+Q, tooltip enabled), plus the already-hinted dim/caffeine/touchpad/mic/volume. Date/time tooltips document their new right-click actions. Workspaces/window/tray don't take custom tooltips; timer has no keybind.
-
-** CANCELLED [#C] Smooth waybar expansion animation :waybar:
-CLOSED: [2026-07-02 Thu]
-The cluster expansion jumps instead of animating, and a few systray icons pop in one-by-one afterward, which reads as glitchy. Animate the expansion smoothly if waybar allows it — width transitions are limited, so feasibility is uncertain (hence [#C]). Origin: roam inbox capture.
-
-Assessed infeasible 2026-07-02: collapse works by config rewrite + SIGUSR2 reload, which rebuilds every widget — nothing survives to transition, GTK3 can't animate add/remove without Revealer (an upstream waybar change), and the tray pop-in is async StatusNotifier re-registration. Full findings + revisit conditions: [[file:docs/design/2026-07-02-waybar-expansion-animation-feasibility.org]].
-
-** DONE [#C] Optional label on timer/alarm/stopwatch items :feature:waybar:
-CLOSED: [2026-07-02 Thu]
-Let each wtimer item carry an optional short text label. The data model already supports it (=add_timer/add_alarm/add_stopwatch/add_pomodoro= all take =label=""=, and =_describe= shows =label or type=); the gap is the fuzzel-driven creation flow, which doesn't prompt for a label. Add the optional label prompt on create. Origin: roam inbox capture.
-
-Shipped 2026-07-02 (dotfiles ca35642): =wtimer new= gained a "label (optional)" fuzzel prompt after the type/value prompts; empty keeps the unlabeled default. 2 new tests (89 total in the suite).
-
 ** TODO [#C] Open meeting links in the browser instead of the Zoom app :feature:
 Route Zoom (and similar) meeting links to the browser web client rather than launching the native Zoom app — e.g. firefox/chrome extensions, or a =zoommtg://= URL handler / desktop-file override that rewrites to the web-client URL. Decide the mechanism (extension vs handler) before building. Origin: roam inbox capture.
 
-** DONE [#C] Alarm tooltip shows time remaining, not alarm time :bug:waybar:quick:
-CLOSED: [2026-07-01 Wed]
-The =wtimer= alarm tooltip displays the countdown (time remaining) instead of the alarm's wall-clock fire time. For an alarm set to 2:00pm, the tooltip should name the target time, not "1h 23m left". Fix the tooltip rendering in =wtimer= (dotfiles repo). Origin: roam inbox capture.
-
-Fixed 2026-07-01 (dotfiles): =_describe= now renders an alarm's wall-clock target via a new =format_clock= helper instead of =format_time(remaining)=. TDD test added; full wtimer suite (87) green.
-
-** DONE [#C] Waybar right-cluster module order :waybar:quick:
-CLOSED: [2026-07-01 Wed]
-Move the timer module to the rightmost position, just left of the systray, and move the battery/sysmonitor module to second-to-rightmost. Config edit in the waybar config (dotfiles hyprland tier). Origin: roam inbox capture.
-
-Done 2026-07-01 (dotfiles waybar config): =custom/timer= now sits just left of =tray= with =custom/sysmon= second-to-rightmost. waybar regenerated + reloaded live on velox; visual confirmation pending Craig.
-
 ** TODO [#B] Scrolling/Carousel layout: frame fit + wrap-around :hyprland:
 :PROPERTIES:
 :LAST_REVIEWED: 2026-06-13
@@ -203,26 +64,6 @@ Refiled from the archsetup task audit (2026-06-28), landed via ~/.dotfiles/inbox
 - Check dotfiles for uninstalled packages and remove orphaned configs.
 - Verify all stowed files are actually used.
 
-** DONE [#B] Pocketbook finish-or-cancel decision :pocketbook:
-CLOSED: [2026-07-02 Thu]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-06-24
-:END:
-Decided by Craig 2026-07-02, ahead of the scheduled checkpoint: remove pocketbook altogether. Executed same day — pip package uninstalled (user site clean), running instance killed, launcher gone, =pocketbook/= tree removed from the repo, Super+P rebound to toggle-touchpad (P for Pointers; Super+Shift+I unbound, waybar tooltip hint updated — dotfiles a750cb4). The org-capture popup remains the quick-notes surface.
-
-** DONE [#B] Provision Eask in archsetup :tooling:eask:
-CLOSED: [2026-07-02 Thu]
-Shipped 2026-07-02 (speedrun): npm global install block added after the nvm line — runs as $username with --prefix $HOME/.local, display/error_warn wrapped, output to $logfile, matching the claude-code block's shape. The npmrc decision went yes: dotfiles common/.npmrc pins prefix=${HOME}/.local (stowed; hand-linked live, npm config get prefix confirms ~/.local — dotfiles 01627cc). VM assertion added: ~/.local/bin/eask present + ~/.npmrc stowed. Live smoke: eask 0.12.9 on PATH. Full acceptance (fresh-install chime make setup/test) rides the next VM pass.
-:PROPERTIES:
-:LAST_REVIEWED: 2026-05-26
-:END:
-Add =@emacs-eask/cli= to archsetup's provisioning so fresh machines get it. Eask is installed by hand today and declared nowhere in archsetup or the dotfiles repo, yet both chime and linear-emacs depend on it (their =make setup/test/coverage= shell out to =eask=). Source: handoff from linear-emacs 2026-05-23.
-
-- Add a global npm install after the node block (=archsetup= ~2030, after =aur_install nvm=), modeled on the claude-code native-install block: run as =$username=, wrapped in =display=/=error_warn=, output to =$logfile=. Roughly =sudo -u "$username" bash -c 'npm install -g --prefix "$HOME/.local" @emacs-eask/cli'=.
-- Pin the prefix to =~/.local= so eask lands at =~/.local/bin/eask= (already on PATH) and the install runs as the user, not root. On the current machine =npm config get prefix= returns =/usr=, so eask was installed with an explicit =--prefix=.
-- Decision: also set a persistent user npm prefix (=~/.npmrc= with =prefix=${HOME}/.local=)? If yes, that =~/.npmrc= is a legitimate dotfile to stow; if no, rely on the explicit =--prefix= flag alone. =~/.eask/= is a regenerable cache — leave un-stowed.
-- Acceptance: fresh run leaves =eask= on PATH at =~/.local/bin/eask= (no root); =cd ~/code/chime && make setup && make test= works.
-
 ** TODO [#B] Network panel redesign — no terminals, verify-everything, full failure coverage :feature:waybar:network:
 Major evolution of the shipped =custom/net= module ([[file:docs/design/2026-06-29-waybar-network-module-spec.org]]).
 Reverses the spec's "privileged tiers run in a net-popup terminal" decision. Origin:
@@ -509,18 +350,6 @@ Initial spec written 2026-07-02: [[file:docs/design/2026-07-02-timer-panel-spec.
 
 From Craig's roam capture 2026-07-02: give the timer a GTK UI/UX like the network panel.
 
-** DONE [#C] Waybar timer dialog styling :waybar:
-CLOSED: [2026-07-02 Thu]
-From Craig's roam capture 2026-07-02: style the timer module dialogs like the screenshot dialog — tighter window, icons on the selections, colon+space after the prompt.
-
-Shipped 2026-07-02 (dotfiles 9ffcba7): dialogs size to content, type menu carries the kind glyphs, prompts end ": ". Three new tests; screenshot-verified live.
-
-** DONE [#B] Waybar collapse jumps client windows :bug:waybar:hyprland:
-CLOSED: [2026-07-02 Thu]
-From Craig's roam capture 2026-07-02: collapsing/expanding (and any waybar teardown) snapped every tiled window up and back down; hold the clients still and let only the bar change.
-
-Shipped 2026-07-02 (dotfiles 4b1a4ec): waybar now runs exclusive:false and the new waybar-reserve script statically reserves the bar strip per monitor (wired as exec so config reloads re-apply it). Verified live: client geometry held constant through bar kill, relaunch, and a collapse round-trip. Eight new tests (script + pairing).
-
 ** TODO [#B] Desktop-settings dropdown panel :waybar:
 :PROPERTIES:
 :LAST_REVIEWED: 2026-06-24
@@ -544,33 +373,6 @@ Design / open questions (propose before building):
 
 Implementation notes: a small GTK layer-shell app (mirror pocketbook's structure: src-layout Python package, pytest, Makefile) talking to brightnessctl / hyprctl / the touchpad + airplane helpers. Lives in the dotfiles repo or in-tree like pocketbook. TDD the backing toggle/slider logic. Sizable — worth a design doc first.
 
-** DONE [#B] Bluetooth panel + bar module :feature:waybar:bluetooth:
-CLOSED: [2026-07-02 Thu]
-:PROPERTIES:
-:SPEC_ID: 1271a845-4463-4831-9902-990eda6b2265
-:END:
-Spec: [[file:docs/design/2026-07-02-bluetooth-panel-spec.org]] (IMPLEMENTED 2026-07-02 — all five phases shipped same day: engine eb2230f, panel 76b2c05, bar module e372de3, bt-priv + blueman retirement 2a026b1/d8d8c53, install wiring proven by VM assertions). Residual: the phase 4-5 VM assertions run on the next VM pass; ratio picks up the package removal + hand-links on its trip list.
-
-A bluetooth panel driving a CLI underneath (bluetoothctl one-shot verbs), consistent in look and feel with the net panel (GTK4 + layer-shell + Blueprint, humble-object presenter, verify-everything). Minimalistic interface, full functionality, plus a diagnostics/troubleshooting section mirroring the net panel's Diagnostics tab. Bar module glyph opens it. Craig's ask (2026-07-02): follow UX/UI best practices; where the net panel's patterns conflict with best practices, file a net-panel bug task rather than clone the flaw.
-
-*** 2026-07-02 Thu @ 13:30:42 -0400 Shipped phase 1 — the bt engine package (dotfiles eb2230f)
-=bluetooth/src/bt/= mirrors the net engine's layout: btctl parsing boundary (show/devices/info, connect-error classifier), sysio rfkill/airplane, audio module over pw-dump/wpctl (HSP probe + A2DP switch repair with verify-after), redacted eventlog, six repair tiers, and the doctor chain (adapter → rfkill → service → powered → devices → audio profile) with safe auto-repairs behind =--fix= (never auto-connects; airplane blocks are named, not fought). 101 tests over fake binaries; 42 suites green (=make test= glob auto-discovered =tests/bt/= — gate check verified). Live read-only on velox: =bt status= + =bt doctor= read the real adapter/devices/audio graph; =~/.local/bin/bt= hand-linked (no restow under running Hyprland). Ground truth vs spec: profile inventory needs =pw-dump= (wpctl can't enumerate), and the card's =bluez5.profile= prop is unreliable — sink node's =api.bluez5.profile= is authoritative. Deferred INTO phase 2: the shared dupre css factoring (net's css is an inline string in =gui.py=, not an asset — factoring it without the bt-panel consumer just risks the working net panel).
-
-*** 2026-07-02 Thu @ 14:15:27 -0400 Shipped phase 2 — the GTK panel (dotfiles 76b2c05)
-Cloned the net panel's shape over the phase 1 engine: GTK-free PanelModel + viewmodel (69 new tests, display-free), Blueprint pages (Devices with the adapter power row + Paired/Nearby sub-views; Diagnostics with the doctor cascade, inline fix buttons mapped from step =repair= keys, Advanced tool selector), gui.py controller (layer-shell OVERLAY 380x520 TOP+RIGHT, Esc closes, single-instance toggle, bg worker, passkey + confirm dialogs), pairing pty state machine (=bt/pairing.py=: confirm-passkey yes/no with default-deny, display-passkey, bounded deadline, tested over the fake's interactive mode), manage.py op envelopes shared by CLI and panel (cli refactored onto it; power + discoverable verbs added), =bt panel= subcommand + =bt-panel= toggle wrapper (hand-linked into =~/.local/bin=, no restow under live Hyprland). Shared dupre css factored: net's inline =_CSS= → =hyprland/.config/themes/dupre/panel.css= with =dupre-*= classes, both panels load it (stowed path first, repo-relative fallback; hand-linked =~/.config/themes/dupre/panel.css=). Super+Shift+B rebound blueman-manager → bt-panel (hyprctl reloaded, live). 43 suites green (=make test= exit 0). DEFERRED pending Zoom ending: the AT-SPI smoke (=make test-panel-bt=, written and wired) and any visual check of either panel with the factored css — both need a visible window. Gotcha for posterity: the old =test_panel_stub_exits_two= CLI test ran =bt panel= for real once cmd_panel was wired and launched the panel on the live compositor mid-meeting for ~30s before the test timeout killed it; it now asserts parser wiring only — never run =bt panel= inside =make test=.
-
-*** 2026-07-02 Thu @ 15:06:00 -0400 Shipped phase 3 — the bar module + blueman retirement (dotfiles e372de3)
-=custom/bluetooth= over the engine: waybar-bt shim + =bt/indicator.py= (state-following glyph — slashed off/blocked/absent, plain dim idle, connected mark white; low-battery <15% adds a red pango percentage to the glyph; tooltip = connected devices with battery + Super+Shift+B hint). Signal 10; the panel pokes =pkill -RTMIN+10 -x waybar= after each status reload. Blueman retired from the Hyprland session: exec-once + both windowrules removed, applet killed live; waybar relaunched on the runtime config and the module verified on the bar (connected glyph, blueman tray icon gone). Theme drift guard caught that themes/*/waybar.css edits must mirror into the live =waybar/style.css= — all three updated. 43 suites green. ALSO closed this pass: the deferred phase 2 visual batch (bt AT-SPI smoke green after fixing its Connect/Disconnect state-following assertion — c1a8219; net smoke green on the factored css; both panels eyeballed correct in dupre). Left for phase 4: package removal (blueman out of archsetup), sxhkdrc's dwm blueman-manager binding decision rides that pass.
-
-*** 2026-07-02 Thu @ 15:16:51 -0400 Shipped phase 4 — bt-priv shim, blueman out, VM assertions
-Dotfiles =2a026b1=: the stowed =bt-priv= shim over the phase-2 =bt.priv= module (one verb, =restart-bluetooth=; verified end-to-end against the symlinked fake-systemctl — rc 0 with =BT_SUDO= empty, rc 2 on bad verb/usage; hand-linked into =~/.local/bin=), and the sxhkd =Super+Shift+B= bind repointed from the retired blueman-manager to =st -e bluetoothctl= (the decided terminal fallback — the GTK panel is Wayland-only, and the bt CLI is hyprland-tier so dwm never gets it). 43 dotfiles suites green. archsetup: blueman dropped from the =desktop_environment= bluetooth loop (bluez + bluez-utils stay, solaar untouched); VM assertions added to =test_packages.py= (bluez/bluez-utils installed, blueman NOT installed as the retirement regression guard — collected 15, exercised on the next VM run since VM tests run committed code); =bash -n= + =py_compile= + =make test-unit= green. SUDOERS: no new rule needed, same conclusion as net-priv (2026-07-01 entry) — archsetup:1089 grants the primary user blanket =NOPASSWD: ALL=, which covers =systemctl restart bluetooth=; a narrow bt-priv rule would be dead config under the blanket grant, so phase 5's "sudoers placed" item is satisfied by the existing grant. LIVE: blueman package removed from velox (=pacman -Rns=, decision "drop it outright, both machines"); ratio needs the same + the bt-priv hand-link on its trip list.
-
-*** 2026-07-02 Thu @ 15:19:58 -0400 Shipped phase 5 — install-default wiring proven by VM assertions
-No new install code was needed: the waybar =custom/bluetooth= module, the =Super+Shift+B= → =bt-panel= bind (hyprland.conf), and the shared =themes/dupre/panel.css= all live in the dotfiles hyprland tier, so the existing clone + =make stow hyprland= step lands them on a fresh install; sudoers is covered by the blanket grant (phase 4 conclusion). The phase's substance is the proof: =test_desktop.py= gained hyprland-gated assertions that the four bt bins (=bt=, =bt-panel=, =bt-priv=, =waybar-bt=) are stowed executable in =~/.local/bin= (either stow shape — per-file symlink or folded dir), the waybar config carries =custom/bluetooth=, hyprland.conf carries the =bt-panel= bind, and the stowed theme has =panel.css=. Collected 30 in =test_desktop.py=; exercised on the next VM run (VM tests run committed code).
-
-*** 2026-07-02 Thu @ 15:19:58 -0400 Test surface complete across all phases
-Everything the surface named exists and is green: engine suites over fake binaries (phase 1, 101 tests — btctl parse, doctor chain, A2DP repair verify), PanelModel presenter suite (phase 2, 69 tests), pairing state-machine suite (passkey confirm / NoInputNoOutput / timeout, over the fake's interactive mode), bar-module suite (phase 3), gated AT-SPI smoke (=make test-panel-bt=, run green live), and the phase 4-5 VM assertions (=test_packages.py= bluetooth stack + blueman-absent guard; =test_desktop.py= panel wiring). 43 dotfiles suites green; VM assertions await the next VM run.
-
 ** TODO [#B] Local offline LLM runtime + per-host model cache :tooling:llm:
 :PROPERTIES:
 :LAST_REVIEWED: 2026-05-29
@@ -831,20 +633,6 @@ Some entries are libraries likely pulled in as dependencies (blas-openblas, open
 - [ ] webkit2gtk
 - [ ] whisper.cpp
 
-** DONE [#B] All error messages should be actionable with recovery steps
-CLOSED: [2026-07-02 Thu]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-06-24
-:END:
-Shipped 2026-07-02 (speedrun). Structural fix at the helper: =error_fatal= now takes an optional third recovery-hint arg and every fatal prints the last five log lines inline, the full log path, the per-site "Fix:" when given, and the resume pointer (step markers mean a re-run continues where it stopped) — so even a hint-less fatal is actionable. All 17 fatal call sites got specific hints (keyring reinit, mirrorlist switch, userdel/USERNAME conflict, base-devel for makepkg, DESKTOP_ENV values, dotfiles-dir cleanup, tmpfs sizing, aur.archlinux.org reachability). The end-of-run Error Summary now closes with the grep-the-log line and the fix-and-re-run pointer. =error_warn= already carried what-failed + exit code into the summary; unchanged.
-
-** DONE [#B] Improve logging consistency
-CLOSED: [2026-07-02 Thu]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-06-24
-:END:
-Shipped 2026-07-02 (speedrun), paired with the actionable-errors task. Audit result: the install helpers (pacman_install/aur_install/retry_install/run_task/git_install/pipx) and error helpers already tee/append everything to $logfile — the gaps were direct mutations whose stderr went to the console and vanished. Swept every =sed -i= and file-write mutation lacking capture (locale.gen uncomment, pacman.conf ParallelDownloads/Color + multilib, waybar battery removal x3, wireless-regdom, geoclue BeaconDB, paccache, BRIO udev rule, fstab fmask, mkinitcpio HOOKS, sudoers append, ufw status read): each now sends stderr to $logfile, and the previously-silent ones (locale.gen, pacman.conf, multilib, waybar, regdom, geoclue, paccache, udev) gained =error_warn= handlers so failures land in the summary instead of passing silently. Verified: bash -n clean, 10 unit suites green, shellcheck warning-diff vs HEAD empty (no new findings).
-
 ** TODO [#B] Security hardening + audit :security:
 :PROPERTIES:
 :LAST_REVIEWED: 2026-06-28
@@ -885,86 +673,12 @@ Practical guidelines for working in public spaces
 :END:
 Ensure new tools integrate with the Hyprland environment and don't break workflow (the fleet is all Hyprland now; archsetup still supports DWM/X11 but no current machine uses it)
 
-** DONE [#B] Add NVIDIA preflight check for Hyprland
-CLOSED: [2026-07-02 Thu]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-05-21
-:END:
-Shipped 2026-07-02 (speedrun), TDD. =nvidia_preflight_report= is a pure sed-extractable core (same harness pattern as zig-pin): modalias scan for vendor 10DE — DRM first, PCI display-class (bc03) fallback so an NVIDIA audio function can't false-trigger — then the repo's =nvidia-utils= candidate major checked against 535. Prints the Wayland guidance + env vars (LIBVA_DRIVER_NAME, GBM_BACKEND, __GLX_VENDOR_LIBRARY_NAME, ELECTRON_OZONE_PLATFORM_HINT) and the pre-Turing/AUR-legacy note. preflight_checks aborts on <535/unknown (rc 11), prompts continue/abort on a healthy NVIDIA box (rc 10), silent on non-NVIDIA (rc 0). 9 Normal/Boundary/Error tests over fake modalias trees + a fake pacman (=tests/nvidia-preflight/=, glob-discovered by test-unit — 10 suites green).
-
-** DONE [#C] Wlogout exit-menu buttons are rectangular, not square
-CLOSED: [2026-07-02 Thu]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-06-24
-:END:
-The wlogout exit menu renders its buttons taller than they are wide on velox, so the cells read as vertical rectangles instead of squares. They render square (centered) correctly on ratio, so this is a per-host / resolution difference, not a flat bug. Fix the button sizing in the wlogout style (=~/.dotfiles/hyprland/.config/wlogout/style.css=) so each cell is square on both hosts. Noticed 2026-05-21. Related: the [#D] VERIFY about wlogout sizing across displays.
-
-The wlogout config uses fixed pixel margins, which is the likely reason sizing differs across the two displays — adjusting them for the laptop screen is part of the fix (folded in from the former "Test wlogout menu on laptop" VERIFY, 2026-06-24).
-
-Add a regression test so the square-cell fix doesn't silently break on a resolution change: assert the rendered (or computed) wlogout button cells are square across ratio's and velox's resolutions. Dropped :quick: — the cross-host test pushes this past a spare-moment fix.
-
-Shipped 2026-07-02 (dotfiles 775771b). Keybind now calls a =wlogout-menu= wrapper computing centered margins from the focused monitor (the old fixed L/R 1200 exceeded velox's 1436 logical width). Also fixed two styling defects the geometry hid: invisible unfocused borders (now muted, so the square edge is visible) and hover/focus sharing one gold rule (lock button glowed at launch; focus is now a muted ring). Tests: unit margin-math suite across both hosts' resolutions + portrait + small + bad-geometry, CSS regression suite, and a compositor-gated =make test-wlogout= smoke that launches a no-op probe, screenshots, and measures squareness (velox: 361x361 px, PASS). Ratio's visual eyeball rides the pending ratio sync.
-** DONE [#C] Net panel: error toasts auto-dismiss unread :bug:network:waybar:
-CLOSED: [2026-07-02 Thu]
-Fixed in dotfiles 0f017d4: viewmodel.toast_plan owns the toast policy — errors show sticky and ignore the post-op refresh's empty clear (worst case: a forget failure's error was wiped within ~2s by its own refresh), and the next real status replaces them. Successes keep the 4s fade. 7 policy tests added; 41 suites green.
-
-** DONE [#C] Net panel: verify claimed keyboard navigation :test:network:waybar:
-CLOSED: [2026-07-02 Thu]
-Found during the bluetooth-panel UX pass (2026-07-02). The V2 spec claims tab-between-sections, arrow-key row navigation, and type-to-filter, but no custom keyboard code exists in the panel — arrows and type-ahead may ride GTK ListBox defaults, tab-between-sections likely doesn't. Verify each claim against the live panel (AT-SPI smoke can assert focus order); implement or strike the claims from the spec so spec and panel agree.
-
-*** 2026-07-02 Thu @ 13:05:00 -0400 Code-level pass done; live probe deferred (Craig in a Zoom meeting)
-Code reality (dotfiles net/src/net): Esc close is wired (gui EventControllerKey); row-activated -> primary is wired for both connection lists (pages.py:122,126), so Enter-on-row rides the GTK ListBox activate binding; arrows ride ListBox defaults; NOTHING implements type-to-filter (no search/filter code exists — that claim is false as written); Tab is the plain GTK focus chain, widget by widget, not section jumps. Live AT-SPI probe plan: launch panel in test mode, drive keys via hyprctl dispatch sendshortcut targeted AT THE PANEL WINDOW (never the focused surface — wtype/ydotool absent anyway), gate every key on the panel holding focus, never send Enter on the available list (real connect risk). BLOCKED at 13:05: active window is zoom (meeting) — no test windows, no synthetic input until clear. Then: verify focus order + arrows + no-filter live, strike/reword the spec's keyboard bullet to match.
-
-*** 2026-07-02 Thu @ 14:57:18 -0400 Ran the live probe; spec bullet reworded to match reality
-Zoom ended ~14:45; probe ran per plan (panel in test mode, hyprctl dispatch sendshortcut targeted at the panel address, every key gated on panel focus, Enter never sent). Verdicts: arrows move row focus and Enter rides the ListBox activate binding (TRUE — kept); Esc closes (TRUE — kept); Tab is the plain GTK widget-by-widget chain and inside a list crawls row by row, no section jumps (claim FALSE — struck); type-to-filter does not exist (claim FALSE — struck; typing into the 24-row Saved list filtered nothing). Spec's Keyboard bullet reworded with the live evidence and a note that section-jump Tab or filtering would be new work. Probe gotchas for reuse: AT-SPI list items have empty accessible names, so row identity needs get_index_in_parent(); a killed test panel can leave a windowless single-instance process that eats the next launch via D-Bus activation — pkill -9 -f 'net panel$' and wait before relaunching.
-
 ** TODO [#C] Window focus lost when unhiding stashed windows :bug:hyprland:
 :PROPERTIES:
 :LAST_REVIEWED: 2026-06-24
 :END:
 From the roam inbox: hiding a window (e.g. the org-capture popup) then unhiding it should leave the unhidden window focused, but another window typically takes focus. Also =ctrl+j/k= (layout-navigate) can't reach the unhidden window afterward — it should always reach any visible window except the waybar. Involves stash-restore + layout-navigate; needs interactive reproduction with Craig.
 
-** CANCELLED [#C] Pocketbook development backlog :pocketbook:
-CLOSED: [2026-07-02 Thu]
-Cancelled with the 2026-07-02 remove-pocketbook decision — the app and its in-tree package are gone.
-:PROPERTIES:
-:LAST_REVIEWED: 2026-05-26
-:END:
-Pocketbook (GTK4 layer-shell notes panel, toggled via waybar) was pulled from publication 2026-05-26 — github repo + cjennings.net repo deleted, mirror hook removed — and folded into this repo at =pocketbook/= until it's ready to spin back out. Src-layout Python package with pytest tests and a Makefile. Develop it in-tree; the backing modules are =store/note/panel/layer_shell/app/note_widget= + =style.css=.
-
-Backlog (unordered; promote items to their own dated tasks as they're picked up):
-
-- Configurable options, possibly a dedicated configuration panel.
-- Lose-focus hides pocketbook — configurable on/off.
-- Configurable display order: chronological by creation date (asc/desc), manual, alphabetical (asc/desc).
-- Search / filter notes.
-- Global toggle keybind (Hyprland =bind=) alongside the waybar click; document the waybar integration.
-- Note CRUD polish (create/edit/delete) + optional markdown rendering.
-- Pin / favorite notes.
-- Tags or notebooks / categories.
-- Persistence: confirm store format + =~/.local/share/pocketbook/= location, add versioning/migration, decide a backup/sync story.
-- Theming: track the dupre/hudson theme system so =style.css= follows =set-theme=.
-- Layer-shell geometry config (anchor edge, width, margins) + HiDPI / multi-monitor behavior — ties into [[file:docs/PLAN-per-host-overrides.org][per-host overrides]] scaling work.
-- Config file format (toml) + reload-without-restart.
-- Expand test coverage (TDD per testing standards; =tests/= already exists).
-- Release prep for the eventual spin-back-out: pyproject metadata, version, license.
-- Re-wire the archsetup install (gtk4-layer-shell dep + install step + post-install clone) when pocketbook ships. Removed 2026-05-26 — see git history of =archsetup= / =scripts/post-install.sh=.
-
-** CANCELLED [#C] Fn+F9 toggles pocketbook — source unlocated :hyprland:pocketbook:
-CLOSED: [2026-07-02 Thu]
-Retired with pocketbook itself (2026-07-02 removal) per this task's own exit condition — with the app uninstalled and unbound, whatever Fn+F9 emitted has nothing to toggle.
-:PROPERTIES:
-:LAST_REVIEWED: 2026-06-23
-:END:
-On velox, pressing Fn+F9 (physical function key) toggles the pocketbook panel. It shouldn't. Raised from a home-project session 2026-06-23.
-
-Investigated 2026-06-23 and could not locate the trigger in any config. Ruled out, three ways:
-- No F9 bind (bare / $mod / keycode) in the live =hyprland.conf= (now a stow symlink), the velox host tier =conf.d/local.conf=, or the waybar config.
-- =hyprctl binds= runtime (all 90 active binds, authoritative) execs pocketbook on ONLY =SUPER+P=. No F9/XF86 path reaches it. The old touchpad toggle that used to sit on =$mod+F9= was moved to =$mod+M=, so F9 is unbound in Hyprland.
-- No input remapper (keyd/xremap/input-remapper) and no hotkey daemon (sxhkd/swhkd) running or configured; pocketbook's own source has no F9 / GlobalShortcuts / portal / dbus listener (its GTK ShortcutController binds only Esc/Ctrl-n/Ctrl-j/Ctrl-k/Del/Return). pocketbook is a single-instance Gtk.Application, so any path that re-runs =pocketbook= toggles it.
-
-Parked at Craig's call (not worth deeper investigation now). If it resurfaces, the one unfinished step is to capture what keysym Fn+F9 actually emits (=wev -f wl_keyboard:key=, press Fn+F9, read the =sym:= / =code:=) and grep for that. Most likely folds into removing pocketbook from the waybar setup — if pocketbook leaves the bar, retire this with it.
-
 ** TODO [#C] Ensure sleep/suspend works on laptops
 :PROPERTIES:
 :LAST_REVIEWED: 2026-06-09
@@ -1009,24 +723,6 @@ The goal is a single place to edit each config, not two.
 :END:
 Once-yearly systematic inventory of known deficiencies and friction points in current toolset
 
-** CANCELLED [#C] Waybar emacs-service status + control :feature:waybar:
-CLOSED: [2026-07-02 Thu]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-06-24
-:END:
-From the roam inbox (2026-06-22): with Emacs integrated into the system as file manager and instant note-taker, make bouncing it trivial. A waybar component showing the emacs service status, with detail on hover, that turns the server on / off / bounce via right-click. Pairs with running the Emacs daemon as a managed systemd user service.
-
-Cancelled 2026-07-02 per Craig during the task-batch pick: no current need. Re-add or pull back from Resolved if a need surfaces.
-
-** DONE [#C] set-wallpaper detaches waypaper config from its stow symlink :bug:hyprland:quick:solo:
-CLOSED: [2026-07-02 Thu]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-06-28
-:END:
-=set-wallpaper= persists with =mv "$tmp" "$CONFIG"=, which replaces the =~/.config/waypaper/config.ini= stow symlink with a real file. After the first run the live config is detached from =~/.dotfiles/hyprland/.config/waypaper/config.ini=, so a later =git pull= + restow won't update it and set-wallpaper changes never flow back to the repo. Fix: write in place rather than =mv= over the symlink — e.g. =cp "$tmp" "$CONFIG"= (follows the symlink to the real dotfiles file), or resolve the link target and write there. Lives in =~/.dotfiles/hyprland/.local/bin/set-wallpaper=; it has a test suite, so add a Boundary case for "CONFIG is a symlink".
-
-Shipped 2026-07-02 (dotfiles d826be4): write-back now redirects through the symlink instead of mv-ing over it; two boundary tests pin the invariant (replace + append paths). velox's live config was still a healthy symlink, so no repair needed.
-
 ** TODO [#D] Consider Customizing Hyprland Animations
 Current: windows pop in, scratchpads slide from bottom.
 
@@ -1317,110 +1013,6 @@ Verified: the only =eval= left in =archsetup= is line 578 in =retry_install=, an
 
 * Archsetup Resolved
 
-** DONE [#B] Idle-inhibitor keybind + synced waybar indicator :hyprland:waybar:
-CLOSED: [2026-06-23 Tue]
-Shipped 2026-06-23 as dotfiles commit =a004201=. Super+I toggles the hypridle daemon (kill = inhibit, relaunch = restore). The built-in waybar =idle_inhibitor= module was replaced with a =custom/idle= module backed by a =waybar-idle= script, so the keybind, the bar click, and the icon share one source of truth (whether hypridle is running) and stay in sync. Icons 󰒳 inhibited / 󰒲 active, with a 5s poll safety net. Freed =Super+I= by pruning the unused ai-term pyprland scratchpad from both host configs. TDD'd (=waybar-idle= + =hypridle-toggle= suites); dupre/hudson theme CSS updated. From a home-project handoff 2026-06-23; Craig confirmed it works live.
-** DONE [#B] Verify package signature verification not bypassed by --noconfirm
-CLOSED: [2026-06-23 Tue]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-05-21
-:END:
-Audited 2026-06-23. =--noconfirm= does not bypass signature verification — it only auto-answers interactive prompts. Signature checking is governed by =SigLevel= in =/etc/pacman.conf=, which archsetup leaves at the Arch default (=Required DatabaseOptional=): its only pacman.conf edits are ParallelDownloads, Color, and enabling multilib (=archsetup:913,917=), none of which touch =SigLevel=. So every repo package stays signature-verified regardless of =--noconfirm=.
-
-One real integrity bypass exists, and it is not =--noconfirm=: =archsetup:2403= runs =yay -S --noconfirm --mflags --skipinteg python-lyricsgenius=, where =--skipinteg= skips makepkg's checksum and PGP-signature checks for that one AUR package (a documented workaround for an expired-signature issue upstream). It's scoped to a single package, not global. Tracked for periodic re-check below.
-** DONE [#C] Harden sshd in the installer (explicit prohibit-password) :solo:
-CLOSED: [2026-06-24 Wed]
-Done 2026-06-24: the openssh block (=archsetup:1271-1277=) now writes =/etc/ssh/sshd_config.d/10-hardening.conf= with =PermitRootLogin prohibit-password= and reloads sshd, right after starting the service. =PasswordAuthentication= left untouched so ssh-copy-id to the user still works. Makes the posture intentional rather than dependent on the upstream default. Velox and ratio (which carried an explicit =PermitRootLogin yes= at =sshd_config:33= from earlier provisioning) were already fixed by hand 2026-06-23. Verified =bash -n= + =shellcheck -S error= clean; full drop-in-on-fresh-install confirmation is VM-deferred (the unit harness covers helpers, not inline install steps).
-** DONE [#C] Build security dashboard command :solo:
-CLOSED: [2026-06-23 Tue]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-05-21
-:END:
-Shipped 2026-06-23 as dotfiles commit =1b9b205=: =security-status= (=common/.local/bin=, on PATH). Read-only dashboard showing disk encryption (LUKS *and* ZFS native — the fleet runs ZFS, so a LUKS-only check would have falsely reported "no encryption"), ufw state, externally-reachable ports (counts all listening, lists only the non-loopback exposures), and running/failed service counts. Command lookups are env-overridable; parsing covered by unit tests against canned output. New file, so ratio needs =git pull && make stow hyprland= to link it.
-** DONE [#C] Teach archsetup to stow the host tier :solo:
-CLOSED: [2026-06-23 Tue]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-06-11
-:END:
-Already implemented in =user_customizations()= (=archsetup:1049-1058=): after stowing =common= + the DE package, it derives =host_tier="$(cat /etc/hostname 2>/dev/null || uname -n)"= and stows that package when =$dotfiles_dir/$host_tier= exists, else prints "no host tier for '<host>' — skipping". The =/etc/hostname=-first detection is the right call for install time (=uname -n= still reports the ISO's name until reboot), and it's the same skip-if-absent semantics as the dotfiles Makefile. Verified by reading the installer 2026-06-23; no code change needed.
-** DONE [#C] Waybar indicators unevenly spaced :quick:solo:waybar:
-CLOSED: [2026-06-24 Wed]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-06-24
-:END:
-The right-side module icons don't sit at even intervals — spacing reads as inconsistent across the group. Noticed 2026-05-21 after adding the airplane indicator.
-
-Done 2026-06-24: a screenshot showed the standalone module icons were already even — the unevenness was the tray, whose icons clustered tight (tray =spacing: 4= vs the ~0.3rem margins on every other module). Bumped tray =spacing= 4 → 10 in the waybar =config=; restarting waybar and re-screenshotting confirmed the row reads even. The lever was the tray spacing, not the per-module CSS the original body guessed at.
-** DONE [#B] Separate mpd playlist_directory from music_directory :mpd:music:quick:
-CLOSED: [2026-06-24 Wed]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-06-24
-:END:
-Done 2026-06-24 (dotfiles a9bfdf3): set =playlist_directory= to =~/.local/share/mpd/playlists= (separate from =music_directory= ~/music). git-moved the 73 radio-stream playlists from =common/music/= into =common/.local/share/mpd/playlists/= (history preserved); dropped the empty =60s Sounds.m3u= (Craig's call); git rm'd the stray =Black Flamingos - Space Bar.m4a= and moved the real track into the music library. Curated playlists left flat in ~/music (Craig's call — avoids rewriting the 7 relative-path ones). The ~/music/radio orphan was already gone. Relinked surgically (a pre-existing =whereami= stow conflict blocked a full =stow common=). mpd restarted clean: 73 radio playlists load from playlist_directory (verified SomaFM stream URLs), 24 curated browsable from the music tree. ratio needs the same restow + mpd restart on its next pull (reminder filed). Decisions answered: 60s dropped, curated flat.
-Spec written and approved (option 1), pinned before execution on 2026-06-03. Root issue: mpd.conf has =playlist_directory= == =music_directory= == ~/music, so the whole audio library is the playlist store and radio streams mix with curated playlists. Option 1: radio stream playlists (portable, 73 in the dotfiles repo) move to a dedicated =playlist_directory= (=~/.local/share/mpd/playlists=) via stow; the 22 curated local playlists (machine-specific track refs) live in the music tree. Also removes the broken ~/music/radio/ orphan (73 dead symlinks).
-
-Full step-by-step spec (mpd.conf edit, repo restructure of =common/music/= → =common/.local/share/mpd/playlists/=, curated relocation, restow, verification incl. the 7 relative-path curated playlists, ratio propagation) is in the 2026-06-03 session record under .ai/sessions/. Two open decisions before executing: (1) drop the empty =60s Sounds.m3u= or refill with the SomaFM 60s URL; (2) curated playlists into =~/music/playlists/= subdir vs leave flat in ~/music/. Side cleanup surfaced: a stray audio file =Black Flamingos - Space Bar.m4a= is wrongly committed in the dotfiles repo's =common/music/= — git rm it and move to the synced library.
-** DONE [#C] Install adopted modern CLI tools :tooling:solo:
-CLOSED: [2026-06-24 Wed]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-06-24
-:END:
-Done 2026-06-24: added bat/dust/hyperfine/doggo to archsetup General Utilities (tealdeer was already declared), installed all five on velox, set =BAT_THEME=ansi= in =common/.profile.d/tools.sh= (tracks the dupre terminal palette), seeded the tldr cache. ratio still needs the =pacman -S= (additive; lands on its next archsetup run).
-Decision (Craig, 2026-06-24): adopt all five recommended tools — =bat=, =dust=, =hyperfine=, =tealdeer=, =doggo= (all in extra). Add them to archsetup's package list and install on both machines. Optional candidates (=xh=/=jless=/=sd=/=ouch=) declined for now. Full evaluation: [[file:docs/2026-06-10-modern-cli-tools-evaluation.org][docs/2026-06-10-modern-cli-tools-evaluation.org]].
-
-- Add the five to the appropriate pacman package section in =archsetup=.
-- =pacman -S bat dust hyperfine tealdeer doggo= on velox + ratio.
-- =bat=: set =BAT_THEME= to match the dupre palette once installed.
-- =tealdeer=: run =tldr --update= to seed the cache after install.
-** DONE [#C] Review file manager options for Wayland
-CLOSED: [2026-06-24 Wed]
-Decision (Craig, 2026-06-24): keep nautilus only; skip yazi. File management lives in Emacs dired plus the Super+F dirvish popup, so a TUI file manager has no daily user here. ranger was already ruled out (frozen upstream). Full evaluation: [[file:docs/2026-06-10-file-manager-evaluation.org][docs/2026-06-10-file-manager-evaluation.org]]. Follow-on surfaced: nautilus needs dark theming (filed as its own task).
-** DONE [#B] Theme nautilus to a dark theme :bug:solo:
-CLOSED: [2026-06-24 Wed]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-06-24
-:END:
-nautilus rendered blindingly white (Craig, 2026-06-24). As a GTK4/libadwaita app it follows the appearance portal's =org.freedesktop.appearance color-scheme=, which mirrors =org.gnome.desktop.interface color-scheme=. Two stacked causes:
-
-1. velox had no system-wide dconf db at all — no =/etc/dconf/profile/user=, no =/etc/dconf/db/site.d/00-archsetup-defaults=, no compiled =site= db — so archsetup's declared default (=color-scheme='prefer-dark'=, =archsetup:1109-1119=) never reached the machine (velox predates that block). Created the profile + site defaults as archsetup writes them and ran =dconf update=. =gsettings get= then returned =prefer-dark=.
-
-2. That alone did NOT fix the running session: a system-db default emits no GSettings change signal, so the appearance portal kept reporting =0= (no-preference → light), and libadwaita reads the portal, not =GTK_THEME=. (An early screenshot looked dark only because the shell env carries =GTK_THEME=Adwaita:dark=, which Hyprland-launched apps don't inherit — masking the real state.) Fix: a user-level =gsettings set org.gnome.desktop.interface color-scheme prefer-dark=, which signals the portal live. It now reports =1=, and a portal-driven nautilus (GTK_THEME unset) renders dark — screenshot-verified.
-
-Durable: the user value persists in =~/.config/dconf/user=; archsetup's system-db handles fresh installs (the portal reads the default fresh at login, so no signal is needed there). No archsetup change. ratio may need the same one-two — see the Active Reminder.
-** CANCELLED [#D] Test wlogout menu on laptop
-CLOSED: [2026-06-24 Wed]
-Merged into the "Wlogout exit-menu buttons are rectangular, not square" task ([#C]) — same effort (per-host wlogout button sizing across velox/ratio). The fixed-pixel-margins hint was folded into that task's body.
-** DONE [#B] Enlarge org-capture popup to scratchpad size :hyprland:
-CLOSED: [2026-06-24 Wed]
-From a .emacs.d inbox handoff (2026-06-15, captured via roam): the quick-capture / org-protocol popup is too small to be effective — it should be about the size of a terminal scratchpad.
-
-*** 2026-06-24 Wed @ 17:21:11 -0400 Sized the popup to the scratchpad, per-host in pixels
-The 06-15 read was wrong: the real size lever is the Hyprland window rule, not the quick-capture char-cell count. The =size 900 500= rule on the org-capture window pinned it to 900x500 regardless of the frame's requested geometry (demoing 120x24 vs 180x32 looked identical because both clamped to 900x500). Tried a percentage rule (=size 75% 70%=) to auto-adapt per host like the pyprland scratchpad — native window rules do NOT honor percentages (only pyprland does), so the frame fell back to char-cell geometry and overflowed the screen. Fix: absolute pixels matching each host's terminal scratchpad, placed in the host tier (=<host>/conf.d/local.conf=) since pixels don't adapt across monitors. velox = 1078x671 (75%x70% of its 1437x958 logical desktop) — verified on-screen. ratio = 1892x936 (55%x65% of 3440x1440) — set but not yet eyeballed on ratio (tracked as an Active Reminder in notes.org). The shared hyprland.conf keeps float/center/stay_focused and a comment pointing at the per-host size. dotfiles change — needs commit in =~/.dotfiles=.
-
-*** 2026-06-15 Mon @ 19:19:55 -0500 AI Response: popup size is the frame's char-cell count, not the Hyprland rule
-Triaged under auto inbox-zero. The popup is the emacsclient frame named "org-capture", created by =~/.dotfiles/hyprland/.local/bin/quick-capture= with =(width . 90) (height . 22)= — 90 columns by 22 lines. Emacs sizes by character cells and overrides the Hyprland rule =windowrule = match:title ^(org-capture)$, size 900 500= (hyprland.conf:182). The live frame measured ~889x860 px; the width tracks the 90-column count, not the window rule. Setting the Hyprland rule to =size 55% 65%= (the scratchpad's pyprland spec) did not change the frame width, so I reverted it — dotfiles left clean.
-
-Real lever: the column/line count in the quick-capture script. Scratchpad reference on ratio (DP-4, 3440x1440) is 55% 65% ~= 1892x936 px ~= 190 cols by 24 lines. Why this isn't a solo auto-fix — it needs a tradeoff decision:
-- The script lives in the shared =hyprland/= stow tier, so a fixed ~190 columns overflows velox's 1920-wide laptop, and 24+ lines overflows velox's 1080 height (22 lines ~= 860 px is already near the safe max there).
-- Emacs char-cell sizing doesn't adapt to the monitor the way pyprland's percentage does, so "scratchpad-size on both machines" needs one of: a fixed compromise count, a per-host override via the ratio/velox tiers, or a script that computes columns from the active monitor.
-Options to weigh: (a) a safe-on-both compromise like width 120-130 / height 24; (b) per-host width through the ratio/velox tiers; (c) dynamic sizing in quick-capture from =hyprctl monitors=. Pick the tradeoff and I'll implement.
-** DONE [#C] Highlight current month and year in the calendar hover :feature:waybar:quick:solo:
-CLOSED: [2026-06-24 Wed]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-06-24
-:END:
-From the roam inbox (2026-06-24): the waybar clock's calendar tooltip highlights today's date in goldenrod; the current month and year header should be goldenrod too.
-
-Done 2026-06-24: the date module is the custom =waybar-date= script (not the built-in clock), so the highlight lives in its tooltip markup. Added a sed wrapping line 1 of the current-month =cal= output (the centered "Month Year") in the same =#daa520= goldenrod the day highlight uses. Verified the tooltip JSON carries =<span color='#daa520'><b>June 2026</b></span>= with today's highlight intact and waybar live; the on-hover look is Craig's spot-check.
-** DONE [#C] Wallpaper-set from dirvish doesn't work on Wayland :hyprland:
-CLOSED: [2026-06-24 Wed]
-From the roam inbox (2026-06-24, claimed for archsetup by Craig): typing =bg= in the dirvish popup doesn't change the wallpaper — Craig's read is it may still be wired to feh/X11 instead of a Wayland utility.
-
-Findings (2026-06-24): the Wayland wallpaper utility on this setup is =awww= (waypaper's configured =backend = awww=; =set-theme= sets the default via =awww img <file>=). There was no shared wallpaper script (=bg= on PATH is just the shell builtin), and the dirvish =bg= command lives in the Emacs config, so it was calling the wrong (or no Wayland) setter.
-
-Done 2026-06-24 (dotfiles 8be2484): added =set-wallpaper <image>= to the hyprland tier — sets live via =awww img= and persists the choice into =waypaper/config.ini=, the single Wayland-correct entry point. Resolves relative paths, validates the file, exits non-zero without persisting if awww fails. 8 Normal/Boundary/Error tests green; live-verified (awww set it, config rewrote). Notified =.emacs.d= to point the dirvish =bg= command at =set-wallpaper <file>= — that wiring is its piece (dependency cleared, =:blocker:= dropped).
-
-Follow-up (separate, small): the login restore =exec-once= in =hyprland.conf= is hardcoded to =trondheim-norway.jpg=, so a wallpaper set via =set-wallpaper= shows live but won't survive a relogin until the exec-once becomes =waypaper --restore= (which reads the now-persisted config). Filed below.
 ** DONE [#B] Add backup before system file modifications :solo:
 CLOSED: [2026-06-25 Thu]
 :PROPERTIES:
@@ -1493,22 +1085,6 @@ A design doc (not yet written) should cover:
 - Tiered testing strategy (smoke/integration/end-to-end)
 - How to run tests and integrate with run-test.sh
 - Comparison with alternatives (Goss)
-** DONE [#C] Proton Mail Bridge font size :chore:quick:
-CLOSED: [2026-06-24 Wed]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-06-24
-:END:
-From the roam inbox (2026-06-22): adjust the Proton Mail Bridge UI font to a comfortable size. The bridge is a Qt app, so it likely keys off Qt scaling or the qt5ct/qt6ct config like the other Qt apps (QT_SCALE_FACTOR or a font setting).
-
-Done 2026-06-24 (dotfiles =hyprland.conf:47=): the bridge is a Qt6 *QML* app, so it ignores the qt6ct General font — bumped the UI font via =QT_FONT_DPI= on the autostart instead. Changed the exec-once to =env QT_FONT_DPI=108 protonmail-bridge --no-window= (default DPI is 96; 108 = 1.125x). Iterated live with Craig: 120 too big, 108 comfortable. hyprland.conf is a stow symlink so the change is already live; applies at every login. The =~/.config/autostart/Proton Mail Bridge.desktop= entry is dormant under Hyprland (no XDG-autostart), so it was left as-is.
-** DONE [#C] Wallpaper login-restore is hardcoded, not waypaper --restore :hyprland:quick:solo:
-CLOSED: [2026-06-24 Wed]
-:PROPERTIES:
-:LAST_REVIEWED: 2026-06-24
-:END:
-The Hyprland =exec-once= (=hyprland.conf:26=) restores the wallpaper with a hardcoded =awww img ~/pictures/wallpaper/trondheim-norway.jpg=, so any wallpaper set later (via =set-wallpaper=, waypaper, or the dirvish =bg=) reverts on relogin. =set-wallpaper= now persists the choice to =waypaper/config.ini=, so switch the exec-once to =waypaper --restore= (after =awww-daemon= is up) to make set wallpapers survive a relogin. Small, dotfiles-only; verify by setting a different wallpaper, relogging, and confirming it sticks.
-
-Done 2026-06-24 (dotfiles): swapped the line-26 exec-once from the hardcoded =awww img …/trondheim-norway.jpg= to =awww-daemon & sleep 1 && waypaper --restore=. waypaper has a real =awww= backend (in its =--backend= list), the stowed =waypaper/config.ini= carries =backend = awww= plus a default =wallpaper == line, so =--restore= works on a fresh install too. Mechanism verified live: =waypaper --restore= reapplied the persisted wallpaper via awww, exit 0. Relogin confirmation filed under "Manual testing and validation". Follow-up filed: =set-wallpaper='s =mv= detached the live =waypaper/config.ini= from its stow symlink, so set-wallpaper changes no longer flow back to dotfiles.
 ** DONE [#B] VM test harness shared one NVRAM file across filesystem profiles :bug:test:
 CLOSED: [2026-06-27 Sat]
 The harness shared one OVMF NVRAM file (=vm-images/OVMF_VARS.fd=) across the btrfs
@@ -1663,3 +1239,274 @@ this is fixed. zfs-profile testing works (=make test FS_PROFILE=zfs=).
 Companion hardening (defense-in-depth, archangel-side): install the bootloader
 with a removable =\EFI\BOOT\BOOTX64.EFI= fallback so a base boots even from
 fresh/empty NVRAM, and real installs survive firmware that drops boot entries.
+** DONE [#B] Network panel UI — review findings :feature:waybar:network:
+CLOSED: [2026-07-01 Wed]
+Full UI review 2026-07-01 (visual walk of every state + code pass over the view logic), 30 findings: 21 from the agent review + color audit, 9 from Craig. All fixed the same night in a no-approvals speedrun, five commits, each test-gated (33 suites) and the whole panel re-verified via AT-SPI smoke + screenshots.
+
+*** 2026-07-02 Wed @ 00:00 -0400 Theme pass: contrast, hierarchy, focus, hover, destructive, dialogs (dotfiles 82aad0b, 998829b)
+Selected-row captions turned cream (the dim gray measured 2.2:1 on the slate fill, a WCAG fail on the auto-selected active row). Names gained weight and captions dropped a size (type hierarchy beyond color). Focus rings gold, scrollbars slim slate, rows got a hover wash. Disconnect and Forget became terracotta destructive-action (red = off). The Join/Add dialogs carry the dupre contract now (ground, mono, styled entries with a gold focus border, no stock blue anywhere), the Add dialog disables its action button until the SSID is non-empty, its password field gained the peek icon, and long SSIDs ellipsize.
+
+*** 2026-07-02 Wed @ 00:00 -0400 Connections logic: lies, gaps, races (dotfiles 693f820)
+Saved profiles stopped claiming "open" (security shows only when the scan knows it; subtitles are view-aware — Available adds signal %, Saved says just "out of range"). An active wired connection pins above the wifi scan in Available. Add connects immediately (Craig's decision), both add dialogs' action reads Connect. Rescan went through the op state machine (the dead guard let double-clicks double-scan). A failed initial load shows in the boxes with a retry instead of stranding "Loading…". Available's first message is "Scanning networks…". Live-info ages humanize (7m, not 445s). A held portal replaces the internet line (stable row height under the poll). The poll pauses on other tabs. VPN profiles got a VPN glyph. Forgetting the active network warns it will disconnect you.
+
+*** 2026-07-02 Wed @ 00:00 -0400 Diagnostics restructure: selector, live speed test, streamed verdicts, leaner chrome (dotfiles 787b475)
+The six-button wall became a dropdown tool selector (full tool names, a description of the selection, one Run button) revealed under Advanced. Speed Test became Network Performance: a reveal with Run Speedtest + Stop (no button-flips-to-Cancel), a once-a-second elapsed ticker while running, and Download / Upload / Ping (high-latency warn) / Server / Tip rows in the diagnose aesthetic. Get Me Online streams the diagnose rows first, announces each attempt ("Trying Reset Connection…"), then its result, and closes with a bold verdict row — as do diagnose, the single tools, and the portal login; verdicts left the status line. Get Me Online at a held captive portal opens the login flow (doctor runs the safe, reversible portal-login when fix is requested). Connecting to a network that turns out captive sends a desktop heads-up, waits a beat, then opens the login page. Diagnose evidence humanized ("open internet (HTTP 204)", "names resolve (captive.apple.com)"). The title row and Close button are gone (Esc + focus-loss auto-hide cover a transient popup) and the status line became a self-clearing toast. The AT-SPI driver anchors on the Diagnostics tab now.
+** DONE [#B] Advanced repair buttons: half width, two per row :feature:waybar:network:quick:
+CLOSED: [2026-07-01 Wed]
+The wide Advanced buttons shrink the panel and leave the diagnostics output impossible to read. Make each half width, two to a row, and rename where needed to fit. Origin: roam inbox capture.
+
+Done 2026-07-01 (dotfiles aca6827): the Advanced reveal became a 3-column, 2-row grid (Diagnose, Unblock WiFi, Reset / Restart, Test DNS, Force Portal) per Craig's follow-up; labels shortened, tooltips carry the full descriptions. Verified live + AT-SPI smoke.
+** DONE [#B] Panel action-button rows fill the panel width :feature:waybar:network:quick:
+CLOSED: [2026-07-01 Wed]
+Disconnect / Rescan / Add / Add Hidden, and the Saved row, should be as wide as the panel and the buttons above them. Apply the same homogeneous full-width treatment used on the Available / Saved sub-tabs. Origin: roam inbox capture.
+
+Done 2026-07-01 (dotfiles aca6827): both Connections action rows are homogeneous full-width, which also fixed the panel resizing when Connect flips to Disconnect. Verified live.
+** DONE [#B] Live connection info in the row subtitle :feature:waybar:network:
+CLOSED: [2026-07-01 Wed]
+The live connection information shown in the row hover should also appear in the small print under the connection name, updated in realtime like the hover. Origin: roam inbox capture.
+
+Done 2026-07-01 (dotfiles aca6827): a 1.5s poll fills the active row's subtitle with the bar-tooltip fields minus the SSID (signal, interface, internet + age, portal note, throughput), sharing the bar's per-field formatters. Verified live.
+** DONE [#B] Right-click date/time: ntp sync + timezone update :feature:waybar:
+CLOSED: [2026-07-02 Thu]
+Right-click on the date updates the clock from ntpd (or whatever keeps the clock in sync); right-click on the time runs the update-timezone script. Neither opens a terminal — both run transparently in the background. Errors surface as notifications. The module should detect whether we're online and, if not, show a message instead of running the script. Origin: roam inbox capture 2026-07-02.
+
+Shipped 2026-07-02 (dotfiles 2f7993d). Date right-click = clock-sync (chronyc makestep behind a connectivity gate; offline/captive get explanatory notifications). Time right-click = timezone-set, rewritten to prefer WiFi geolocation (whereami → timeapi.io) over IP lookup — the hotel IP geolocated two timezones off, and ipapi.co is paywalled now (ipinfo.io is the fallback). Both promptless (sudo -n), all outcomes notify, worldclock gained signal 16 for an instant refresh. 13 new tests across clock-sync + timezone-set; live-verified on velox (WiFi path resolved Rhode Island → America/New_York correctly).
+** DONE [#B] Screenshot "view image" option :feature:hyprland:
+CLOSED: [2026-07-02 Thu]
+The screenshot flow should also offer a "view image" selection: saves the shot, opens it in a viewer, and puts the path on the clipboard. Origin: roam inbox capture 2026-07-02.
+
+Shipped 2026-07-02 (dotfiles 10e5961). View Image entry in the post-capture fuzzel menu: opens the shot via xdg-open (default viewer, currently feh) and puts the path on the clipboard. Script gained env seams + a nine-test suite (menu dispatch, both capture modes, cancel, failure). Live check pending: take a shot and pick View Image once.
+** DONE [#C] Collapse-triangle buttons: dimmer, inlaid styling :waybar:
+CLOSED: [2026-07-02 Thu]
+The triangle collapse buttons should look embedded (inlaid) and be slightly dimmed so they don't compete for eye attention with the other components. Origin: roam inbox capture 2026-07-02.
+
+Shipped 2026-07-02 (dotfiles 15cb93c): muted color + dark inset well + smaller glyph; hover still brightens. Hudson theme carries the same shape. Screenshot-verified.
+** DONE [#C] Contrast button ignores the white=on / red=off paradigm :bug:waybar:
+CLOSED: [2026-07-02 Thu]
+The contrast button doesn't respect the white=on, red=off color paradigm the other waybar modules follow. Cosmetic × every time = P3. Origin: roam inbox capture 2026-07-02.
+
+Shipped 2026-07-02 (dotfiles 15cb93c). The "contrast button" is the auto-dim module — its ON icon (nf-fa-adjust) is the classic contrast glyph. It showed gold when on and nothing when off; now on = default silver, off = terracotta, matching every other toggle. Screenshot-verified both states.
+** DONE [#C] Off-state red inconsistent across waybar modules :bug:waybar:
+CLOSED: [2026-07-02 Thu]
+Terracotta red isn't applied uniformly for "off": the sleep icon is a different shade than the mouse/trackpad when off, and the dim indicator doesn't show red when off at all. Cosmetic × every time = P3. Origin: roam inbox capture 2026-07-02.
+
+Shipped 2026-07-02 (dotfiles 15cb93c + 7f1f334). The touchpad script's Pango-markup red predated the terracotta theme pass (#d47c59 vs the CSS's #cb6b4d) — unified on #cb6b4d. Dim-off now red (see the contrast task). Bonus find: waybar hands every pulseaudio instance the sink's .muted class, so a muted speaker also painted the mic red — scoped with :not(.mic) so each glyph keys on its own device.
+** DONE [#B] Waybar volume/mic toggle like the touchpad module :feature:waybar:
+CLOSED: [2026-07-02 Thu]
+Make the volume/mic waybar component look and behave like the touchpad/mouse toggle.
+- Move the mic to the other side of the volume so the percentage isn't in the way. The mic and speaker icons sit the same distance apart as the hand and mouse.
+- One keybinding cycles the four states: volume on / mic on, volume on / mic off (red), volume off (red) / mic on, volume off (red) / mic off (red).
+- Move the trackpad/mouse toggle to another keybinding (discuss an open mnemonic, e.g. =d= for disable) and assign Super+M to this module (for mute).
+Origin: roam inbox capture.
+
+Shipped 2026-07-02 (dotfiles 7f1f334). New audio-cycle script walks the four-state ring in the exact order above (wpctl-backed, explicit set-mute so the pair can't desync, 6 tests) on Super+M; live-verified the full ring on velox. Mic moved left of the speaker and hugs it via paired margins, percentage on the outside. Touchpad toggle moved to Super+Shift+I ("input devices" — d-for-disable was taken at both levels by removemaster and dim-toggle); its tooltip and tests follow.
+** DONE [#C] Timer end sends no notification :bug:waybar:
+CLOSED: [2026-07-02 Thu]
+The end of a wtimer timer didn't fire a desktop notification. Needs reproduction to confirm frequency; priority follows the severity-by-frequency matrix once known (a reliably-missing timer-end alert would rate higher). Origin: roam inbox capture.
+
+Root-caused 2026-07-02 (dotfiles ca35642). The pipeline works (a live 3s timer fired and persisted), but notify sent everything --urgency=normal and dunstrc delays normal-urgency popups while a fullscreen window has focus — a timer ending mid-video sat invisible until fullscreen exit. Alarms now go critical urgency, which rides the fullscreen_show_critical rule; verified CRITICAL in dunst history. The alarm sound (paplay, separate from dunst) was never affected.
+** DONE [#B] Bake captive-portal login into the net panel :feature:network:
+CLOSED: [2026-07-01 Wed]
+Make the captive-portal login a first-class net-panel feature instead of the one-off =~/.local/bin/hotel-wifi= script. When the engine sees a held portal, offer "Log in to this network" that runs the plain-DNS + clean-browser flow reversibly (disable DoT -> recover the portal URL from the redirect -> open a clean Chrome profile -> restore DoT when online). Reconcile with the existing =net portal= / =captive= helper, whose DNS-hijack-to-gateway model did NOT match the real Hyatt portal.
+
+Full mechanism writeup, the working script, and the integration plan: [[file:docs/design/2026-06-30-captive-portal-login.org]]. From the 2026-06-30 Hyatt saga.
+
+*** 2026-06-30 Tue @ 11:40 -0400 Engine core landed (dotfiles a7d7559)
+Replaced =net portal='s old captive-helper hand-off with a =portal-login= repair tier: drop DoT to plain DNS, probe the portal URL (302 / meta-refresh), open a throwaway browser profile, spawn a detached watcher that restores DoT once online (or on timeout). =net portal --restore= is the manual fallback. 7 tests. So =net doctor= / the bar's =net portal= hookups already run the real flow now. Remaining: (1) name the DoT-blocking cause in =net diagnose=; (2) a dedicated "Log in to this network" button in the panel's Diagnose/Repair tab (today it rides the generic =net portal=); (3) live validation against a real captive portal (unit-tested only — didn't run it live to avoid disrupting a meeting).
+
+*** 2026-07-01 Wed @ 22:41:51 -0400 Live-validated end to end against a local captive simulator (dotfiles c1401db)
+The last remainder. tests/net/captive_sim.py is a local redirect portal (302s to a login page until "logged in", then a clean 204). NET_PROBE_URL and NET_PORTAL_TRIGGERS point the whole flow at it (an overridden probe skips the interface binding, which can't reach loopback). Ran live on velox, both restore paths verified: online-detect (login click, watcher saw the 204, DoT drop-in restored within ~2s, clean exit) and the timeout fallback (a watcher that never saw online restored DoT at its 300s deadline). Real sudo mv, real resolved restarts, real redirect URL recovery, real clean-profile Chrome — against a temp drop-in dir, so live DNS was untouched. All three remainders are done; the task is closed. The remaining what-if is a real venue's walled-garden quirks, which only an actual portal exercises.
+
+*** 2026-07-01 Wed @ 21:44:05 -0400 Diagnose names the DoT block; panel gained Log in to This Network (dotfiles 51e0e2d)
+Remainders 1 and 2 landed. The dns-resolve step names the DoT pin when resolution is dead and the drop-in exists (sysio.dot_forced), and routes next_action to the portal login. The panel's hidden Open Portal button became a first-class suggested-action "Log in to This Network", shown whenever the report holds a portal signal (portal step with or without a URL, or the DoT-blocked resolution) via the unit-tested viewmodel.wants_portal_login. TDD, 33 suites green. Remainder 3 (live validation against a real portal) still open.
+
+*** 2026-06-30 Tue @ 14:59:53 -0400 Live test on velox surfaced two fixed bugs + a deeper follow-up
+Force portal (panel Repair tab) = =net-popup net portal= = the same portal-login tier. Tested live on @Hyatt_WiFi (already authorized, so no real intercept). Two bugs fixed in dotfiles (TDD, full suite green):
+- Chrome first-run wizard fired on every launch — =_open_portal= made a fresh tempfile profile but passed no first-run flags. Added =--no-first-run --no-default-browser-check= + a unit test.
+- Flashing sudo prompt for the DoT drop + pointless resolved restart on velox, where the DoT drop-in the code looks for (=/etc/systemd/resolved.conf.d/dns-over-tls.conf=) doesn't exist. Guarded =_disable_dot=/=_restore_dot= to be true no-ops (no sudo, no restart) when there's no DoT drop-in to move; tests assert no systemctl call fires.
+** DONE [#B] Consistent red=off across waybar toggle modules :waybar:
+CLOSED: [2026-07-01 Wed]
+Extend the red=off convention (just added to the touchpad/mouse indicator) to the other toggles — sound volume, microphone mute, and caffeine — so a disabled / muted / off state reads red across the board. Skip the "cross"/slash; the color alone carries it. Origin: roam inbox capture.
+
+Already implemented (verified 2026-07-01): =style.css= gives =#pulseaudio.muted=, =#pulseaudio.mic.source-muted=, and =#custom-caffeine.inhibited= the off-state color =#d47c59=, matching =#custom-touchpad.disabled=. Note: caffeine's red fires on =.inhibited= (caffeine ON / staying awake), which is arguably the inverse of "off" — leave as-is unless you want strict off=red semantics there.
+** DONE [#B] Microphone-mute keybind :feature:waybar:quick:
+CLOSED: [2026-07-01 Wed]
+A keyboard shortcut to toggle the mic mute. The pulseaudio#mic module shows the state but there's no hotkey to flip it. Wire a hyprland bind to a mic-mute toggle. Origin: roam inbox capture.
+
+Already implemented (verified 2026-07-01): hyprland.conf binds both =XF86AudioMicMute= and =Super+Shift+A= to =mic-toggle= (no conflict — airplane is Super+Shift+X).
+** DONE [#C] Keybind hints in waybar module tooltips :waybar:
+CLOSED: [2026-07-02 Thu]
+Every module's hover tooltip should list its keyboard shortcut(s), for discoverability. Audit the modules and add the bindings to each tooltip. Origin: roam inbox capture.
+
+Shipped 2026-07-02 (dotfiles 4c32aec). Audited every module: arrows (Super+[ / Super+]), sysmon (Super+R), net (Super+Shift+N), layout (Super+Shift+←/→), menu (Super+Space / Super+Shift+Q, tooltip enabled), plus the already-hinted dim/caffeine/touchpad/mic/volume. Date/time tooltips document their new right-click actions. Workspaces/window/tray don't take custom tooltips; timer has no keybind.
+** CANCELLED [#C] Smooth waybar expansion animation :waybar:
+CLOSED: [2026-07-02 Thu]
+The cluster expansion jumps instead of animating, and a few systray icons pop in one-by-one afterward, which reads as glitchy. Animate the expansion smoothly if waybar allows it — width transitions are limited, so feasibility is uncertain (hence [#C]). Origin: roam inbox capture.
+
+Assessed infeasible 2026-07-02: collapse works by config rewrite + SIGUSR2 reload, which rebuilds every widget — nothing survives to transition, GTK3 can't animate add/remove without Revealer (an upstream waybar change), and the tray pop-in is async StatusNotifier re-registration. Full findings + revisit conditions: [[file:docs/design/2026-07-02-waybar-expansion-animation-feasibility.org]].
+** DONE [#C] Optional label on timer/alarm/stopwatch items :feature:waybar:
+CLOSED: [2026-07-02 Thu]
+Let each wtimer item carry an optional short text label. The data model already supports it (=add_timer/add_alarm/add_stopwatch/add_pomodoro= all take =label=""=, and =_describe= shows =label or type=); the gap is the fuzzel-driven creation flow, which doesn't prompt for a label. Add the optional label prompt on create. Origin: roam inbox capture.
+
+Shipped 2026-07-02 (dotfiles ca35642): =wtimer new= gained a "label (optional)" fuzzel prompt after the type/value prompts; empty keeps the unlabeled default. 2 new tests (89 total in the suite).
+** DONE [#C] Alarm tooltip shows time remaining, not alarm time :bug:waybar:quick:
+CLOSED: [2026-07-01 Wed]
+The =wtimer= alarm tooltip displays the countdown (time remaining) instead of the alarm's wall-clock fire time. For an alarm set to 2:00pm, the tooltip should name the target time, not "1h 23m left". Fix the tooltip rendering in =wtimer= (dotfiles repo). Origin: roam inbox capture.
+
+Fixed 2026-07-01 (dotfiles): =_describe= now renders an alarm's wall-clock target via a new =format_clock= helper instead of =format_time(remaining)=. TDD test added; full wtimer suite (87) green.
+** DONE [#C] Waybar right-cluster module order :waybar:quick:
+CLOSED: [2026-07-01 Wed]
+Move the timer module to the rightmost position, just left of the systray, and move the battery/sysmonitor module to second-to-rightmost. Config edit in the waybar config (dotfiles hyprland tier). Origin: roam inbox capture.
+
+Done 2026-07-01 (dotfiles waybar config): =custom/timer= now sits just left of =tray= with =custom/sysmon= second-to-rightmost. waybar regenerated + reloaded live on velox; visual confirmation pending Craig.
+** DONE [#B] Pocketbook finish-or-cancel decision :pocketbook:
+CLOSED: [2026-07-02 Thu]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-06-24
+:END:
+Decided by Craig 2026-07-02, ahead of the scheduled checkpoint: remove pocketbook altogether. Executed same day — pip package uninstalled (user site clean), running instance killed, launcher gone, =pocketbook/= tree removed from the repo, Super+P rebound to toggle-touchpad (P for Pointers; Super+Shift+I unbound, waybar tooltip hint updated — dotfiles a750cb4). The org-capture popup remains the quick-notes surface.
+** DONE [#B] Provision Eask in archsetup :tooling:eask:
+CLOSED: [2026-07-02 Thu]
+Shipped 2026-07-02 (speedrun): npm global install block added after the nvm line — runs as $username with --prefix $HOME/.local, display/error_warn wrapped, output to $logfile, matching the claude-code block's shape. The npmrc decision went yes: dotfiles common/.npmrc pins prefix=${HOME}/.local (stowed; hand-linked live, npm config get prefix confirms ~/.local — dotfiles 01627cc). VM assertion added: ~/.local/bin/eask present + ~/.npmrc stowed. Live smoke: eask 0.12.9 on PATH. Full acceptance (fresh-install chime make setup/test) rides the next VM pass.
+:PROPERTIES:
+:LAST_REVIEWED: 2026-05-26
+:END:
+Add =@emacs-eask/cli= to archsetup's provisioning so fresh machines get it. Eask is installed by hand today and declared nowhere in archsetup or the dotfiles repo, yet both chime and linear-emacs depend on it (their =make setup/test/coverage= shell out to =eask=). Source: handoff from linear-emacs 2026-05-23.
+
+- Add a global npm install after the node block (=archsetup= ~2030, after =aur_install nvm=), modeled on the claude-code native-install block: run as =$username=, wrapped in =display=/=error_warn=, output to =$logfile=. Roughly =sudo -u "$username" bash -c 'npm install -g --prefix "$HOME/.local" @emacs-eask/cli'=.
+- Pin the prefix to =~/.local= so eask lands at =~/.local/bin/eask= (already on PATH) and the install runs as the user, not root. On the current machine =npm config get prefix= returns =/usr=, so eask was installed with an explicit =--prefix=.
+- Decision: also set a persistent user npm prefix (=~/.npmrc= with =prefix=${HOME}/.local=)? If yes, that =~/.npmrc= is a legitimate dotfile to stow; if no, rely on the explicit =--prefix= flag alone. =~/.eask/= is a regenerable cache — leave un-stowed.
+- Acceptance: fresh run leaves =eask= on PATH at =~/.local/bin/eask= (no root); =cd ~/code/chime && make setup && make test= works.
+** DONE [#C] Waybar timer dialog styling :waybar:
+CLOSED: [2026-07-02 Thu]
+From Craig's roam capture 2026-07-02: style the timer module dialogs like the screenshot dialog — tighter window, icons on the selections, colon+space after the prompt.
+
+Shipped 2026-07-02 (dotfiles 9ffcba7): dialogs size to content, type menu carries the kind glyphs, prompts end ": ". Three new tests; screenshot-verified live.
+** DONE [#B] Waybar collapse jumps client windows :bug:waybar:hyprland:
+CLOSED: [2026-07-02 Thu]
+From Craig's roam capture 2026-07-02: collapsing/expanding (and any waybar teardown) snapped every tiled window up and back down; hold the clients still and let only the bar change.
+
+Shipped 2026-07-02 (dotfiles 4b1a4ec): waybar now runs exclusive:false and the new waybar-reserve script statically reserves the bar strip per monitor (wired as exec so config reloads re-apply it). Verified live: client geometry held constant through bar kill, relaunch, and a collapse round-trip. Eight new tests (script + pairing).
+** DONE [#B] Bluetooth panel + bar module :feature:waybar:bluetooth:
+CLOSED: [2026-07-02 Thu]
+:PROPERTIES:
+:SPEC_ID: 1271a845-4463-4831-9902-990eda6b2265
+:END:
+Spec: [[file:docs/design/2026-07-02-bluetooth-panel-spec.org]] (IMPLEMENTED 2026-07-02 — all five phases shipped same day: engine eb2230f, panel 76b2c05, bar module e372de3, bt-priv + blueman retirement 2a026b1/d8d8c53, install wiring proven by VM assertions). Residual: the phase 4-5 VM assertions run on the next VM pass; ratio picks up the package removal + hand-links on its trip list.
+
+A bluetooth panel driving a CLI underneath (bluetoothctl one-shot verbs), consistent in look and feel with the net panel (GTK4 + layer-shell + Blueprint, humble-object presenter, verify-everything). Minimalistic interface, full functionality, plus a diagnostics/troubleshooting section mirroring the net panel's Diagnostics tab. Bar module glyph opens it. Craig's ask (2026-07-02): follow UX/UI best practices; where the net panel's patterns conflict with best practices, file a net-panel bug task rather than clone the flaw.
+
+*** 2026-07-02 Thu @ 13:30:42 -0400 Shipped phase 1 — the bt engine package (dotfiles eb2230f)
+=bluetooth/src/bt/= mirrors the net engine's layout: btctl parsing boundary (show/devices/info, connect-error classifier), sysio rfkill/airplane, audio module over pw-dump/wpctl (HSP probe + A2DP switch repair with verify-after), redacted eventlog, six repair tiers, and the doctor chain (adapter → rfkill → service → powered → devices → audio profile) with safe auto-repairs behind =--fix= (never auto-connects; airplane blocks are named, not fought). 101 tests over fake binaries; 42 suites green (=make test= glob auto-discovered =tests/bt/= — gate check verified). Live read-only on velox: =bt status= + =bt doctor= read the real adapter/devices/audio graph; =~/.local/bin/bt= hand-linked (no restow under running Hyprland). Ground truth vs spec: profile inventory needs =pw-dump= (wpctl can't enumerate), and the card's =bluez5.profile= prop is unreliable — sink node's =api.bluez5.profile= is authoritative. Deferred INTO phase 2: the shared dupre css factoring (net's css is an inline string in =gui.py=, not an asset — factoring it without the bt-panel consumer just risks the working net panel).
+
+*** 2026-07-02 Thu @ 14:15:27 -0400 Shipped phase 2 — the GTK panel (dotfiles 76b2c05)
+Cloned the net panel's shape over the phase 1 engine: GTK-free PanelModel + viewmodel (69 new tests, display-free), Blueprint pages (Devices with the adapter power row + Paired/Nearby sub-views; Diagnostics with the doctor cascade, inline fix buttons mapped from step =repair= keys, Advanced tool selector), gui.py controller (layer-shell OVERLAY 380x520 TOP+RIGHT, Esc closes, single-instance toggle, bg worker, passkey + confirm dialogs), pairing pty state machine (=bt/pairing.py=: confirm-passkey yes/no with default-deny, display-passkey, bounded deadline, tested over the fake's interactive mode), manage.py op envelopes shared by CLI and panel (cli refactored onto it; power + discoverable verbs added), =bt panel= subcommand + =bt-panel= toggle wrapper (hand-linked into =~/.local/bin=, no restow under live Hyprland). Shared dupre css factored: net's inline =_CSS= → =hyprland/.config/themes/dupre/panel.css= with =dupre-*= classes, both panels load it (stowed path first, repo-relative fallback; hand-linked =~/.config/themes/dupre/panel.css=). Super+Shift+B rebound blueman-manager → bt-panel (hyprctl reloaded, live). 43 suites green (=make test= exit 0). DEFERRED pending Zoom ending: the AT-SPI smoke (=make test-panel-bt=, written and wired) and any visual check of either panel with the factored css — both need a visible window. Gotcha for posterity: the old =test_panel_stub_exits_two= CLI test ran =bt panel= for real once cmd_panel was wired and launched the panel on the live compositor mid-meeting for ~30s before the test timeout killed it; it now asserts parser wiring only — never run =bt panel= inside =make test=.
+
+*** 2026-07-02 Thu @ 15:06:00 -0400 Shipped phase 3 — the bar module + blueman retirement (dotfiles e372de3)
+=custom/bluetooth= over the engine: waybar-bt shim + =bt/indicator.py= (state-following glyph — slashed off/blocked/absent, plain dim idle, connected mark white; low-battery <15% adds a red pango percentage to the glyph; tooltip = connected devices with battery + Super+Shift+B hint). Signal 10; the panel pokes =pkill -RTMIN+10 -x waybar= after each status reload. Blueman retired from the Hyprland session: exec-once + both windowrules removed, applet killed live; waybar relaunched on the runtime config and the module verified on the bar (connected glyph, blueman tray icon gone). Theme drift guard caught that themes/*/waybar.css edits must mirror into the live =waybar/style.css= — all three updated. 43 suites green. ALSO closed this pass: the deferred phase 2 visual batch (bt AT-SPI smoke green after fixing its Connect/Disconnect state-following assertion — c1a8219; net smoke green on the factored css; both panels eyeballed correct in dupre). Left for phase 4: package removal (blueman out of archsetup), sxhkdrc's dwm blueman-manager binding decision rides that pass.
+
+*** 2026-07-02 Thu @ 15:16:51 -0400 Shipped phase 4 — bt-priv shim, blueman out, VM assertions
+Dotfiles =2a026b1=: the stowed =bt-priv= shim over the phase-2 =bt.priv= module (one verb, =restart-bluetooth=; verified end-to-end against the symlinked fake-systemctl — rc 0 with =BT_SUDO= empty, rc 2 on bad verb/usage; hand-linked into =~/.local/bin=), and the sxhkd =Super+Shift+B= bind repointed from the retired blueman-manager to =st -e bluetoothctl= (the decided terminal fallback — the GTK panel is Wayland-only, and the bt CLI is hyprland-tier so dwm never gets it). 43 dotfiles suites green. archsetup: blueman dropped from the =desktop_environment= bluetooth loop (bluez + bluez-utils stay, solaar untouched); VM assertions added to =test_packages.py= (bluez/bluez-utils installed, blueman NOT installed as the retirement regression guard — collected 15, exercised on the next VM run since VM tests run committed code); =bash -n= + =py_compile= + =make test-unit= green. SUDOERS: no new rule needed, same conclusion as net-priv (2026-07-01 entry) — archsetup:1089 grants the primary user blanket =NOPASSWD: ALL=, which covers =systemctl restart bluetooth=; a narrow bt-priv rule would be dead config under the blanket grant, so phase 5's "sudoers placed" item is satisfied by the existing grant. LIVE: blueman package removed from velox (=pacman -Rns=, decision "drop it outright, both machines"); ratio needs the same + the bt-priv hand-link on its trip list.
+
+*** 2026-07-02 Thu @ 15:19:58 -0400 Shipped phase 5 — install-default wiring proven by VM assertions
+No new install code was needed: the waybar =custom/bluetooth= module, the =Super+Shift+B= → =bt-panel= bind (hyprland.conf), and the shared =themes/dupre/panel.css= all live in the dotfiles hyprland tier, so the existing clone + =make stow hyprland= step lands them on a fresh install; sudoers is covered by the blanket grant (phase 4 conclusion). The phase's substance is the proof: =test_desktop.py= gained hyprland-gated assertions that the four bt bins (=bt=, =bt-panel=, =bt-priv=, =waybar-bt=) are stowed executable in =~/.local/bin= (either stow shape — per-file symlink or folded dir), the waybar config carries =custom/bluetooth=, hyprland.conf carries the =bt-panel= bind, and the stowed theme has =panel.css=. Collected 30 in =test_desktop.py=; exercised on the next VM run (VM tests run committed code).
+
+*** 2026-07-02 Thu @ 15:19:58 -0400 Test surface complete across all phases
+Everything the surface named exists and is green: engine suites over fake binaries (phase 1, 101 tests — btctl parse, doctor chain, A2DP repair verify), PanelModel presenter suite (phase 2, 69 tests), pairing state-machine suite (passkey confirm / NoInputNoOutput / timeout, over the fake's interactive mode), bar-module suite (phase 3), gated AT-SPI smoke (=make test-panel-bt=, run green live), and the phase 4-5 VM assertions (=test_packages.py= bluetooth stack + blueman-absent guard; =test_desktop.py= panel wiring). 43 dotfiles suites green; VM assertions await the next VM run.
+** DONE [#B] All error messages should be actionable with recovery steps
+CLOSED: [2026-07-02 Thu]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-06-24
+:END:
+Shipped 2026-07-02 (speedrun). Structural fix at the helper: =error_fatal= now takes an optional third recovery-hint arg and every fatal prints the last five log lines inline, the full log path, the per-site "Fix:" when given, and the resume pointer (step markers mean a re-run continues where it stopped) — so even a hint-less fatal is actionable. All 17 fatal call sites got specific hints (keyring reinit, mirrorlist switch, userdel/USERNAME conflict, base-devel for makepkg, DESKTOP_ENV values, dotfiles-dir cleanup, tmpfs sizing, aur.archlinux.org reachability). The end-of-run Error Summary now closes with the grep-the-log line and the fix-and-re-run pointer. =error_warn= already carried what-failed + exit code into the summary; unchanged.
+** DONE [#B] Improve logging consistency
+CLOSED: [2026-07-02 Thu]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-06-24
+:END:
+Shipped 2026-07-02 (speedrun), paired with the actionable-errors task. Audit result: the install helpers (pacman_install/aur_install/retry_install/run_task/git_install/pipx) and error helpers already tee/append everything to $logfile — the gaps were direct mutations whose stderr went to the console and vanished. Swept every =sed -i= and file-write mutation lacking capture (locale.gen uncomment, pacman.conf ParallelDownloads/Color + multilib, waybar battery removal x3, wireless-regdom, geoclue BeaconDB, paccache, BRIO udev rule, fstab fmask, mkinitcpio HOOKS, sudoers append, ufw status read): each now sends stderr to $logfile, and the previously-silent ones (locale.gen, pacman.conf, multilib, waybar, regdom, geoclue, paccache, udev) gained =error_warn= handlers so failures land in the summary instead of passing silently. Verified: bash -n clean, 10 unit suites green, shellcheck warning-diff vs HEAD empty (no new findings).
+** DONE [#B] Add NVIDIA preflight check for Hyprland
+CLOSED: [2026-07-02 Thu]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-05-21
+:END:
+Shipped 2026-07-02 (speedrun), TDD. =nvidia_preflight_report= is a pure sed-extractable core (same harness pattern as zig-pin): modalias scan for vendor 10DE — DRM first, PCI display-class (bc03) fallback so an NVIDIA audio function can't false-trigger — then the repo's =nvidia-utils= candidate major checked against 535. Prints the Wayland guidance + env vars (LIBVA_DRIVER_NAME, GBM_BACKEND, __GLX_VENDOR_LIBRARY_NAME, ELECTRON_OZONE_PLATFORM_HINT) and the pre-Turing/AUR-legacy note. preflight_checks aborts on <535/unknown (rc 11), prompts continue/abort on a healthy NVIDIA box (rc 10), silent on non-NVIDIA (rc 0). 9 Normal/Boundary/Error tests over fake modalias trees + a fake pacman (=tests/nvidia-preflight/=, glob-discovered by test-unit — 10 suites green).
+** DONE [#C] Wlogout exit-menu buttons are rectangular, not square
+CLOSED: [2026-07-02 Thu]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-06-24
+:END:
+The wlogout exit menu renders its buttons taller than they are wide on velox, so the cells read as vertical rectangles instead of squares. They render square (centered) correctly on ratio, so this is a per-host / resolution difference, not a flat bug. Fix the button sizing in the wlogout style (=~/.dotfiles/hyprland/.config/wlogout/style.css=) so each cell is square on both hosts. Noticed 2026-05-21. Related: the [#D] VERIFY about wlogout sizing across displays.
+
+The wlogout config uses fixed pixel margins, which is the likely reason sizing differs across the two displays — adjusting them for the laptop screen is part of the fix (folded in from the former "Test wlogout menu on laptop" VERIFY, 2026-06-24).
+
+Add a regression test so the square-cell fix doesn't silently break on a resolution change: assert the rendered (or computed) wlogout button cells are square across ratio's and velox's resolutions. Dropped :quick: — the cross-host test pushes this past a spare-moment fix.
+
+Shipped 2026-07-02 (dotfiles 775771b). Keybind now calls a =wlogout-menu= wrapper computing centered margins from the focused monitor (the old fixed L/R 1200 exceeded velox's 1436 logical width). Also fixed two styling defects the geometry hid: invisible unfocused borders (now muted, so the square edge is visible) and hover/focus sharing one gold rule (lock button glowed at launch; focus is now a muted ring). Tests: unit margin-math suite across both hosts' resolutions + portrait + small + bad-geometry, CSS regression suite, and a compositor-gated =make test-wlogout= smoke that launches a no-op probe, screenshots, and measures squareness (velox: 361x361 px, PASS). Ratio's visual eyeball rides the pending ratio sync.
+** DONE [#C] Net panel: error toasts auto-dismiss unread :bug:network:waybar:
+CLOSED: [2026-07-02 Thu]
+Fixed in dotfiles 0f017d4: viewmodel.toast_plan owns the toast policy — errors show sticky and ignore the post-op refresh's empty clear (worst case: a forget failure's error was wiped within ~2s by its own refresh), and the next real status replaces them. Successes keep the 4s fade. 7 policy tests added; 41 suites green.
+** DONE [#C] Net panel: verify claimed keyboard navigation :test:network:waybar:
+CLOSED: [2026-07-02 Thu]
+Found during the bluetooth-panel UX pass (2026-07-02). The V2 spec claims tab-between-sections, arrow-key row navigation, and type-to-filter, but no custom keyboard code exists in the panel — arrows and type-ahead may ride GTK ListBox defaults, tab-between-sections likely doesn't. Verify each claim against the live panel (AT-SPI smoke can assert focus order); implement or strike the claims from the spec so spec and panel agree.
+
+*** 2026-07-02 Thu @ 13:05:00 -0400 Code-level pass done; live probe deferred (Craig in a Zoom meeting)
+Code reality (dotfiles net/src/net): Esc close is wired (gui EventControllerKey); row-activated -> primary is wired for both connection lists (pages.py:122,126), so Enter-on-row rides the GTK ListBox activate binding; arrows ride ListBox defaults; NOTHING implements type-to-filter (no search/filter code exists — that claim is false as written); Tab is the plain GTK focus chain, widget by widget, not section jumps. Live AT-SPI probe plan: launch panel in test mode, drive keys via hyprctl dispatch sendshortcut targeted AT THE PANEL WINDOW (never the focused surface — wtype/ydotool absent anyway), gate every key on the panel holding focus, never send Enter on the available list (real connect risk). BLOCKED at 13:05: active window is zoom (meeting) — no test windows, no synthetic input until clear. Then: verify focus order + arrows + no-filter live, strike/reword the spec's keyboard bullet to match.
+
+*** 2026-07-02 Thu @ 14:57:18 -0400 Ran the live probe; spec bullet reworded to match reality
+Zoom ended ~14:45; probe ran per plan (panel in test mode, hyprctl dispatch sendshortcut targeted at the panel address, every key gated on panel focus, Enter never sent). Verdicts: arrows move row focus and Enter rides the ListBox activate binding (TRUE — kept); Esc closes (TRUE — kept); Tab is the plain GTK widget-by-widget chain and inside a list crawls row by row, no section jumps (claim FALSE — struck); type-to-filter does not exist (claim FALSE — struck; typing into the 24-row Saved list filtered nothing). Spec's Keyboard bullet reworded with the live evidence and a note that section-jump Tab or filtering would be new work. Probe gotchas for reuse: AT-SPI list items have empty accessible names, so row identity needs get_index_in_parent(); a killed test panel can leave a windowless single-instance process that eats the next launch via D-Bus activation — pkill -9 -f 'net panel$' and wait before relaunching.
+** CANCELLED [#C] Pocketbook development backlog :pocketbook:
+CLOSED: [2026-07-02 Thu]
+Cancelled with the 2026-07-02 remove-pocketbook decision — the app and its in-tree package are gone.
+:PROPERTIES:
+:LAST_REVIEWED: 2026-05-26
+:END:
+Pocketbook (GTK4 layer-shell notes panel, toggled via waybar) was pulled from publication 2026-05-26 — github repo + cjennings.net repo deleted, mirror hook removed — and folded into this repo at =pocketbook/= until it's ready to spin back out. Src-layout Python package with pytest tests and a Makefile. Develop it in-tree; the backing modules are =store/note/panel/layer_shell/app/note_widget= + =style.css=.
+
+Backlog (unordered; promote items to their own dated tasks as they're picked up):
+
+- Configurable options, possibly a dedicated configuration panel.
+- Lose-focus hides pocketbook — configurable on/off.
+- Configurable display order: chronological by creation date (asc/desc), manual, alphabetical (asc/desc).
+- Search / filter notes.
+- Global toggle keybind (Hyprland =bind=) alongside the waybar click; document the waybar integration.
+- Note CRUD polish (create/edit/delete) + optional markdown rendering.
+- Pin / favorite notes.
+- Tags or notebooks / categories.
+- Persistence: confirm store format + =~/.local/share/pocketbook/= location, add versioning/migration, decide a backup/sync story.
+- Theming: track the dupre/hudson theme system so =style.css= follows =set-theme=.
+- Layer-shell geometry config (anchor edge, width, margins) + HiDPI / multi-monitor behavior — ties into [[file:docs/PLAN-per-host-overrides.org][per-host overrides]] scaling work.
+- Config file format (toml) + reload-without-restart.
+- Expand test coverage (TDD per testing standards; =tests/= already exists).
+- Release prep for the eventual spin-back-out: pyproject metadata, version, license.
+- Re-wire the archsetup install (gtk4-layer-shell dep + install step + post-install clone) when pocketbook ships. Removed 2026-05-26 — see git history of =archsetup= / =scripts/post-install.sh=.
+** CANCELLED [#C] Fn+F9 toggles pocketbook — source unlocated :hyprland:pocketbook:
+CLOSED: [2026-07-02 Thu]
+Retired with pocketbook itself (2026-07-02 removal) per this task's own exit condition — with the app uninstalled and unbound, whatever Fn+F9 emitted has nothing to toggle.
+:PROPERTIES:
+:LAST_REVIEWED: 2026-06-23
+:END:
+On velox, pressing Fn+F9 (physical function key) toggles the pocketbook panel. It shouldn't. Raised from a home-project session 2026-06-23.
+
+Investigated 2026-06-23 and could not locate the trigger in any config. Ruled out, three ways:
+- No F9 bind (bare / $mod / keycode) in the live =hyprland.conf= (now a stow symlink), the velox host tier =conf.d/local.conf=, or the waybar config.
+- =hyprctl binds= runtime (all 90 active binds, authoritative) execs pocketbook on ONLY =SUPER+P=. No F9/XF86 path reaches it. The old touchpad toggle that used to sit on =$mod+F9= was moved to =$mod+M=, so F9 is unbound in Hyprland.
+- No input remapper (keyd/xremap/input-remapper) and no hotkey daemon (sxhkd/swhkd) running or configured; pocketbook's own source has no F9 / GlobalShortcuts / portal / dbus listener (its GTK ShortcutController binds only Esc/Ctrl-n/Ctrl-j/Ctrl-k/Del/Return). pocketbook is a single-instance Gtk.Application, so any path that re-runs =pocketbook= toggles it.
+
+Parked at Craig's call (not worth deeper investigation now). If it resurfaces, the one unfinished step is to capture what keysym Fn+F9 actually emits (=wev -f wl_keyboard:key=, press Fn+F9, read the =sym:= / =code:=) and grep for that. Most likely folds into removing pocketbook from the waybar setup — if pocketbook leaves the bar, retire this with it.
+** CANCELLED [#C] Waybar emacs-service status + control :feature:waybar:
+CLOSED: [2026-07-02 Thu]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-06-24
+:END:
+From the roam inbox (2026-06-22): with Emacs integrated into the system as file manager and instant note-taker, make bouncing it trivial. A waybar component showing the emacs service status, with detail on hover, that turns the server on / off / bounce via right-click. Pairs with running the Emacs daemon as a managed systemd user service.
+
+Cancelled 2026-07-02 per Craig during the task-batch pick: no current need. Re-add or pull back from Resolved if a need surfaces.
+** DONE [#C] set-wallpaper detaches waypaper config from its stow symlink :bug:hyprland:quick:solo:
+CLOSED: [2026-07-02 Thu]
+:PROPERTIES:
+:LAST_REVIEWED: 2026-06-28
+:END:
+=set-wallpaper= persists with =mv "$tmp" "$CONFIG"=, which replaces the =~/.config/waypaper/config.ini= stow symlink with a real file. After the first run the live config is detached from =~/.dotfiles/hyprland/.config/waypaper/config.ini=, so a later =git pull= + restow won't update it and set-wallpaper changes never flow back to the repo. Fix: write in place rather than =mv= over the symlink — e.g. =cp "$tmp" "$CONFIG"= (follows the symlink to the real dotfiles file), or resolve the link target and write there. Lives in =~/.dotfiles/hyprland/.local/bin/set-wallpaper=; it has a test suite, so add a Boundary case for "CONFIG is a symlink".
+
+Shipped 2026-07-02 (dotfiles d826be4): write-back now redirects through the symlink instead of mv-ing over it; two boundary tests pin the invariant (replace + append paths). velox's live config was still a healthy symlink, so no repair needed.