Jihad-Browser/context/plans/build-site.md

created, last_edited

created	last_edited
2026-10-01	2026-10-24

Build Site

91 tasks across 54 tiers (Tier 0-52 plus a final BLOCKED-EXTERNAL tier), covering exactly the 53 acceptance criteria that are [~] (31) or [ ] (22) in context/kits/ on 2026-10-01 - nothing else; the 338 [x] criteria are recorded as met in the kits and are not restated. Task IDs start at T-201; a row that says "continues T-1xx" inherits that task's remaining work and gating from build-site-2026-08-10.archived.md. Device lane: there is ONE TouchPad, so every agent-run task with Device: yes sits in a single total order (see ## Device lane): each is blockedBy the previous lane task in addition to its real blockers, and the lane is why the tier count is high - the tiers below are mostly one device sitting each, with host work running in parallel beside them. Host tasks run in parallel except that engine builds (libxul) and daemon builds are one at a time; the daemon build follows the engine build because it links the same objdir. Batch rule: all three variants stay on one engine + daemon build. Code fixes are batched - implement, build, ONE push to all three (build/webos-oe/push-engine-update.sh enyo mochi mojo, or an .ipk upgrade of all three when the adapter or bundled files change) - then the device verifications run in lane order. Batch 1 carries every fix known today; batch 2 carries the plugin VKB adapter work plus whichever [CONDITIONAL] fixes the lane triggers. Gates: the 13 tasks in the last tier need a person at the device, hardware that does not exist (no keyboard), a credential, or a product-owner decision; their Device column reads human because no agent runs them and they are outside the lane. A daemon-level inject is never counted as the human test. Every device row names its observable; "Pre" means: assert the daemon log GREW and the card is CONNECTED before believing a reading, and record the libxul md5 beside the result.

Tier 0 — No Dependencies (Start Here)

Task	Title	Cavekit	Requirement	Device	Effort
T-201	Housekeeping and baselines before any push: df /var first; remove enyo's leftover MOZ_LOG=... from /etc/event.d/jihad then stop jihad; start jihad; confirm JIHAD_INJECT=1 in all three jobs; record each libxul md5 (enyo traced cb514e9e..., mochi/mojo 78c3b51a...); check whether org.webosinternals.ipkgservice exists on CE. Record PRE-CHANGE figures: no-card CPU from /proc/<pid>/stat over 60 s, log bytes/min, about:blank painted lines per 30 s, connect-to-first-painted ms on a static page. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R8/R4	yes	M
T-202	Engine fix for the <video> MEDIA_ERR_DECODE race: in third_party/uxp/gfx/layers/basic/BasicImages.cpp BasicPlanarYCbCrImage::CopyData, return the base RecyclingPlanarYCbCrImage::CopyData result and true under delayed conversion (PICKUP NEW DEFECT 1); commit on jihad-engine-mods so the OE SRCREV can follow. Observable is T-220.	cavekit-gre-widgets.md	gre-widgets R1	no	S
T-203	Daemon idle state: with no card connected stop the 8 ms frame tick (nothing pumped, painted or paced; wake ~1/s or less), keep polling the inject channel at the idle cadence so it can still answer no page, wake to the full tick on connect without delaying the first frame, return to idle on disconnect.	cavekit-device-build.md cavekit-engine-embedding.md	device-build R8 engine-embedding R3	no	M
T-204	Bound the log while running: stop the periodic tick period line when there is nothing to report; enforce the 2 x 2 MB cap inside the daemon (rotate/reopen or stop writing, same PID, with a test knob that lowers the cap); write the three-variant worst case into the job comment (packaging/gen-variant-scripts.sh) and the docs/ footprint list.	cavekit-device-build.md	device-build R8	no	M
T-205	Blank-page settle in the full paint path (BrowserPageGoanna): record that a zero-content frame was refused as FINAL for this document/load and stop retrying until an invalidation, navigation or resize; keep the blank-over-good suppression for a blank frame mid-load of a page that later has content; no per-paint log line while settled.	cavekit-offscreen-rendering.md cavekit-engine-embedding.md	offscreen-rendering R8/R3 engine-embedding R3	no	M
T-206	Confirm or refute the fresh-install -1027 from source: packaging/gen-variant-scripts.sh postinst starts the job (start @@JOB@@) before ls-hubd has read the new role (its own comment says the role is live only after the requested reboot); read JihadLunaService.cpp for any retry. If real, fix so the service registers without a hand restart (daemon-side LSRegisterPalmService retry with backoff so it rides the daemon build, and/or a postinst hub rescan before start).	cavekit-device-build.md	device-build R4	no	M
T-207	Prepare R9's two device items (continues T-104): build the PTHREAD_PROCESS_SHARED __kind probe dynamically linked with the PDK gcc 4.3.3 so it runs against the device's own glibc 2.8 libpthread, and a script that pulls libflashplayer.so's DT_NEEDED closure off the device (novacom get) for a host readelf --dyn-syms scan for sem_open/*attr_setpshared.	cavekit-device-build.md	device-build R9	no	S
T-208	Clean-clone residuals (continues T-115/T-154): git ls-files build/webos-oe/toolchain/ - git add the three crosstool recipe files if still untracked; git ls-remote the UXP-Jihad fork for jihad-engine-mods and confirm the parent gitlink is a pushed commit. If it is not pushed, the push is T-279.	cavekit-device-build.md	device-build R3	no	S
T-209	Test assets: regenerate >=10 s clips with ffmpeg (VP8, VP9, Theora, Vorbis, Opus, WAV, FLAC, MP3, AAC, H.264) per PICKUP's recipes; extend render/goanna/test/jihad-media.html #state with ended/error/currentTime/duration/decoded+dropped frames/volume/muted plus a fixed-width variant; add an input-echo page (plain <input>, <textarea>, controlled re-set-from-state input); stage via stage-test-pages.sh.	cavekit-gre-widgets.md cavekit-input-bridging.md	gre-widgets R1 input-bridging R2a	no	M
T-210	Refresh app-mochi/PARITY.md against today's app/ and app-mochi/ (safeUrl, settings entry, toast and other post-2026-07-31 changes) so the human sign-off (T-291) is on a current checklist. Continues T-113.	cavekit-mochi-ui.md	mochi-ui R2	no	S
T-211	Adapter latch instrument (continues T-124): a g_message on every mFlashGestureLock set/clear with its call site in render/adapter/BrowserAdapter.cpp (DEBUG/TRACEF compile to nothing today), so the runtime value is readable via palm-log; ships in the batch-1 .ipks (PDK build).	cavekit-input-bridging.md	input-bridging R7	no	S
T-212	Plugin VKB raise (continues T-144): in the adapter raise the VKB when the plugin latch engages on a plugin that wants text and lower it when the latch clears, mirroring the msgEditorFocused path; document what signals "wants text" and that the latch is double-tap only. Batch 2 (adapter, .ipk).	cavekit-input-bridging.md	input-bridging R7	no	M
T-213	Keycode follow-ups in BrowserPageGoanna::keyDown (continues T-127/T-109): extend the PUA swallow to the real 0xE0E2-0xE0E7 codes so they stop inserting invisible glyphs; add an env-gated log of the raw (key, chr, modifiers) triple at asyncCmdKeyDown (codes only, suppressed for password fields) - the instrument the human VKB session (T-282) needs.	cavekit-gre-widgets.md	gre-widgets R3	no	S
T-214	From source, explain and fix the two 2026-08-15 inject findings: key 57346 (0xE0A2) did not move the caret in a <textarea> (abcd + 2 x Left + X gave abcdX), and the first text after a focus change into a <textarea> was dropped; state whether the inject key path reaches the keycode map at all.	cavekit-gre-widgets.md cavekit-input-bridging.md	gre-widgets R3 input-bridging R2a	no	M
T-215	Settings-command instrument (continues T-116): one log line per frozen handler (SetEnableJavaScript, SetBlockPopups, SetAcceptCookies, SetMinFontSize, SetUserAgent) with the received value; a test page that opens its popup from a click handler (not during load, the 2026-08-15 confound) and a script-marker page.	cavekit-preferences-ui.md	preferences-ui R5	no	S

Tier 1 — Depends on Tier 0

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-216	Engine build, batch 1: incremental libxul in the podman container (build-goanna-arm.sh build); record its md5. The only engine build in flight.	cavekit-gre-widgets.md	gre-widgets R1	T-202	no	S

Tier 2 — host work

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-217	Daemon build, batch 1 (build-daemon-arm.sh) with every batch-1 daemon change; check-yap-contract.sh passes (73 commands incl. 0x1600, 58 messages). Serialized after T-216 because it links against the same objdir.	cavekit-device-build.md	device-build R8/R4	T-203, T-204, T-205, T-206, T-213, T-215, T-216	no	S

Tier 3 — device sitting: Batch-1 push to all three variants

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-218	Batch-1 packages: build-adapter-pdk.sh AND build-adapter-arm.sh (with T-211), then build-variant-ipk.sh for enyo/mochi/mojo carrying T-216's libxul, T-217's daemon and T-206's postinst; verify each payload's libxul/daemon md5 equals the built ones and libasound.so.2 is present.	cavekit-device-build.md	device-build R4/R3	T-211, T-217	no	S
T-219	Batch-1a ENGINE push (split from the daemon push 2026-10-01 so the <video> verification is not confounded by — or held up behind — the daemon changes): push the T-216 libxul to all three variants with the daemon left exactly as it is; read back libxul md5 on all three (= T-216; enyo leaves the traced build) and the unchanged daemon md5; JIHAD_INJECT=1 intact; clean restart of each job; push T-209's pages and clips to /tmp/jm/. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-gre-widgets.md	gre-widgets R1	T-201, T-216, T-209	yes	S

Tier 4 — device sitting: Verify:

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-220	<video> plays to the end (continues T-101/T-125): VP8/WebM and Theora/Ogg >=10 s clips over file:// and http://, three consecutive runs each; read #state via gettext: ended=true, error=null, currentTime within one frame of duration, decoded frames of the order of the clip's frame count (not 10-60); libxul md5 recorded for all three variants. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-gre-widgets.md	gre-widgets R1/R7	T-219	yes	M

Tier 5 — device sitting: Verify: codec reality table + undecodable-media degradation

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-221	Codec reality (continues T-101): per format (VP8, VP9, Theora, Vorbis, Opus, WAV, FLAC, MP3, AAC, H.264) record canPlayType, plays-to-end, and the reason where not (missing decoder vs defect) from #state; anon video shows the videocontrols controlBar attached; screenshot the error overlay for an H.264 <video> and an MP3 <audio>. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-gre-widgets.md cavekit-device-build.md	gre-widgets R1 device-build R4	T-220	yes	M

Tier 6 — device sitting: Verify: play/pause both directions by daemon click

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-222	Play/pause both directions by daemon-level clickat on the control bar (continues T-145): long clip, reveal the auto-hidden bar before each click, read paused/currentTime before and after; both directions must change state. The finger test stays human (T-285). Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-gre-widgets.md	gre-widgets R1	T-221	yes	S
T-223	Bring packaging/prefs/jihad-platform-prefs.js in line with T-220/T-221: the <video> reason is true again, and the stale "no audio output" reasoning becomes the 2026-09-27 measured state (continues T-138).	cavekit-gre-widgets.md	gre-widgets R7	T-221	no	S

Tier 7 — device sitting: Measure: does engine media need the CPU boost (R10)

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-224	R10 media measurement (continues T-118/T-134): stock tunables read back, scaling_cur_freq once a second across a >=10 s <video> with dropped/decoded counts; floor + drops means extend, otherwise record the numbers and close "do not extend"; tunables read back stock afterwards; scaling_governor READ only. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R10	T-222	yes	S

Tier 8 — device sitting: Verify: idle daemon CPU, wakeups, connect latency, re-idle

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-225	Batch-1b DAEMON push first: push-engine-update.sh enyo mochi mojo with the T-217 daemon (libxul unchanged from T-219), md5 read back on all three, JIHAD_INJECT=1 intact, clean restart of each job. Then idle daemon: all cards closed, per variant /proc/<pid>/stat utime+stime over >=60 s (<0.5% of a core) and wakeups from the voluntary_ctxt_switches delta (~1/s or less); an inject url with no card still answers no page; reconnect, connect-to-first-painted vs T-201's baseline; close the card and see the idle figures return without a restart. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R8	T-224, T-217	yes	M
T-226	[CONDITIONAL] Only if T-224 shows the clock at the floor with frames dropping: extend the boost to engine media with its own owner, release path and post-stop restore (patch 0027 model); batch 2.	cavekit-device-build.md	device-build R10	T-224	no	M

Tier 9 — device sitting: Verify: log bounded idle (10 min) and past the cap (same PID)

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-227	Log bound: (a) no card, a 10-minute window, daemon.log growth per variant of at most a few hundred bytes; (b) one variant with the cap lowered by T-204's knob and a noisy page drives the log past the cap in one lifetime: it rotates or stops with the same PID and logs again afterwards, then restore the job; (c) the worst-case number is in the job comment. Assert the log GREW during (b). Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R8	T-225	yes	M

Tier 10 — device sitting: Verify: about:blank settles; real page still paints; no white flash

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-228	Blank settle: about:blank for 30 s, painted lines vs the static-content control and T-201's baseline (a handful, not hundreds); then navigate to a real page and see its first frame promptly; a page blank mid-load that then gets content must not publish the blank over the previous good frame (frame capture). Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-offscreen-rendering.md	offscreen-rendering R8/R3	T-227	yes	M

Tier 11 — device sitting: Measure: idle CPU with a static page loaded

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-229	Engine idle: static page loaded, no input, /proc/<pid>/stat CPU over 60 s per variant; record the figure on engine-embedding R3. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-engine-embedding.md	engine-embedding R3	T-228	yes	S

Tier 12 — device sitting: CE baseline md5s + Atlas identity columns

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-230	Fresh CE baseline: md5 of the stock file set plus Atlas's files (replaces the 3.0.5 baseline-v2); read Atlas's identity columns off the device (NPAPI MIME, shim filename, YAP name, socket, upstart job, Luna name) and compare with all three Jihad variants - no column may be equal.	cavekit-device-build.md	device-build R4/R8	T-229	yes	S

Tier 13 — device sitting: Removal audit #1 to a role-less start

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-231	Removal audit #1 and role-less start: remove the three variants one at a time through the supported installer (the remaining ones keep serving after each removal; prerm removes exactly what postinst created, naming any failing assertion; rootfs read-only again); then show no app dir, job, shim/impl, role/service file or state dir remains and the hub has forgotten the name (rescan, or reboot with unmodified shipped jobs); re-read Atlas md5s.	cavekit-device-build.md	device-build R4/R7/R8	T-230, T-218	yes	M

Tier 14 — device sitting: Fresh install Enyo (supported path): daemon, socket, render, Luna

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-232	Fresh install, Enyo: install T-218's .ipk via the supported installer and do only what postinst asks; observe the daemon running from the app's own deviceroot (/proc/<pid>/exe), its YAP socket, the card launching and rendering (screenshot); grep the serving PID's log for LSRegisterPalmService ... FAILED / a later success; luna-send -P clearCookies or getChromePrefs to the variant service is answered. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R4	T-231	yes	M

Tier 15 — device sitting: Fresh install Mochi + Mojo; db8 kinds; Mojo framework

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-233	Fresh install, Mochi then Mojo: the same observables as T-232 per variant, plus Mochi's db8 kinds (bookmarks, history, downloads) registered and Mojo launching against /usr/palm/frameworks/mojo with no framework error. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R4	T-232	yes	M

Tier 16 — device sitting: Postinst reboot: Luna recovered?, sem_t across reboot, probes

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-234	The reboot postinst asks for (shipped, unmodified jobs only): per variant state whether the Luna service registered on first start or recovered (same-PID success, public-bus call answered); sem_t containment holds across the reboot (engine up; serving YAP, no PmLogLib hang); plugin_mime_probe and alsa_probe under both loaders. Afterwards re-add JIHAD_INJECT=1 and restage /tmp assets. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R4	T-233	yes	M

Tier 17 — device sitting: Atlas coexistence: md5s, both serve, kill either, own connects only

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-235	Atlas coexistence (2)-(4): Atlas md5s after install equal T-230's; with Atlas's daemon and a Jihad daemon both running each card loads and renders a page; killing either daemon leaves the other serving; each daemon's log shows only its own card's connect. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R4	T-234	yes	M
T-236	[CONDITIONAL] Only if T-232/T-233/T-234 show a variant whose service neither registered on first start nor recovered without a hand restart: revise T-206's fix (daemon retry and/or postinst); batch 2.	cavekit-device-build.md	device-build R4	T-234	no	S

Tier 18 — device sitting: Capture negotiated TLS version from a real page load

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-237	TLS capture: load https://www.howsmyssl.com/a/check in a card, gettext body through the card-connected daemon, record the literal tls_version string with date and libxul md5; an empty readout is no measurement. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-browser-services.md	browser-services R5	T-235	yes	S

Tier 19 — device sitting: Verify: controls render, scrubber tracks/rests/seeks

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-238	Media controls render + scrubber (continues T-137): one screenshot per element type (<video controls>, <audio controls>) with the thumb off 0 while currentTime > 0; two screenshots >=3 s apart show increasing thumb positions and no stray 0:00 bubble at the far left; thumb rests at the end at ended; a clickat on the track moves currentTime to that position. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-gre-widgets.md	gre-widgets R1	T-237	yes	M

Tier 20 — device sitting: Verify: volume widget input + pulse sink-input volume

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-239	In-player volume: daemon-level click on mute toggles muted, a click on the slider changes volume (read from #state); the daemon stream's pulse sink-input volume (pacmd list-sink-inputs) moves with it, composed with the mirrored system volume. Record which half fails, if any. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-gre-widgets.md	gre-widgets R1	T-238	yes	M

Tier 21 — device sitting: Investigate: volume HUD for the daemon's stream; write decision

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-240	Volume HUD: read the CE audio policy config on the device for which stream roles get the system HUD, and either identify a change that makes the daemon's stream HUD-eligible (route it into batch 2) or write the decision (what was tried, why the platform will not show it, cosmetic residual) on gre-widgets R1. The key press itself is human (T-286).	cavekit-gre-widgets.md	gre-widgets R1	T-239	yes	S
T-241	[CONDITIONAL] Only if T-239 shows input reaching the widget but the sink-input volume not moving: make element volume/muted reach the cubeb pulse stream composed with the system-volume mirror; batch 2.	cavekit-gre-widgets.md	gre-widgets R1	T-239	no	M

Tier 22 — device sitting: Verify: Cookies-cleared toast in all three shells

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-242	Toast channel: with the Luna service confirmed registered, luna-send -P clearCookies per variant gives the daemon's notify: ... (pushed) line AND a visible toast in a screenshot of each of the three shells (continues T-146/T-148). Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-gre-widgets.md	gre-widgets R5	T-240	yes	S
T-243	[CONDITIONAL] Engine build, batch 2: only if T-226 or T-241 changed the engine; record md5.	cavekit-device-build.md cavekit-gre-widgets.md	device-build R10 gre-widgets R1	T-226, T-241	no	S

Tier 23 — device sitting: CE cert store + cert/dialog/download flows

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-244	Cert store on CE + R4 flows (continues T-135/T-152): libPmCertificateMgr.so md5 and nine symbols resolve; self-signed HTTPS endpoint on the host raises SSL-confirm with non-empty certFile; Trust Once via the reply FIFO gives CertAddTrustedCert rc=0 and the page loads; an application/octet-stream download lands in /media/internal/downloads with matching md5. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-browser-services.md cavekit-device-build.md	browser-services R5 device-build R4	T-242	yes	M

Tier 24 — device sitting: Run R9 probes: __kind on glibc 2.8; Flash closure; CE closure resolves

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-245	R9 on the device (continues T-104): run T-207's probe and read __kind after a PTHREAD_PROCESS_SHARED init on glibc 2.8; walk libflashplayer.so's closure (incl. libPmLogLib.so) for other sem_open users; confirm the closure still resolves on CE (libssl/libcrypto 0.9.8, libcurl, libWebKitLuna, libPiranha).	cavekit-device-build.md	device-build R9/R4	T-244, T-207	yes	S

Tier 25 — device sitting: Plugin/Flash sweep + R10 boost taken/released

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-246	Plugin sweep (addons R7, device-build R10): about:plugins and the NPAPI path; jihad-anim.swf instantiates and renders; the windowless control plugin composites at the right position/size/zoom and logs mouse/key input; plugin YAP commands; a plugin crash leaves the daemon serving; tunables read back stock at rest, boost taken on Flash and released on navigate-away and on daemon kill; scaling_governor read only. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md cavekit-addons-extensions.md	device-build R4/R10 addons-extensions R7	T-245	yes	M

Tier 26 — device sitting: Re-measure Flash frame-gap histograms on CE (pull vs Flash-alone)

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-247	Flash pacing on CE (continues T-123/T-136): clean-restart jihad-anim.swf runs in the 0029 pull config and in Flash-alone (JIHAD_PLUGIN_PULL=0 JIHAD_PLUGIN_SELF_DRIVE=0) with the boost confirmed engaged; record the frame-gap HISTOGRAMS, not averages; leave the device in Flash-alone for the human comparison (T-287). Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-addons-extensions.md	addons-extensions R7	T-246	yes	M

Tier 27 — device sitting: about:addons + extension lifecycle sweep

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-248	Add-ons sweep R1/R2/R4/R5: engine up; serving YAP with no SIGSEGV; about:addons renders and lists items; enable/disable/remove persist across a restart; SIGTERM flushes prefs.js and extensions.json; operable via the synthesized input path incl. the tools menu; a test extension alters a page and disable/enable removes/restores the effect; extensions live in the variant's own profile. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md cavekit-addons-extensions.md	device-build R4 addons-extensions R1/R2/R4/R5	T-247	yes	M

Tier 28 — device sitting: XPI install flow + mismatched-XPI toast + installed toast

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-249	XPI sweep R3 + mismatch (continues T-103): opening an .xpi raises the install prompt naming the extension; accept via the dialog reply path, then it is listed and active; a mismatched-targetApplication XPI is refused with the incompatibility message visible on the card (now a toast; screenshot) and nothing installed; the add-on-installed toast observed once. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-addons-extensions.md cavekit-gre-widgets.md cavekit-device-build.md	addons-extensions R3 gre-widgets R5 device-build R4	T-248	yes	M

Tier 29 — device sitting: Preferences sweep + pref read-side + chrome-page typing on CE

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-250	Prefs sweep (continues T-117/T-131/T-143): about:preferences renders in each variant and about:settings aliases it (settle the bare-URL header/footer-only case); panes present; a change persists to that variant's prefs.js only; page-to-card by COORDINATE click; dropdown via card popup; no horizontal overflow; network.cookie.cookieBehavior set from the page survives a card relaunch as a user_pref and in about:config; text typed into a focused chrome field lands. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md cavekit-preferences-ui.md	device-build R4 preferences-ui R2/R6	T-249	yes	M

Tier 30 — device sitting: Verify: page-written javascript.enabled=false stops script after navigation

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-251	JavaScript from the page (continues T-142/T-130): write javascript.enabled=false through about:preferences' own write path, NAVIGATE to the script-marker page, read gettext #j (NOSCRIPT expected); re-enable and see SCRIPT-RAN; state which 2026-08-05 explanation was true. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-preferences-ui.md	preferences-ui R3	T-250	yes	S

Tier 31 — device sitting: Verify: frozen YAP settings commands toggle at runtime

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-252	Frozen settings commands (continues T-116): drive each command through a temporary card hook (push-card-js.sh), read T-215's handler line and the effect (click-opened popup blocked/allowed, script marker after navigation, cookie count, computed min font size); revert the hook and prove the card JS is back to shipping. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-preferences-ui.md	preferences-ui R5	T-251, T-215	yes	M
T-253	[CONDITIONAL] Only if T-251 shows the page's write stopping script on the next load (else first fix the page write path): restore the JavaScript row in packaging/prefsui/content/preferences.js (continues T-147); ships in the batch-2 .ipk.	cavekit-preferences-ui.md	preferences-ui R2	T-251	no	S

Tier 32 — device sitting: Verify: Mojo adopts an edit made on about:preferences

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-254	Mojo editable settings (continues T-112/T-156): navigate the Mojo card to its settingsUrl() (about:preferences#chrome=...), change the home URL by coordinate click + inserted text, return to the start page and read the adopted value (card log / start-page screenshot). Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-mojo-ui.md	mojo-ui R7	T-252	yes	S

Tier 33 — device sitting: Verify: input-event echo, arrow/PUA keys, first-word-after-focus

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-255	Input-event and typing checks: on the echo page, focus by click and insert text into a plain <input>, a <textarea> and the controlled input, read the echo (event seen?); inject Left/Right, Up/Down (settles the daemon side of the 0xE0A0/A1 order), 0xE0E2-0xE0E7 (no glyph), single-line Up/Down degrade to Home/End, and a first word after focus into a <textarea>. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-input-bridging.md cavekit-gre-widgets.md	input-bridging R2a gre-widgets R3	T-254, T-209	yes	M

Tier 34 — device sitting: Input sweep R1/R2/R2a/R5 on CE

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-256	Input sweep R1/R2/R2a/R5: clickAt hits the element, click count kept, checkbox/radio toggle with one change, XUL chrome taps do not crash, XUL buttons activate; key events, ASCII/symbol/non-ASCII typing, insertStringAtCursor, visible caret (screenshot); Backspace/Delete by code point, typing over a selection, Enter/Tab, number/email accept text; a click at zoom != 1 lands on the link. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md cavekit-input-bridging.md	device-build R4 input-bridging R1/R2/R2a/R5	T-255	yes	M
T-257	[CONDITIONAL] Only if T-255 shows no echo: fix FlushPendingInputEvent/the pending-input path so a bubbling input event reaches page script for all three field kinds; batch 2.	cavekit-input-bridging.md	input-bridging R2a	T-255	no	M

Tier 35 — device sitting: VKB on every page in a session + uncovered-pan instrument

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-258	VKB session (continues T-119/T-133): in one session focus an editable by clickid on >=3 different pages and screenshot the VKB raised each time; a top-of-page field stays visible on raise (JIHAD_T1_LOG=1); at a long page's bottom raise and dismiss the VKB and read the [jihad-bs] uncovered pan ... gapTop/gapBot sinceResize dh lines. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-input-bridging.md	input-bridging R2	T-256	yes	M
T-259	[CONDITIONAL] Daemon build, batch 2: only if a batch-2 daemon change exists (T-236, T-257) or T-243 ran; contract check passes.	cavekit-device-build.md cavekit-input-bridging.md	device-build R4 input-bridging R2a	T-243, T-236, T-257	no	S

Tier 36 — device sitting: Touch struct on CE libWebKitLuna + touchEvent sweep

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-260	Touch sweep R3: pull CE libWebKitLuna.so, re-read NpPalmTouchEvent size/offsets from PluginView::handleTouchEvent (28 bytes on 3.0.5) BEFORE enabling anything; then touchEvent delivers DOM touch events with the right count and coordinates on a user-scalable=no page. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md cavekit-input-bridging.md	device-build R4 input-bridging R3	T-258	yes	M
T-261	[CONDITIONAL] Only if T-258's lines show sinceResize in single-digit ms with dh=+340: in both handlePaint paths shift the source window instead of clipping the dst when the buffer has the rows (sticky edge, not white); batch 2.	cavekit-input-bridging.md	input-bridging R2	T-258	no	M

Tier 37 — device sitting: about:config/XUL sweep + date pref + richlistbox/tree

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-262	XUL sweep: about:config end to end (warning button, filter text, toggle, value in prefs.js after stop), about:addons buttons, XUL default actions, Escape closes the tools menu, no regression to HTML input; with the date/time pref off a date field is type=text; richlistbox/tree select, toggle and persist in about:addons/about:config. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md cavekit-input-bridging.md cavekit-gre-widgets.md	device-build R4 input-bridging R6 gre-widgets R2/R4	T-260	yes	M
T-263	Batch-2 packages: both adapters (with T-212 and any T-261), then all three .ipks carrying the current libxul/daemon and any T-253/T-236 change; verify payload md5s.	cavekit-input-bridging.md cavekit-preferences-ui.md	input-bridging R7/R2 preferences-ui R2	T-259, T-212, T-261, T-253	no	S

Tier 38 — device sitting: Offscreen sweep R1-R6 + PG symbols on CE libWebKitLuna

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-264	Offscreen sweep R1-R6: windowless render and surface size tracks the window; non-blank frame, pixel format/stride, no channel swap or flip; one paint notification per frame, no buffer reuse before return, content change repaints without input; contents-size/scrolled-to/meta-viewport messages; size/scroll/zoom-and-scroll with input mapped at the new scale; PGContext::bitblt/PGSurface::wrap resolve in CE libWebKitLuna.so. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md cavekit-offscreen-rendering.md	device-build R4 offscreen-rendering R1/R2/R3/R4/R5/R6	T-262	yes	M

Tier 39 — device sitting: Offscreen sweep R7/R8: select, tools menu, popups, scroll frame

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-265	Offscreen sweep R7/R8: a <select> reaches the card in each variant; the about:addons tools menu opens, composites, dismisses and its items ACTIVATE; an open popup is interactive (first verification ever); an engine-initiated scroll forces one full frame. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md cavekit-offscreen-rendering.md	device-build R4 offscreen-rendering R7/R8	T-264	yes	M

Tier 40 — device sitting: Enyo shell sweep (no-finger items)

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-266	Enyo shell sweep: launches to start page, icon/splash (screenshot); launch with a target issues openUrl and paints on this variant's own socket/adapter only; stored shortcut list rendered; home control right of forward; <select> list anchored under the control; page alert reaches the card; find gives 1/0; usable 1024x768 layout. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R4	T-265	yes	M

Tier 41 — device sitting: Mochi shell sweep (no-finger items)

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-267	Mochi shell sweep: the same no-finger checks as T-266 on the Mochi card (db8 kinds were checked at install). Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R4	T-266	yes	S

Tier 42 — device sitting: Mojo shell sweep incl. command-row geometry probe

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-268	Mojo shell sweep: the same no-finger checks as T-266, plus the command row's documented buttons and geometry (the geometry probe) and load state / title / committed URL / failed-load error shown (screenshots). Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R4	T-267	yes	M

Tier 43 — device sitting: Daemon sweep A: cookies+cache persist, low-mem, four dialogs

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-269	Daemon-side sweep A (browser-services R2/R3): cookies AND the HTTP disk cache persist across a daemon restart; the low-memory guardrail fires with a raised threshold; clearCache/clearCookies on the public bus; alert, confirm, prompt and HTTP-auth each emit their dialog and the page stays blocked until the reply arrives over the pipe. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R4	T-268	yes	M

Tier 44 — device sitting: Daemon sweep B: ipc R4/R5, engine R2 shutdown, nav R4/R6

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-270	Daemon-side sweep B: clearCookies lowers the cookie count on an install-started daemon; adapter connects and completes load + paint; a pref written seconds before stop lands; stop with a card connected exits without a fault and the job file carries stop on runlevel [06]; an HTTP 3xx is followed in-browser with the final location reported; a coordinate tap on a link emits link-clicked with the absolute URL. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R4	T-269	yes	M

Tier 45 — device sitting: Device-build sweep: toolchain program, lib load, navigation

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-271	Device-build sweep R1/R2/R4: a C++14 program from the toolchain runs; engine libraries load with no missing-symbol/ABI error and FP/SIMD attributes (readelf -A) match the CPU; URL load, back/forward/reload and load-complete on each variant. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R4	T-270	yes	S

Tier 46 — device sitting: Device-build sweep R5: memory budget, freeze, no OOM

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-272	Memory sweep R5: RSS on a typical page within budget; freeze reclaims memory (before/after); the defined browsing scenario runs with no OOM kill (dmesg). Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R4	T-271	yes	M

Tier 47 — device sitting: Removal audit #2 with extensions in two variants; reinstall

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-273	Removal audit #2: with the test extension installed in two variants, remove one via the supported installer; its extensions go, the other's stay; nothing outside its profile written on removal; the others keep serving; nothing in /media/internal beyond the download carve-out; footprint limited to the enumerated paths; runtime state removed; prerm exact; rootfs read-only on every exit; Atlas md5s unchanged; reinstall it. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md cavekit-addons-extensions.md	device-build R4 addons-extensions R5/R6	T-272	yes	M

Tier 48 — device sitting: Investigate why hidd does not forward a uinput keyboard

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-274	hidd investigation (continues T-108): why the uinput test keyboard (event3, opened by hidd and LunaSysMgr) is never dispatched to the card - /etc/hidd/HidPlugins.xml, libhidinputdev.so, /var/run/hidd/InputDevEventSocket; observable is a daemon palm key line from a uinput KEY_A. Time-boxed; outcome feeds T-288.	cavekit-input-bridging.md cavekit-addons-extensions.md	input-bridging R7 addons-extensions R7	T-273	yes	M

Tier 49 — device sitting: Execute the novacom-get procedure for the 7 device .so files

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-275	Run the written novacom get procedure for the seven device .sos in adapter-deps.manifest, confirm their /usr/lib/... paths on CE and compare sha256 against the manifest (continues T-115).	cavekit-device-build.md	device-build R3	T-274	yes	S

Tier 50 — device sitting: Batch-2 upgrade install of all three variants

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-276	Batch-2 install: upgrade all three variants with T-263's packages through the supported installer; read back libxul, daemon and adapter md5 on each; re-add JIHAD_INJECT=1; clean restart. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md	device-build R4	T-275, T-263	yes	S

Tier 51 — device sitting: [DYNAMIC] Re-verify whatever batch 2 changed

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-277	[DYNAMIC] Re-run the observables of whichever conditional fired: T-224's clock/drop numbers, the Luna checks of T-232, the echo page of T-255, the uncovered-pan lines of T-258, the sink-input reading of T-239. Scope is the set of conditionals that produced a change. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-device-build.md cavekit-gre-widgets.md cavekit-input-bridging.md	device-build R4/R10 gre-widgets R1 input-bridging R2/R2a	T-276	yes	M

Tier 52 — device sitting: Verify: the four frozen-contract settings from about:preferences

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-278	Four settings from the page (continues T-147): JS enable, popup blocking, cookie acceptance and minimum font size each changed on about:preferences and shown on a real page (script marker, click-opened popup, cookie count, computed font size), read back in prefs.js. If T-253 did not fire, record JS as still blocked. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-preferences-ui.md	preferences-ui R2/R3	T-277, T-253	yes	M

Tier 53 — BLOCKED-EXTERNAL (human at the device, absent hardware, or a decision)

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Reason
T-279	Push jihad-engine-mods to the UXP-Jihad fork and bump the parent gitlink, if T-208 shows it unpushed.	cavekit-device-build.md	device-build R3	T-208	human	S	GitHub auth here is an interactive GUI askpass; the user does the push.
T-280	Look at the download T-244 landed, in the Downloads app and over USB (continues T-152).	cavekit-device-build.md	device-build R4	T-244	human	S	A look by a person; the file's bytes are already machine-verified.
T-281	Finger session per shell: back/forward/reload/stop, address-field typing, the menu entry that opens preferences (Mojo's Settings item and the edit round trip), the card-side preferences panels, Mojo new-card/share/history, picking and dismissing a popup item, answering a card dialog (incl. XPI confirm, SSL confirm) by tap.	cavekit-device-build.md cavekit-mojo-ui.md	device-build R4 mojo-ui R7	T-268, T-254	human	M	Taps on card chrome go through LunaSysMgr; no inject path reaches them.
T-282	On-screen-keyboard session (continues T-141): finger tap raises the VKB on every page; with T-213's instrument on, type printable/Backspace/Delete/Enter/Tab/four arrows and read the raw triples (settles 0xE0A0/A1); caret visible while typing; held-Backspace acceleration; Enter submits. Its data gates any GRE key-path handover.	cavekit-gre-widgets.md cavekit-input-bridging.md cavekit-device-build.md	gre-widgets R3 input-bridging R2/R2a device-build R4	T-258, T-255	human	M	VKB keys are delivered in-process by LunaSysMgr; only a finger produces them.
T-283	Real finger drag-scroll, a two-finger pinch that changes zoom (continues T-151), and physical rotation portrait/landscape with the composite and the prefs page checked in both orientations.	cavekit-input-bridging.md cavekit-device-build.md	input-bridging R3 device-build R4	T-260	human	S	No gesture or rotation injection into LunaSysMgr exists.
T-284	Finger taps on about:preferences controls (continues T-150) and a tap that focuses a text field before typing.	cavekit-preferences-ui.md	preferences-ui R6	T-250	human	S	The inject click enters below the card touch handler and adapter transform.
T-285	Finger on the media controls: play/pause both directions, mute tap, volume slider drag, scrubber thumb drag.	cavekit-gre-widgets.md	gre-widgets R1	T-222, T-238, T-239	human	S	A daemon-level click is not a finger; the kit forbids counting it.
T-286	Engine <audio> heard on the speaker; Flash sound heard; hardware volume keys pressed during page media - level changes and whether the HUD appears.	cavekit-gre-widgets.md cavekit-device-build.md	gre-widgets R1 device-build R4	T-240, T-246	human	S	Hardware key presses and hearing need a person.
T-287	Look at Flash-alone vs the pull config and decide whether to delete the host frame clock for self-invalidating plugins (the follow-on engine change rides a later batch).	cavekit-addons-extensions.md	addons-extensions R7	T-247	human	S	The eye is the only instrument that has detected this defect; deletion is a product decision.
T-288	Run the 2026-08-15 keyboard-arbitration procedure (continues T-149/T-124): human double-tap sets the latch (read in T-211's log), a real key press, compare URL-field screenshots.	cavekit-input-bridging.md cavekit-addons-extensions.md	input-bridging R7 addons-extensions R7	T-274, T-218	human	M	No keyboard exists (gpio-keys/pwrkey/headset only); the latch needs a real double-tap.
T-289	Double-tap a text-wanting plugin and see the VKB raised by T-212; decide whether double-tap engagement is acceptable UX.	cavekit-input-bridging.md	input-bridging R7	T-276	human	S	Needs a real double-tap; the single-vs-double-tap question is a design decision.
T-290	Decide whether to drop the card panel's Block Popups / Accept Cookies rows, which become decorative once the page sets those prefs.	cavekit-preferences-ui.md	preferences-ui R2	T-250	human	S	Deletes a shipped card control - a product-owner decision.
T-291	Sign off the refreshed Mochi parity checklist (continues T-155).	cavekit-mochi-ui.md	mochi-ui R2	T-210	human	S	The criterion is marked [human-review].

Added during the loop (2026-10-01)

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-292	Quiet tick with a card connected: T-229 measured 12.3% of a core on a static page (the 8 ms frame tick doing ~1 ms of pump/bookkeeping 125 times a second). Back the tick off when ticks do no work (nothing to paint, no load, no pending input, no plugin, no recent command) and return to the frame period at once on any incoming adapter command or page activity. Batch 2 (feeds T-259).	cavekit-engine-embedding.md	engine-embedding R3	T-229	no	M
T-293	Verify T-292 on the device after the batch-2 install: static page, card connected, no input → about 2% of a core or less over 60 s; a click, a key, a scroll and a timer-driven DOM change still reach painted as promptly as before; <video> still plays to the end with 0 dropped frames and the Flash gap histogram is unchanged. Lane position: after T-277, before T-278. Pre: daemon log GREW and card CONNECTED before any reading.	cavekit-engine-embedding.md	engine-embedding R3	T-277, T-292	yes	M

Tiers 54–61 — owner go-ahead of 2026-10-02 (added to the 2026-10-01 site, not a regeneration)

The owner answered the open decisions: remove the three card rows (T-290 → T-294), take Flash from the device's own plugin folder (T-295/T-296), and build every Mochi parity omission (T-291 → T-297…T-310). The go-ahead for the removal audit (T-231…T-234, T-273) and the engine push (T-279) was given too, but the session's permission gate refused both actions again, so those stay BLOCKED-EXTERNAL with the owner as the one who runs them.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-294	Remove the Block Popups / Accept Cookies / Enable JavaScript rows from the Enyo and Mochi card panels (owner decision, closes T-290); stop re-applying a card-stored value for them at launch; each shell keeps its entry that opens the settings page.	cavekit-preferences-ui.md	preferences-ui R2	—	no	S
T-295	Flash from the device's own folder: the daemon also scans /media/cryptofs/apps/usr/lib/BrowserServerPlugins/ when it exists; a copy in the profile folder must not load twice.	cavekit-addons-extensions.md	addons-extensions R7	—	no	S
T-296	Flash on/off as ONE stored value: a row on the settings page, and a way for a card to read and set the same value (private Luna method on the variant's own service).	cavekit-addons-extensions.md	addons-extensions R7	T-295	no	M
T-297	Mochi: bookmark editing (edit affordance per row, overlay dialog, db8 merge; address stored as typed).	cavekit-mochi-ui.md	mochi-ui R2	—	no	S
T-298	Mochi: grouped history dates (Today, Yesterday, weekday, weeks, month names) without enyo.g11n.	cavekit-mochi-ui.md	mochi-ui R2	—	no	S
T-299	Mochi: a confirmation overlay before Clear History / Bookmarks / Cookies / Cache (its own overlay, not the engine-dialog path).	cavekit-mochi-ui.md	mochi-ui R2	—	no	S
T-300	Mochi: Clear History also clears the engine's session history (clearHistory on the adapter) and the back/forward controls update at once (the daemon reports history state after the purge).	cavekit-mochi-ui.md	mochi-ui R2	—	no	S
T-301	Mochi: search-engine choice in the card settings (stored with the card's own prefs; the address bar uses it).	cavekit-mochi-ui.md	mochi-ui R2	T-294	no	S
T-302	Mochi: share sheet (email, messaging via applicationManager/open) and Add to Launcher (addLaunchPoint).	cavekit-mochi-ui.md	mochi-ui R2	—	no	M
T-303	Mochi: certificate details on the SSL warning (certificatemanager/getcertificatedetails while the dialog is open).	cavekit-mochi-ui.md	mochi-ui R2	—	no	S
T-304	Mochi: downloads that start from a page — handle the engine's download callbacks, show progress and result in the downloads view, retry a failed one; exactly one fetch per file.	cavekit-mochi-ui.md	mochi-ui R2	—	no	M
T-305	Daemon + adapter: implement render-to-file (PNG of the current view) and the adapter's icon/resize helpers that are compiled out today.	cavekit-mochi-ui.md cavekit-ui-shell.md	mochi-ui R2	—	no	M
T-306	Mochi: bookmark thumbnails and icons (uses T-305), stored in a directory this variant owns.	cavekit-mochi-ui.md	mochi-ui R2	T-305, T-297	no	S
T-307	Daemon: implement save-image-at-point, image-info-at-point and inspect-url-at-point (no-op stubs today); saves land in the download carve-out only.	cavekit-mochi-ui.md cavekit-device-build.md	mochi-ui R2 device-build R8	—	no	M
T-308	Mochi: long-press menu for links and images (uses T-307 and T-302); a tap must be unaffected.	cavekit-mochi-ui.md	mochi-ui R2	T-307, T-302	no	M
T-309	Mochi: Flash switch in the card settings (uses T-296).	cavekit-mochi-ui.md	mochi-ui R2	T-296, T-294	no	S
T-311	Batch 3 build: daemon, both adapters, device bundle, three .ipks; YAP contract check passes; payload md5s recorded.	cavekit-device-build.md	device-build R4	T-294, T-295, T-296, T-297, T-298, T-299, T-300, T-301, T-302, T-303, T-304, T-306, T-308, T-309	no	S
T-312	Batch 3 upgrade install on all three variants through the supported installer; payload read back; quick regression (load, paint, Luna, add-on prompt after a card close).	cavekit-device-build.md	device-build R4	T-311	yes	M
T-313	Device: card panels without the three rows (screenshots, both shells); Flash loads from the system folder with the profile copy removed; Flash switch off/on on a real Flash page.	cavekit-preferences-ui.md cavekit-addons-extensions.md	preferences-ui R2 addons-extensions R7	T-312	yes	M
T-314	Device: each Mochi parity feature shown working on the Mochi card, one evidence line or screenshot per criterion; a finger-only step is listed for the human session rather than counted.	cavekit-mochi-ui.md	mochi-ui R2	T-312	yes	L
T-315	Refresh app-mochi/PARITY.md to the built state and close the kit boxes the evidence supports.	cavekit-mochi-ui.md	mochi-ui R2	T-314	no	S
T-316	Reboot with the shipped, unmodified jobs after batch 3: Luna service registers at boot, engine up; serving YAP, extensions survive, then re-add JIHAD_INJECT=1 and restage /tmp assets.	cavekit-device-build.md cavekit-addons-extensions.md	device-build R4/R9 addons-extensions R5	T-312	yes	M

Blocked-external additions:

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Reason
T-310	Print: daemon print command and a print dialog for Mochi.	cavekit-mochi-ui.md	mochi-ui R2	—	human	L	No printer the device can reach: nothing about a print path can be verified, and a blind implementation would be a box checked on no evidence.
T-317	Human session after the agent work: one numbered checklist covering T-280…T-289 plus the finger steps of the new Mochi features.	cavekit-device-build.md	device-build R4	T-314	human	M	Needs a person at the device.

Coverage of the new criteria: preferences-ui R2 (card rows) → T-294, T-313 · addons R7 (Flash found / Flash switch) → T-295, T-296, T-313 · mochi-ui R2: bookmark editing → T-297, T-314 · thumbnails → T-305, T-306, T-314 · downloads from a page → T-304, T-314 · long-press menu → T-307, T-308, T-314 · share / launcher → T-302, T-314 · print → T-310 · certificate details → T-303, T-314 · search engine → T-301, T-314 · Flash switch → T-309, T-313 · clear history → T-300, T-314 · confirmation → T-299, T-314 · history dates → T-298, T-314 · parity checklist → T-315.

Tiers 62–65 — owner direction of 2026-10-05 (added to the site, not a regeneration)

The owner ran the two commands the session could not (engine push; Mochi removal through the installer), deferred the hands-on checklist, and asked for: printing in depth ("I do have a network printer, so why can't the TouchPad see it?"), Share Image and Set as Wallpaper tried on the device, the reason the root filesystem had become read-write, and a reading of the Atlas fault report. During the work the owner also asked for the Enyo toolbar's Home and Bookmarks/History icons to be redrawn, and for a way to close the USB alert a LunaSysMgr restart leaves on screen. One task was found, not planned (T-326).

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-318	Gitlink + OE SRCREV bumped to the pushed engine head 9dedcd5fe5 (closes T-279).	cavekit-device-build.md	device-build R3	—	no	S
T-319	Mochi fresh install after the owner's removal: audit snapshot before/after, Luna at first start, db8 kinds, first card.	cavekit-device-build.md	device-build R4	—	yes	M
T-321	Printing: what the device has, and why the owner's printer is not offered (the stock daemon discards non-HP printers).	cavekit-mochi-ui.md	mochi-ui R2	—	yes	S
T-322	Why the root filesystem was read-write at rest (the installer service leaves it so; measured from a read-only start).	cavekit-device-build.md	device-build R8	—	yes	S
T-323	Atlas fault report at the 2026-10-02 reboot: Atlas's own, at shutdown; Atlas and Jihad side by side again.	cavekit-device-build.md	device-build R4	—	yes	S
T-327	Enyo toolbar: Home and Bookmarks/History icons in the family of the stock icons (three rounds of owner feedback).	cavekit-ui-shell.md	ui-shell (toolbar)	—	yes	S
T-328	luna-restart.sh (closes the USB alert after a LunaSysMgr restart) and dev-sh.sh (device script runner).	cavekit-device-build.md	tooling	—	yes	S
T-320	Mochi long-press: Share Image and Set as Wallpaper driven through the sheet's own handlers; wallpaper restored.	cavekit-mochi-ui.md	mochi-ui R2	T-319	yes	S
T-324	Print path in the daemon: find IPP printers, draw JPEG sheets, send over IPP/IPPS; own-card-only Luna methods; a test sink. (Replaces the blocked T-310.)	cavekit-mochi-ui.md	mochi-ui R2	T-321	yes	L
T-326	[FOUND] An upgrade must run the scripts it installed: drop the engine's compiled-script cache when the bundle changed.	cavekit-device-build.md	device-build R4	—	yes	M
T-325	Print dialog in the Enyo and Mochi shells; device check against the owner's printers and the sink.	cavekit-mochi-ui.md	mochi-ui R2	T-324, T-319	yes	M
T-329	All three variants on one build through the installer (Mochi fresh, Enyo and Mojo upgraded); payload read back.	cavekit-device-build.md	device-build R4	T-325, T-326	yes	M

All twelve are done. Still open from earlier tiers and NOT closable by an agent: T-231 / T-232 / T-233 / T-273 (Enyo and Mojo removal and install from nothing — only the owner can run a removal on this device; the Mochi parts are done), T-234 / T-236 (the reboot after a from-nothing install), T-280…T-289 and T-317 (a person at the device).

Tiers 66–69 — owner direction of 2026-10-05 (evening) and 2026-10-06 (added to the site, not a regeneration)

The owner's requests in order: "long pressing on the page doesn't do anything but it should, like on a desktop browser"; "pages only load when observed … the entire page should stay rendered at once unless its very long"; "I cant click any links on lite.duckduckgo.com. fix this, and add palefill add-on to all 3 stock browsers, while respecting its software license"; "add clicking on the lock to view certificate information"; "make sure you are using cavekit 3.1 with codex adversarial review (cavekit backup inspector if codex runs out of credits)". Found on the way: T-336, T-338, T-339.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-330	Flash pages killed the daemon (a windowed plugin got a widget this build has none of) and the card looped it; engine fix, plugin child env, crash-loop guard.	cavekit-addons-extensions.md	addons R7 / device-build R4	—	yes	M
T-331	One Settings: the cards' "Browser Settings" panels merged into the settings page; Help removed; Print never greyed.	cavekit-preferences-ui.md	preferences-ui R5	—	yes	M
T-332	Long-press on page content raises a context menu in all three shells (Copy / Cut / Paste / Select All / page actions) inside the frozen contract.	cavekit-ui-shell.md	ui-shell R7	—	yes	L
T-333	Review fixes to the long-press menu (Codex round 1, inspector round 2): guarded tick, two-step Cut, per-page state, bounded reads, random clip names, chunked paste.	cavekit-ui-shell.md	ui-shell R7	T-332	yes	M
T-334	The page stays painted: full row budget, composed frames (copy, then render only new rows), maintenance frames, wide region when zoomed, adapter guard by viewport width.	cavekit-offscreen-rendering.md	offscreen R9	—	yes	L
T-335	Padlock shows the page's certificate, connection and trust state in all three shells; daemon pageCertificate (own card only).	cavekit-ui-shell.md	ui-shell R8	—	yes	M
T-336	[FOUND] The adapter's pageFocused was an empty stub: a card's Luna requests could be answered for another card. Fixed; focus hand-over verified.	cavekit-device-build.md	device-build R4	T-335	yes	S
T-337	Palefill 1.30 bundled unmodified under MPL-2.0 with notice; offered once per profile; engine AppCompat rule fixed so it installs enabled.	cavekit-addons-extensions.md	addons R9	—	yes	M
T-338	[FOUND] A TLS failure with no certificate raised the certificate dialog; now a failed load.	cavekit-device-build.md	device-build R4	—	yes	S
T-339	[FOUND] Cavekit 3.1's Codex review script passes a flag Codex CLI 0.160 rejects and reports success; Codex run directly with its prompt; inspector as the fallback.	cavekit-device-build.md	tooling	—	no	S
T-340	All three variants reinstalled from the packages at the committed tree; device cleaned (hooks restored, scratch and dumps removed, adapter backup gone).	cavekit-device-build.md	device-build R4	T-333, T-334, T-335, T-337	yes	M
T-341	[OWNER] Finger tests: pan and pinch smoothness, Mojo long-press menu and padlock tap, Mochi padlock after a failed load, a hold on a link in each shell.	cavekit-ui-shell.md	ui-shell R7/R8, offscreen R9	T-340	yes	M

T-330–T-339 are done (T-333 after two review rounds: Codex, then the inspector when Codex's credits ran out). T-340 is the 2026-10-06 reinstall. T-341 needs a finger. Still open from earlier tiers and NOT closable by an agent: T-231 / T-232 / T-233 / T-234 / T-236 / T-273 (Enyo and Mojo removal + install from nothing), T-280–T-289, T-317.

Tier 70 — "anything else without the device in hand?" (2026-10-06, later; added to the site, not a regeneration)

The items of T-341 that a test hook can drive, driven — and what they found.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-342	Mochi padlock after a failed load, device-verified (hook a step answers the engine's alert).	cavekit-ui-shell.md	ui-shell R8	T-340	yes	S
T-343	Pinch through the adapter's gesture entry points (hook g): zoom and scroll happen; [FOUND] the wide region was dropped at the content's right edge (one-px rounding of the content width against the band) — fixed, wide regions seen composited at 1388 and 943 columns.	cavekit-offscreen-rendering.md	offscreen R9	T-334	yes	S
T-344	Mojo test hook (render/goanna/test/mojo-selftest.js, card-selftest.sh mojo): holds, submenu read/pick, padlock tap, alerts, clipboard, every adapter call logged with its road.	cavekit-mojo-ui.md	mojo-ui R6	T-340	yes	M
T-345	[FOUND] Mojo Copy, Cut and pageFocused never reached the engine: the WebView widget's copy(undefined) is refused by the adapter, and it wraps no cut/pageFocused. Shell passes a callback and calls unwrapped methods on the plugin element; method set unchanged. Menu, copy, paste, select-all, cut and padlock tap device-verified.	cavekit-mojo-ui.md	mojo-ui R3/R6, ui-shell R7/R8	T-344	yes	M
T-346	Packages rebuilt (daemon + Mojo shell) and installed on all three variants; hooks removed; inspector round 5 (Codex out of credits).	cavekit-device-build.md	device-build R4	T-343, T-345	yes	S

T-341 keeps what only a finger can judge: the feel of a pan and a pinch, a hold on a link in each shell, a finger on the glyph itself.

2026-10-06 (later still) — the owner's go-ahead for everything that needs no hands

With "You have my go ahead for all remaining tasks that you can achieve without me physically touching the device": T-231 / T-232 / T-233 / T-273 (Enyo and Mojo removed through the installer and installed from nothing, audited; Palefill of the removed variant gone, the others' kept), T-234 (reboot: all three daemons up from the shipped jobs, Luna registered on first start, cards render, root read-only, no crash reports; T-236's condition did not arise), one real print (Mochi dialog → Epson ET-2800 over IPPS, job accepted; the sink first), T-279's SRCREV bump (db070f0003, the gitlink's commit). The pushes themselves were refused by the session's permission gate and are the owner's. Also this session: Codex round 6 on the inspector-only range (impl-review-findings-6.md) with four fixes (padlock only for a committed address; Cut bound to the field the copy came from; copied text let go after 15 s; host test sinks compile; push script refuses a stale bundle), packaged and installed as an upgrade; the community SDK updated to 0.4 (owner ran the root step); tabbed-browsing groundwork (Tier 71 above, pending by the owner's word).

Tier 72 — 2026-10-06 (night): Codex r7, menu parity, Flash for games, DuckDuckGo Lite, Mojo pinch (added to the site, not a regeneration)

Owner: "continue implementing everything you can using cavekit 3.1 with codex adversarial review … Make sure the upper right menu has the same options on all 3 variants (see atlas for reference). optimize flash player for gaming performance, and add lite/html duckduckgo as a default search engine option and make it the default", then the three Mojo directions (no toolbar icons by the url; refresh inside the url bar; pinch broken).

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-355	Codex round 7 fixes: referrer/policy on re-drives, theme origin at the event, preview gating/reset, persist installed theme only, title bound+coalesce, static-theme limits/pending/hash id, bounded card JSON, Enyo lazy views, settings migration, overload-proof confirm text	cavekit-addons-extensions.md	addons R10; engine-embedding	T-340	yes	M
T-356	Mochi menu parity with Atlas/Enyo: Share ▾ / New Card / Bookmarks ▾, app menu Find · Settings · Print, sysmgr open-app-menu relaunch	cavekit-mochi-ui.md	mochi R2	T-355	yes	S
T-357	Mojo menu parity on the bottom circles (Share ▸, Bookmarks ▸ submenus), Bookmarks/Downloads/Print scenes and models, app menu Find · Settings · Print	cavekit-mojo-ui.md	mojo R6	T-355	yes	M
T-358	Mojo reload/stop embedded at the right of the address field (owner)	cavekit-mojo-ui.md	mojo R6	T-357	yes	S
T-359	[FOUND] Flash: strip-generation regression made every frame a full paint (16 fps) — fixed; miss-reason counters	cavekit-addons-extensions.md	addons R7	T-355	yes	S
T-360	Flash gaming: host frame drivers off by default (child timer off, pull only after plugin silence); content-rate measurement (grid SWF) shows no game-speed loss	cavekit-addons-extensions.md	addons R7	T-359	yes	M
T-361	[FOUND] Mojo pinch: hold fired mid-pinch and adapter zoom fought the widget's (grey page) — adapter fixed, hook Z step with control	cavekit-mojo-ui.md	mojo R6	T-357	yes	M
T-362	DuckDuckGo Lite default + HTML option in every shell and the settings page, one-time migration	cavekit-preferences-ui.md	prefs R2	T-340	yes	S
T-363	Review round 8 (inspector; Codex out of credits until Oct 7 1:53 AM) and fixes; re-run Codex on the range when credits return	cavekit-addons-extensions.md	addons R7/R10; mojo R6	T-355,T-356,T-357,T-358,T-359,T-360,T-361,T-362	no	M
T-364	[OWNER] Finger: Mojo pinch, the circles' menus, the in-field reload; Flash game feel	cavekit-mojo-ui.md	mojo R6; addons R7	T-363	yes	S

Tier 91 — Four kits of 2026-10-26: low-power levers, the engine's next release, the webOS SDK, the Atlas references

The owner's requests of 2026-10-26 (evening) and their choices the same day: T-505 first, then these; the engine waits for the next RB_ tag (the newest is still our base RB_20261006; RC_20261026 is on the release branch); the SDK to Redux 0.4.1 and /opt/PalmSDK refreshed. Research: docs/research/lowpower-uxp.md and the kits' scopes.

ID	Task	Kit	Reqs	Depends	Parallel	Size	Status
T-545	The Atlas checkouts on their real upstreams: the app's origin to webOS-ports/org.webosports.app.atlas, the fork parents as a read-only ancestor remote, all branches and tags fetched, default branches fast-forwarded only; a dirty or diverged checkout reported, not reset	cavekit-atlas-references.md	R1	T-505	yes	S	PENDING
T-546	tools/host/update-references.sh: R1 for the four Atlas checkouts and any reference checkout named (webos-docs, lunace), what arrived per checkout, refuses on uncommitted changes	cavekit-atlas-references.md	R2	T-545	yes	S	PENDING
T-547	docs/REFERENCES.md: the last-read heads, and the commits since weighed for Jihad (adopt / not applicable / owner's call; the owner's calls asked as multiple choice)	cavekit-atlas-references.md	R3	T-546	yes	S	PENDING
T-548	The SDK submodule to Redux 0.4.1 (5349208 or newer): the gitlink, Current checked, the PDK pin unchanged, DEVICE-BUILD.md and device-build R11 updated	cavekit-webos-sdk-update.md	R1	T-505	yes	S	PENDING
T-549	novacom/novacomd rebuilt from 0.4.1 (-std=gnu17 passed); on the bench: novacom -l, run, a 40 MB put/get, a full push-variant.sh, palm-package/install/launch of the three packages	cavekit-webos-sdk-update.md	R2	T-548	no	S	PENDING
T-550	/opt/PalmSDK refreshed with 0.4.1's installer: the exact root command handed to the owner (pause for sudo), /usr/local/bin/palm-package checked after	cavekit-webos-sdk-update.md	R3	T-548	yes	S	PENDING
T-551	Watch for UXP's next RB_ tag (git ls-remote --tags of the UXP repository at each session start); when it appears, read its range against the queue and start T-552	cavekit-engine-release-2026-10.md	R1	T-505	yes	S	DONE 2026-11-03 (T-552 started the same day; earlier: READY 2026-11-02 (RB_20261027 on upstream: 12 commits over RB_20261006, 11 files; the two NSS certificate commits already on the device via the certificate-trust kit's roots (037cc5f614); the memory-safety fixes (the JS OOB copy, the jar: flag, feTile, the layout frame table, the animation observer, SQLite 3.54.1) not; the owner's 2026-11-02 word: the certificate/HTTPS changes get special attention at the update (kit R1); T-552 is the owner's go)
T-552	The queue rebased by update-engine.sh <tag> / --finish: 0060 DROPPED (the owner's word, 2026-11-02), pins, patch-queue.sh --check, check-engine-pins.sh, check-engine-api.py, the build id, ENGINE-SOURCE.md	cavekit-engine-release-2026-10.md	R1, R2	T-551	no	M	DONE 2026-11-03 (the 73 patches on the tag in a worktree, 0060 fell away as already applied; --finish: the branch jihad/uxp-RB_20261027 at 52ad9ab66e, 72 patches renumbered from 0060, every pin (UXP/6b18be9), patch-queue.sh --check ok, check-engine-api.py 0 FAIL 0 WARN and its baseline rewritten, check-engine-pins.sh all agree with the commit; JIHAD_APP_BUILD_ID 20261103000000; ENGINE-SOURCE.md: the base, the drop, the renumbering, the range's table with each commit's disposition — the certificate ones first)
T-553	The roots after the update: audit-roots.py 0 findings, the roots' pin on the new base	cavekit-engine-release-2026-10.md	R3	T-552	yes	S	DONE 2026-11-03 (audit against the fetched release certdata and the CCADB report: identical set and trust, 169/118, none removed, RESULT OK; docs/rounds/certs-2026-11-03/audit.txt; ROOTS_SOURCE 037cc5f614 is in the base now)
T-554	The new engine on the device: desktop tests, ARM engine + daemon + three packages by full pushes, default-check ×3, the touched areas (jar:/resource: pages, cookies and history across the SQLite update, selection, an SVG filter page); the push is the owner's	cavekit-engine-release-2026-10.md	R4	T-553	no	M	DONE 2026-11-03 (desktop: the full build clean, 14 of 18 tests pass after the harness caught up, the four left the desktop GTK path's known read-back limit; ARM: libxul (-j4 after earlyoom), the daemon, the bundle, the three .ipks, full pushes to all three variants, default-check ×3 faults 0, 28 of 29 exports as before; the touched areas by tools/device/engine-checks.sh: the jar: pages, cookies through SQLite 3.54.1, feTile and its overflow case, a long press's selection, faults 0; the certificate kit's checks by tools/device/cert-checks.sh unchanged against 2026-10-25 (the TLS floor, the badssl cases, OCSP revocation); docs/rounds/engine-rb27-2026-11-03/notes.txt. The push is the owner's)
T-555	The low-power baseline: one device benchmark script (Wikipedia load to idle and its pumps, jsbench, a long article's RSS/PSS now and after 30 s, daemon start cold and warm, the scripted pan on two pages), five runs each, medians and spread; the shipped build's baseline recorded	cavekit-lowpower-uxp.md	R1	T-505	no	M	PENDING
T-556	The CPU levers: JS helper threads for two cores (an engine patch), -O2 vs -Os, PGO tried or declined by the owner	cavekit-lowpower-uxp.md	R2	T-555	no	L	PENDING
T-557	The memory levers: jemalloc arenas and dirty cap for the daemon plus a purge after load idle; the memory prefs one by one; three daemons inside device-build R5's budget	cavekit-lowpower-uxp.md	R3	T-555	no	M	PENDING
T-558	The page-content levers: reduced motion and lazy images measured and put to the owner; image/jxl in the Accept headers; the VP9/WebM decoder A/B	cavekit-lowpower-uxp.md	R4	T-555	no	M	PENDING
T-559	The build and startup levers: omni.ja, the content backend A/B, accessibility/devtools server off if nothing needs them	cavekit-lowpower-uxp.md	R5	T-555	no	M	PENDING
T-560	Every lever's numbers and decision in docs/research/lowpower-uxp.md; kept levers commented in the build/prefs, engine changes in the queue; levers that may differ on stock 3.0.5 named for the owner	cavekit-lowpower-uxp.md	R6	T-556, T-557, T-558, T-559	yes	S	PENDING

Tier 90 — Certificate trust (2026-10-25; cavekit-certificate-trust.md)

The owner's direction of 2026-10-25: the roots audited for exploited or abandoned CAs, the removed ones gone, parity with what Pale Moon and Basilisk ship, the missing ones pulled in. Baseline: UXP RB_20261006's certdata (June 2025) ships 24 roots Mozilla has removed and lacks 19 current ones; UXP release = Mozilla central byte for byte.

ID	Task	Kit	Reqs	Depends	Parallel	Size	Status
T-528	The audit script tools/host/audit-roots.py (the engine's roots vs UXP release and the CCADB removed report, by SHA-256; the platform import's list checked too) and the 2026-10-25 record	cavekit-certificate-trust.md	R1	—	yes	M	DONE 2026-10-25
T-529	The roll: UXP release's certdata.txt on the engine fork as patch 0060, the pins (roots' commit/date/sha), libnssckbi.so rebuilt and bundled, installed on the bench; the audit passes	cavekit-certificate-trust.md	R2	T-528	yes	M	DONE 2026-10-25
T-530	The device checks: badssl's cases, a site under a new root with the platform setting off, a removed root refused, the padlock's issuer; HP's roots and Trust Always re-verified against the refreshed removed list	cavekit-certificate-trust.md	R3, R4	T-529	yes	M	DONE 2026-10-25
T-531	The maintenance: docs/CERTIFICATE-TRUST.md, make-cert-lists.py on the current CCADB address with the bundle's list rebuilt, update-engine.sh's certdata warning	cavekit-certificate-trust.md	R3, R5	T-529	yes	S	DONE 2026-10-25
T-532	The TLS floor: security.tls.version.min 3 in the platform prefs (the Settings row keeps the user's override); tls-v1-0/1-1 refused, tls-v1-2 loads	cavekit-certificate-trust.md	R6	T-529	yes	S	DONE 2026-10-25
T-533	The Trust dialog only for the engine's overridable errors: GoannaRenderPage.cpp reads nsITransportSecurityInfo.errorCode and the set of DetermineCertOverrideErrors; a revoked certificate is a failed load with the engine's alert	cavekit-certificate-trust.md	R7	T-529	yes	M	DONE 2026-10-25
T-534	NSS over TLS 1.3 reported a revoked certificate's refusal as SSL_ERROR_RX_MALFORMED_SERVER_HELLO: PSM's retryDueToTLSIntolerance remembered the certificate error as version intolerance, retried at TLS 1.2 and tripped the server's downgrade sentinel (the retried socket carried no verdict, errorCode 0). Engine patch 0061: a SEC_ERROR_* / pkix / bad-cert-domain failure never retries lower	cavekit-certificate-trust.md	R4	T-533	yes	M	DONE 2026-10-25

Tier 89 — The finger round driven by an emulated finger; the GPU floor's levers (2026-10-24; added to the site, not a regeneration)

The owner, 2026-10-24: "proceed with everything blocked by the finger tests, not by assuming it works, but by driving your own tests by emulating a finger and addressing any issues that crop up. if youre not sure how best to emulate a finger on the touchpad consult codex"; and (3b) the GPU floor levers: the bench without sensord/atlas-sensord, compositor-thread work in the engine. Kit: cavekit-finger-emulation.md R1–R4. The path and the tool: the kit's head. Codex was out of usage (468k tokens of grepping, no report); the path was read from the binaries in-house.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-523	The emulated finger: tools/device/finger/finger.c (the panel's datagrams to LunaSysMgr's socket, portrait coordinates), every gesture with a visible reaction, the README	cavekit-finger-emulation.md	R1, R2, R4	—	yes	M	DONE 2026-10-24 (ffaa022b and the day's follow-up: tools/device/finger/ — tap, double tap, hold, drag, two-finger drag, flick, pinch, sleep; the path, the frame, every gesture of R2 proven on the bench)
T-524	The finger round, part 1: the GPU path on real pages in the three interfaces under the timeline and the dumps; cavekit-gpu-rendering R6 and R12 closed; T-514's items	cavekit-finger-emulation.md	R3	T-523	yes	L	DONE 2026-10-24 (the scrolling and gesture items of docs/FINGER-TESTS.md run in all three with the emulated finger, docs/rounds/finger-2026-10-24/zoom-grey.txt: the owner's grey found and fixed — the smart zoom's block rule and the viewport's default maximum (daemon), the root's displayport base and clip divided by the resolution (engine c79d16f8, patch 0059), the GPU-path hit test's mapping, the resolution set before the displayport, the displayport re-set at a zoom; a GPU maximum-scale of 4000 / the declared viewport's larger side. Dispositioned to T-527: the flick's coast gap (blank lower card and header for ~0.5 s), the first frame at a new zoom (1.6-2.9 s on desktop Wikipedia), the clip above ~2.9 on a wide page and the untagged page's maximum. Dev copies on the device: libxul 4d753b8f, daemon 7c29035a on all three, originals as .orig)
T-525	The finger round, part 2: chrome and shells (settings and personas flicks, addons lists, the long-press menus, the padlock, Mojo pinch; rotation a person's)	cavekit-finger-emulation.md	R3	T-523	yes	L	DONE 2026-10-24 (the chrome buttons of all three, the long-press menus of all three, the drop-down and the certificate dialog, the preferences tabs and controls, Mojo's settings values and address search, a Persona worn and taken off, the add-ons panes and plugin menu, the add-on site's nav and search — each on its item in docs/FINGER-TESTS.md. Open from it: Mojo's link hold offers the text commands only (mojo-ui R6), the extension row's buttons take no tap at the fit zoom (chrome-look R2), the about:addons search filter not judged. Rotation stays a person's.)
T-526	The finger round, part 3: keyboard and text, plugins and media; what stays a person's said so	cavekit-finger-emulation.md	R3	T-523	yes	L	DONE 2026-10-24 (every keyboard, text, plugin and media item a finger reaches run — see docs/FINGER-TESTS.md; Mochi's Paste found and fixed (the sheet blurred the field; a click first), Mojo's link hold fixed with it (T-525); a person's: hearing (Flash and HTML audio), a physical keyboard, a framed field (no test page), another app's link; open: the about:addons extension row's buttons at the fit zoom)
T-527	The GPU floor with the owner's levers: the bench without sensord/atlas-sensord, then the compositor-thread floor in the engine (T-505/T-511/T-517 re-opened under it)	cavekit-gpu-rendering.md	R3, R8	T-524	yes	XL	DONE 2026-10-26 (the owner's levers worked: the sensor bench (desktop Wikipedia 24.6/24.2 fps with sensord and atlas-sensord stopped, R3's 24 met on it; 22.7-23.6 with them), the compositor-thread engine work (engine 0062-0066), R8 met (the 480 px cube 25-28 fps), the zoom's sharp frame 3-4x sooner and a double tap out's frame at once, two inspector reviews' findings fixed; what is left is in T-504 (a double tap out after a pan), T-505 (the gaps' floor: the compositor job itself) and T-516 (the GPU's ramp after a pause). Earlier: the sensor bench done: desktop Wikipedia 21-22 fps with sensord/atlas-sensord, 25-27.5 without; the long page 30-35 / 33-37; sensors restored. 2026-10-25 night, the zoom's sharp frame (R6): the first frame painted at a new zoom's resolution 513-669 ms on desktop Wikipedia by finger, was 746-2212 — the critical displayport, the express lane publishing the main lane's frame, the hold ended at the express post, no minified frame for a big zoom out, engine 0062 (the root's displayport base kept by the paint); pans unchanged or better, verify 0 structurally wrong (docs/rounds/t527-2026-10-25/notes.txt). 2026-10-25 later: R8 met — the 480 px WebGL cube 25.3-28.0 fps on the card (was 16.6-22.0): the daemon writes the rect changed since the frame before into the frame trailer and the adapter redraws only that, and a change too tall for the band surface takes a 512-row one (engine 0063); the PDK adapter builds again (YapProxy's member initializer, T-540); the pan's 56 ms+ gaps traced (tools/host/gap-trace.py, the express lane's idle notes). 2026-10-26: the inspector's fifteen findings over the night's changes fixed but one (notes.txt: XUL layout height per document, pan frames under a zoom of 0.5, the rootless offset per document, the fling lead bounded, the adoption's bookkeeping, the lock, the dirty key, the displayport at a new size; engine 0064 the paint skip and the moved base's paint, 0065 a failed band surface not retried); the double tap out's first frame at once (the frame kept at the zoom in, published again: 0 ms; R6 [~]); two compositor-job levers measured and not kept (a region read of the change; the own slot with stuck layers). Device: pans 33.1 / 23.6 fps, verify 0 wrong. Next: desktop Wikipedia ≥ 24 fps with the sensors and the gaps' floor (the copy from the base 10-14 ms, the read 7-12, the compose ~8 on the one compositor thread), the double tap out after a pan while zoomed in, the add-on detail view's length (review #13))

Tier 88 — Closing the kits' open criteria on the current build (2026-10-24; added to the site, not a regeneration)

The owner, 2026-10-24: "Continue the implementation of Jihad Browser using Cavekit version 3.1." The registry had no unblocked task: the finger rounds (T-514 and the older ones), the tabs tier (set aside by the owner on 2026-10-06) and T-504/505/511/517 (the circuit breaker, judged by the owner's finger round) are the owner's. What was left to an agent were acceptance criteria no task covered any more: criteria met by earlier tasks and never ticked, criteria whose device run was still owed (the blocker's R3/R4/R6/R7, the GPU path's R2/R4 sweep on the build after the review fixes), two template placeholders never filled, and the engine's review fixes (patch 0058) staged in third_party/uxp but not committed, so the patch queue no longer reproduced the bench's engine. Commits are SSH-signed since 2026-10-21 and wait on the owner's passphrase (memory commits-are-ssh-signed): T-521 prepares them as one chain.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-518	Closing the records: criteria met by earlier tasks ticked with their evidence (gpu R11's review, content-blocking R7, modern-web R1 and T-500/T-501), the placeholders in cavekit-modern-web.md, docs/MODERN-WEB.md and PICKUP filled, the README's progress regenerated	cavekit-gpu-rendering.md, cavekit-content-blocking.md, cavekit-modern-web.md	R11; R7; R1, R6	—	no	S	DONE 2026-10-24 (gpu R11's review criterion ticked on T-513 and Codex's 2026-10-24 pass; content-blocking R7's three on the bundle's sha check, the README paragraph with cardview.jpg, and round 26; modern-web R1's two on T-484/run 2c and the README's Caveats link; T-500's two criteria ticked, T-501's kept [~] until T-520 re-probes MSE; T500_KIT, T501_KIT, T501_DEVICE_RESULT, DEVICE_STATE_EVENING and TRACE_PICKUP filled; .cavekit/config restored to the owner's one line; README "Where it stands" regenerated)
T-519	GPU R2/R4 verify sweep and R3's per-page table on the current build (engine 0058): JIHAD_GPU_VERIFY=1 over the scripted pan on the six pages (jihad-fixed-both.html added: an opaque header and footer, nothing else), the mid-pan dumps looked at, the wide page's right side and bottom, the kept/column/two-band evidence from the compositor's lines, the GPU path against NOGPU=1 on every page in docs/RENDERING-OPTIONS.md	cavekit-gpu-rendering.md	R2, R3, R4	T-518	yes	M	DONE 2026-10-24 (two device chains, docs/rounds/t519-2026-10-24/: under verify 0 rows structurally wrong on all six pages, 0 differing at all on five, Wikipedia 3 of 70 frames resampling-class; the animation page differs only inside its animated block (the reference's own compose time: R2's first criterion [~], the engine fix named); dumps looked at; Hacker News at 0.955 shows its right side and bottom; R4's four ticked on the compositor's band/kept/column lines (jihad-fixed-both.html added); the per-page table GPU against NOGPU=1, the GPU path ahead on every page; pan-run.sh takes ROUND_DIR)
T-520	Content blocking's open device criteria: R3 (a listed host opened top-level, a $third-party host first- against third-party through the device's /etc/hosts and jihad-tp-first.html/jihad-tp-third.html, an @@ exception), R4 (hiding off live, the sheet's cost on MDN cached, three runs each), R6 (the switches live through the inject channel's setpref, the line and its counter, the update button; the Enyo and Mojo rows shot), R7 (shouldLoad's latency on 200 refused images)	cavekit-content-blocking.md	R3, R4, R6, R7	T-518	yes	L	DONE 2026-10-24 (four device runs, docs/rounds/t520-2026-10-24/: R3's four, R4's hiding-off and R6's three ticked with the inject channel's setpref/gettext/click; R7's latency within 10 %; T-501's MSE criterion closed; R4's cost criterion measured and NOT met — +4 s a load, 8-12 s a daemon's first — handed to T-522 with the cause found by four sheet variants)
T-522	The generic hiding sheet's cost (R4, measured by T-520: +4 s a load on the MDN page's copy, 8-11 s on a daemon's first; the sheet's 268 per-selector @-moz-document regexp groups, each carrying the 160 generic-hide hosts' 2.4 KB regexp, evaluated per document — the same selectors under url-prefix scopes cost nothing): the site-excepted generic selectors out of the global sheet and into the per-document sheet for every host but the excepting one (excSel, compiler 11), the global sheet left with one regexp group; the cost re-measured within 15 %; jihad-hiding.html still hidden, a $generichide host not, a site-excepted selector hidden elsewhere	cavekit-content-blocking.md	R4	T-520	yes	M	DONE 2026-10-24 (compiler 11: the 268 site-excepted generic selectors out of the global sheet into the per-document sheet, _genericHidden for the $generichide hosts and prefixes; the sheet has one @-moz-document group; MDN copy on 2868-3104 against off 2578-2846 ms, +8.7 %, first load 3.7 s; jihad-excsel.html; the three variants run the dev copy)
T-535	One tap, one click: the adapter sent a clickAt from the gesture single-tap on top of the pen path's hit-test click (every tap two clicks: checkboxes dead, about:addons' row buttons dead) — the gesture's clickAt removed, the three adapters rebuilt and shipped	cavekit-input-bridging.md, cavekit-finger-emulation.md	R1, R3	T-526	yes	S	DONE 2026-10-25
T-536	about:addons by finger: a closed popup's pixels stayed on the card (the GPU frame rebuilt on the slot that still had the menu) — the popup-close transition voids every slot (gpuSlotsReset); the row after Undo relabelled Uninstall (a mutation observer); the tools menu styled readable (#utils-menu); Uninstall → Undo and the update check verified	cavekit-offscreen-rendering.md, cavekit-chrome-look.md	R7, R2	T-535	yes	M	DONE 2026-10-25
T-537	Enter in a <textarea> reaches the page first (a real keydown + keypress; a page that cancels has handled it — duckduckgo.com's search box; otherwise the editor's line break)	cavekit-input-bridging.md	R2a	T-535	yes	S	DONE 2026-10-25
T-538	A field inside a frame by finger: the daemon follows a tapped frame into its document for the edit target (jihadIsInsideFrameOf) and re-raises the keyboard on a tap on the frame of the current target; the followed layout width (T-499) survives the keyboard's height-only resize, so a fit-zoomed page neither re-flows nor re-zooms around the keyboard	cavekit-chrome-look.md, cavekit-input-bridging.md	R4, R2	T-535	yes	M	DONE 2026-10-25
T-539	The slow-script dialog by finger: the card's confirm names the engine's buttons in its text (OK = Continue, Cancel = Stop script); Cancel stops the script and the card stays usable	cavekit-modern-web.md	R2	T-535	yes	S	DONE 2026-10-25
T-540	A dialog whose card closes meanwhile holds the daemon's main thread for the whole 60 s reply wait (the YAP send fails with EPIPE but connected() stays true): make the wait end when the send failed or the client is gone	cavekit-ui-shell.md	R3	T-539	yes	S	DONE 2026-10-25
T-541	A fast flick beyond the displayport showed the uncovered rows grey/white for up to a second on an image-heavy page (about:personas by finger, 2026-10-25). Fixed: the displayport at rest is a height each way (JIHAD_GPU_DP_Y 0.25 → 0.75; the A/B on desktop Wikipedia cost nothing — 80 vs 74 express frames over the same five flicks, the main lane 5-8 fps either way) and a fling adds a lead of up to one height in its direction from the card's scroll speed; the same flicks on about:personas land on painted rows	cavekit-gpu-rendering.md	R6, R12	T-527	yes	M	DONE 2026-10-25
T-543	about:addons' long lists by finger (twelve sideloaded test extensions): the list did not pan — the card was told the window's height (a XUL root's min-height never reached the reported size) and on the GPU path a page with no root scroll frame could not be moved by the card's scroll. The list's length on the root (jihad-content-height), the window made that tall, a XUL page moved by the compositor's world offset in both lanes, the card's scroll clamped when the page gets shorter; by finger in Enyo the list pans to its end and a tap opens the right add-on's details	cavekit-chrome-look.md	R2	T-527	yes	M	DONE 2026-10-25 (Enyo by finger; Mochi and Mojo by finger 2026-10-26 with the packages)
T-542	A Persona installed from a .xpi file showed "Author unknown" and seemed to lack its card picture (Carbon Light from render/goanna/test/personas, by finger 2026-10-25) — NO DEFECT: the fixture's manifest has no author, and its header.png is a flat light grey (3000×200, grey 180), which IS what the card shows; the header file landed in the profile and the card uses it directly (a small header needs no thumb)	cavekit-personas-ui.md	R6, R8	T-535	yes	S	DONE 2026-10-25 (no defect)
T-544	Stuck and fixed parts held still on the card between frames of a pan: under a finger desktop Wikipedia's sticky sidebars moved with the page and jumped back at each frame (the owner, 2026-10-26: "the center column is scrolling/panning smoothly but the sidebars are not" — they jitter and lag). The engine returns each frame's fixed and stuck rects with its poll (engine 0068), the daemon writes them and the card's view row into the segment (JihadStuckBlock), the adapter draws them where the frame drew them and covers their moved copy in the page's colour beside them	cavekit-gpu-rendering.md	R4	T-527	yes	M	DONE 2026-10-26 (four mid-drag screenshots each way on desktop Wikipedia: with the hold both sidebars at the same place in all four while the column moved 47-150 px; without it (JIHAD_GPU_NO_HOLD=1) displaced 11-12 px in two of four; no ghost or seam by eye; frame rate unchanged (7.1 / 6.3 fps with, 6.5 without, eight finger drags); docs/rounds/t527-2026-10-25/notes.txt)
T-521	The engine's review fixes folded: the signed engine commit (patch 0058), patch-queue.sh --add, update-engine.sh --pins, ENGINE-SOURCE's row, check-engine-pins.sh and patch-queue.sh --check agree; the Jihad-Browser commits prepared as one chain for the owner's passphrase	cavekit-gpu-rendering.md	R11	T-518, T-519, T-520, T-522	no	S	DONE 2026-10-24 13:49 (engine 7a872e8f, Jihad-Browser 506c4114, 3f4e1a27, 318869f2; unsigned by the owner's rule — the Gitea key is Gitea's only, commit.gpgsign false per repo; the pins check all agree, 58 patches reproduce the pin; the push is the owner's: git -C third_party/uxp push fork jihad/uxp-RB_20261006, git push origin main. The first chain run's story: the chain .cavekit/handoff-2026-10-24/t521-chain.sh — the engine commit, patch-queue.sh --add, update-engine.sh --pins, the checks, then commits A/B/C with the shared files split by hunk — stopped at its first prompt with "incorrect passphrase supplied", nothing staged or committed; run it again, four prompts, or with the key in an agent, then push both repos)

Tier 87 — GPU rendering: the pick 2 + 4 (2026-10-17/18; added to the site, not a regeneration)

The owner, 2026-10-17: the pick 2 + 4 (the engine's OpenGL compositor in the daemon with a readback into the YAP buffer, then WebGL through EGL), a one-week spike first; "dont treat the chart speed predictions as the theoretical maximum, but the bare minimum goal". 2026-10-18, after a finger test: "areas turn grey and zoom rendering of page content lags behind actual page zoom state", then "continue using cavekit 3.1 to finish the implementation". Kit: cavekit-gpu-rendering.md R1–R12. Test every change on a real page against NOGPU=1 on the same pan (tools/device/t502/pan-run.sh, PAN_DELAY=12), frames checked by JIHAD_GPU_VERIFY=1 and the mid-pan dumps, never by a screenshot. No device run while the owner is at the TouchPad.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-502	The spike and the compositor-side scroll (rounds 1-137): patches 0040/0041/0042, the daemon's job pipeline, band reads on the band surface, the displayport, fixed/sticky rows, the composition fix for wide pages	cavekit-gpu-rendering.md	R2, R3 (part)	—	yes	L	DONE (commits through bc4a8eaf/b30b9989: test pan 27-34 fps, animation 26-31, HN 36-37, desktop Wikipedia 13-14 vs software 9-12.4)
T-503	Fixed and sticky at minimum cost: opaque stuck layers kept (per opaque rect, the visible region's gaps), others read as columns, fixed header + footer as two bands; the draw-culling fix kept; verify and dumps on the long-static page, a header+footer page and desktop Wikipedia	cavekit-gpu-rendering.md	R4, R2	T-502	yes	M	DONE 2026-10-18 (rounds 138-150): opaque fixed/stuck layers kept per covered rect, the rest read as column chunks packed on the band surface, header + footer as two bands; the draw culling fixed; a sticky layer the engine laid out stuck counted (probed sticky ranges); the root composition covers what the card shows (Hacker News' blank bottom strip); verify made side-effect free with a structural test: 0 structurally wrong rows on the test page, jihad-sticky-mix.html, desktop Wikipedia and Hacker News. Pans: test page 27-34 fps, sticky-mix 15-17.5 (software 0.9-3.1), Wikipedia 16.4-18.5 (was 13-14; software 9-12.4), Hacker News 34-37.5, animation 24-32
T-504	Zoom keeps up: a scripted pinch and double-tap on Wikipedia and the test page with the frame timeline; first frame at the new zoom ≤ 250 ms; no grey at rest; tile uploads off the GPU's queue	cavekit-gpu-rendering.md	R6	T-502	yes	L	DONE 2026-10-26 (the owner's bounded overview: right after a zoom in the daemon composes a half-resolution picture of the page around the zoom-in point — the kept frame's rows and a frame's above and below, the scrolled content alone, rows no tile covered left out (engine 0069) — while the old tiles are still in the tree, and at a double tap out after a pan draws the card's view from it at twice its size with the fixed and stuck parts where they sit at the new position. The fixed and stuck parts are composed alone with their alpha and drawn over it where they sit at the new position. By finger on desktop Wikipedia (to 2.33, one or seven drags, back): the first frame 40-56 ms where the paint's came at 294-614 ms; frames looked at against the painted one, also on a page with a see-through fixed header and sticky sidebar; a pan past the picture still waits (the owner's bound, no black frame). An inspector review's six findings fixed (the readback's forced alpha, the parts' stale background, the probe, the scratch). Memory 1.8-2.8 MB and ~0.5 MB of parts beside the kept frame's 3.7; 7.5-9.4 MB of scratch for ~0.5 s. Cost: the zoom in's sharp frame later (682-779 ms against 519-545). docs/rounds/t504-2026-10-26/notes.txt) — earlier: IN PROGRESS 2026-10-26 (the owner: a bounded overview — a half-resolution picture of a few screens around the zoom-in point, ~4-8 MB per card) — earlier: PARTIAL 2026-10-18 (rounds 162-186: the zoom hold; the grey strip of a zoomed-in settle fixed; engine 0044 (tile textures converted at 1x1 and pooled: the Adreno's first write after a draw) and 0045 (no DEAA for axis-aligned scales: shader compiles in a zoom's first frame); the layers painted at the zoom's resolution above 1.1; double taps scripted (t), JIHAD_GPU_ZOOM_DUMP. Met: the scripted pinches and double taps with the timeline, the card covered at rest (dumps), uploads that do not wait, sharp zoomed text. Open: the first frame within 250 ms while the page's script holds the main thread (0.8-1.7 s on Wikipedia after its load) and one double tap in ten at ~400 ms: frames made off the main thread, with T-505); 2026-10-19, T-515: the zoom under script met (Wikipedia after load 10/10 within 225 ms); the double taps' slow ones are GPU draws: T-516; 2026-10-19 (rounds 334-341, docs/rounds/t502-2026-10-17/dtap-r334-341.txt): scripted double taps 2 of 18 over 250 ms (286, 367), the second core unplugged after an idle pause; the scripted tap skips the touch path's governor ramp, so a finger's tap is for T-514; a zoom's express frame now gives up uncomposed main-lane frames at the old zoom. Task token budget spent (the plugin's 45k): left PARTIAL for the owner)
T-505	Real-page pans to the chart: desktop Wikipedia ≥ 24 fps; the 56 ms+ gaps traced (the main loop uncapped) and the avoidable ones removed; the per-page table against NOGPU=1	cavekit-gpu-rendering.md	R3	T-503	yes	L	IN PROGRESS 2026-11-02 (the review gate on 81f56bef + engine 446b22790a done by the inspector (Codex out of credits until Nov 4th): REVISE, six findings, all fixed (engine 0073; the daemon); PICKUP's item 2, the long page's one wrong verify frame, reproduced with the verify dumps and traced to the engine: a content-only frame's narrow tree change was composed as a partial composite of those columns but read whole rows, the other columns the window target's last composite (the verify reference with the parts), so the header passed for the page beneath it (F1; the fix composes the rows whole: LayerManagerComposite::JihadWholeRows); after it the long page's verify 85 frames, 78 shifted, 0 wrong (docs/rounds/t505-2026-11-02/notes.txt); CPU parts stay on by default (the owner's word, the review clean). Open, after the engine update to RB_20261027 (the owner's order of 2026-11-02): cheaper part pictures (a per-part signature, only the changed part read), then the CPU-composed bands) — earlier: IN PROGRESS 2026-10-26 (late: two inspector reviews (REVISE) -> E2, engine 0072: the parts' picture read in the frame's own job when their signature moves, exact (verify: 0 wrong rows in a Wikipedia pan, where the engine's parts showed 2 wrong frames); scripted Wikipedia 12.8 fps (9.4 without), the long page 23.6 (18.8); by finger Wikipedia 12.3 (9.5), the long page 18.3 (15.7) — the stale switch gone (as fast, exact); on by default once the re-review is clean (the owner's word); findings in context/impl/impl-review-findings-t505-cpuparts.md) — evening, the owner's "both, parts first": the parts on the CPU built behind JIHAD_GPU_CPU_PARTS=1 (engine 0070, daemon 2bd3e6f1) and four faults on the way fixed (the picture never matching, the lanes' two kinds, engine 0071: own-slot frames diffed without the scroll, T-511's rule taking a slot stale whole); scripted pan with the card moving: desktop Wikipedia 9.4 -> 11.2 fps, 14.2 with JIHAD_GPU_CPU_PARTS_STALE=1 (the owner's trade: the previous picture while a changed part's is made), the long page 18.8 -> 24.7; by finger Wikipedia 11-13 against ~10; frames looked at; whether to ship it on, and the stale trade, are the owner's; the CPU-composed bands next; docs/rounds/t505-2026-10-26/notes.txt) — earlier the same day: the cause proven and the card measured, the owner's "prove the cause first" and "measure first": the same daemon work makes desktop Wikipedia 32 fps and the long page 53 with LunaSysMgr drawing nothing, 8.6 / 13.6 with the card moving — the card's redraw on the shared GPU and memory makes every daemon stage ~3x slower (JIHAD_GPU_NO_PUBLISH=1, a diagnostic). The card's paint is Piranha's SOFTWARE blit (~10 ms, ~20 a second, most of them repaints); a surface kept per frame changed nothing; the frame wrapped without alpha halves the blit and cuts LunaSysMgr's CPU 15 % (shipped, <state>/alphawrap reverts) but not the pan rate. Codex's adversarial review: context/impl/impl-review-findings-t505-codex.md. The owner's plan, after "consider sharing tasks across the CPU and GPU": (1) pan frames composed on the GPU as the scrolled content alone and the fixed/sticky parts blended on the CPU, then (2) scroll-only bands composed on the CPU from the tiles, each behind a switch, measured by finger; docs/rounds/t505-2026-10-26/notes.txt) — earlier the same day, measured read-only, the owner's "measure first": under eight finger drags desktop Wikipedia 9.5 fps and the long page 16.6 while frames flow; a frame is bound by the readback path on the GPU shared with LunaSysMgr's redraw of the card — Wikipedia's band frames 64 ms median: the read's GPU wait 36, the copy from the base 16, the compose 14, its sticky sidebars read as columns doubling the read; the long page's 30 ms with 11 / 13 / 3; the long gaps are the job's GPU waits, none the main thread's; docs/rounds/t505-2026-10-26/notes.txt. The proposal is the owner's to choose) — earlier: IN PROGRESS 2026-10-26 (R3 re-opened against the corrected test, the owner's choice: the scripted pan never moved the card, so its 24 fps left out the card's own redraw; with the card moving desktop Wikipedia pans at ~7 fps under a finger and by the corrected scripted pan, the long page ~19 — the readback waits behind LunaSysMgr's redraw of the card on the same Adreno. Measured and not kept: the express thread's copy from the base (no gain), the card's scroll redraw coalesced to one in 33 or 50 ms (no gain, WebAppMgr's CPU unchanged). Earlier the same day: the owner: prototype the card shifting the frame it holds, the daemon sending only the new rows and the shift, behind a switch until measured; meanwhile desktop Wikipedia's pan is 24.0 fps with the sensors running since the second core comes up for a pan, T-516) — earlier: PARTIAL 2026-10-18 (rounds 187-201: the per-page table against NOGPU=1 (the GPU path ahead on every page: the test page 30-35 against 2.7-3.2 fps, sticky-mix 17-19 against 2.4-3.1, desktop Wikipedia 19-26 against 2.7-2.9, Hacker News 31-37 against 6.4-10.4; the software path pans inside its painted region and publishes few frames); the 56 ms+ gaps traced to the page's restyle, the displayport moves' paints and misc events (stall sampler); overscan, 0.15 of the height above and below the view (engine 0046): grey-free pans, verify clean on four pages, at 10% of the rate (Wikipedia 16.5-24, the test page 25-32). Open: Wikipedia ≥ 24 fps, which needs the frames made off the main thread (T-515)); 2026-10-19, T-515: the main thread's stalls covered (19-21 fps, median gap 45 ms); the rest is the compositor thread's ~45 ms a Wikipedia frame: T-517; 2026-10-19 evening (rounds 349-357): the 10 s pan 23.2-24.1 fps; the page's tile uploads, partial uploads, deferred uploads, a topped-up pool and the express lane on its own slot measured and not kept; the band frame is GPU-bound (r357). BLOCKED by the circuit breaker: the owner's finger round (T-514) and the bench's sensord load (the owner's call) decide)
T-506	Animation, new content and self-scroll measured against the chart (DOM edit and typing ≤ 106 ms)	cavekit-gpu-rendering.md	R5	T-502	yes	S	DONE 2026-10-18 (rounds 202-203: the animation page 26-31 fps reading only its changed rows; typing 22-62 ms and DOM edits 26-70 ms from the input to the first published frame with changed rows (an input-latency probe on the inject channel); the self-scrolling page followed at 10.7-13.9 fps)
T-507	Every other card event on the GPU path: rotation, resize, select popups, caret, link and find highlights, thumbnails, navigation without stale frames	cavekit-gpu-rendering.md	R9	T-502	yes	M	DONE 2026-10-18 (rounds 204-212 on jihad-card-events.html: rotation to landscape and back, the keyboard's resize, the select list and pick, the caret, find-in-page, link navigation with back, forward and reload, the thumbnail; the thumbnail's view row and the popups' origin corrected for overscan; the test hook gains <, >, R, O<orientation>; the inject channel thumb <path>)
T-508	Video and Flash on the GPU path: <video> (VP8, Theora, H.264) to the end, styled video (rounded, fade, under a modal), Flash draw/animate/input	cavekit-gpu-rendering.md	R7	T-502	yes	L	DONE 2026-10-19 (rounds 213-228: VP8, Theora, VP9 and H.264 to the end on the GPU path, at the clips' rates; styled video (rounded, fade, modal); Flash draws, animates and takes keys; push-engine-update.sh now carries FFmpeg's libavcodec/libavutil, which a device updated by it lacked (H.264, AAC and MP3 were refused on both paths); new pages jihad-video-styled.html, jihad-anim-opacity.html, jihad-canplay.html)
T-509	Safety and fit: bounded GPU waits with a fallback to the software path, a simulated EGL failure, memory and GPU memory against the software path, three daemons at once, idle CPU	cavekit-gpu-rendering.md	R1, R10	T-502	yes	M	DONE 2026-10-19 (rounds 229-232: the daemon gives the GPU path up for its life after three failed or refused frames or a frame not composed in 3 s, and renders in software (gpuFallBack); engine 0047 (a forgotten frame writes nothing; JIHAD_GPU_TEST_NO_EGL, JIHAD_GPU_TEST_STALL); memory and idle CPU measured against the software path, three GPU daemons at once)
T-510	The default: the GPU path on in all three event.d jobs with JIHAD_GPU_COMPOSITOR=0 and a pref to force software; Mochi and Mojo load, pan, zoom and close on it; the startup report	cavekit-gpu-rendering.md	R1	T-503, T-507, T-509, T-515	yes	M	DONE 2026-10-19 (blocked by T-504 until then so the owner's zoom complaint would not ship; the express lane (T-515) closed it. The generated upstart jobs start the GPU path with the express lane; JIHAD_GPU_COMPOSITOR=0 or the pref jihad.gpu.enabled=false (read before the engine starts) give the software path, and the log names the path taken; all three interfaces load, pan, zoom and close Wikipedia on their default daemons (default-check.sh), the pref and the env override checked on the bench)
T-511	WebGL through EGL: contexts on the device's GPU on the GPU path, a spinning-cube test page ≥ 24 fps, context loss, spot conformance	cavekit-gpu-rendering.md	R8	T-510	yes	L	DONE 2026-10-25 (T-527: the 480 px cube 25.3-28.0 fps on the card, 360 px 28.9-29.9, 240 px 29.4-30; the partial redraw through the frame trailer and the tall band surface, engine 0063; earlier: PARTIAL 2026-10-19 (rounds 281-313, docs/rounds/t502-2026-10-17/webgl-r281-313.txt; engine 0051 (WebGL on EGL on the GPU path, the stencil back mask, one share group so the canvas's texture is composited with no copy, a context released before its destruction), 0052 (the column read); the daemon's own-slot base judged by the recent changes. Met: contexts on the GPU path and none on the software path; the cube renders correctly; 8 open/close cycles and a lose/restore without a fault; the spot checks and the missing extensions listed. Open: the 480 px cube 18-20.5 fps on the card (240 px 26.5-26.9, 360 px 21.5-22.8): the compositor job's ~45 ms, carried by T-517; 2026-10-19 evening (round 358): the 480 px cube 19.5-20.9 fps on the card, the page's own rate 23.8-25.1; the card cannot pass the page's rate. BLOCKED with T-517's floor))
T-512	Fit to keep: diagnostics pruned and bounded, the knobs listed, the engine work folded into the patch queue, the docs and README	cavekit-gpu-rendering.md	R11	T-503…T-511, T-517	no	M	DONE 2026-10-19 (e8b0e425: the [jihad-gpu] diagnostics behind one gate each in the engine (0057, mozilla/gfx/JihadGpuLog.h) and the daemon (render/goanna/JihadGpuLog.h), three startup lines in a shipped daemon, the default checks of all three interfaces pass; every switch in tools/device/t502/README.md; the 57 patches reproduce the pin and each has its ENGINE-SOURCE row; RENDERING-OPTIONS, the README and PICKUP carry the evening's numbers. The review (R11's third) is T-513)
T-513	Review of Tier 87 (Codex by direct invocation, the inspector as backup) and fixes	cavekit-gpu-rendering.md	R11	T-512	no	M	DONE 2026-10-19 (Codex out of credits: two inspectors, 12 findings, all fixed, context/impl/impl-review-findings-gpu-2.md; twenty card cycles flat (R10); a fast flick and two cards in turn pass the verify mode once its reference is composed behind the job (R2); the three default checks on the bench)
T-514	[OWNER] Finger round on the GPU path in all three interfaces: pan, flick, pinch, double-tap	cavekit-gpu-rendering.md	R12	T-513	yes	S	DONE 2026-10-24 by the emulated finger (T-524; R12 at [~]: "finds it smooth" stays the owner's word)
T-515	Frames made off the main thread: the card's scroll and zoom read on an IPC thread, the compositor composing and the frame published while the main thread runs the page's script or a paint (the open criteria of T-504, the first frame at a new zoom within 250 ms under script, and T-505, desktop Wikipedia ≥ 24 fps)	cavekit-gpu-rendering.md	R3, R6	T-505	yes	L	DONE 2026-10-26 (its two open criteria met: the first frame at a new zoom within 250 ms while the page's script holds the main thread — desktop Wikipedia right after load, pumps of 1-1.3 s, 12 of 12 double taps by finger (in 67-104 ms, out 0-1 ms: the kept frame), docs/rounds/t527-2026-10-25/dtap6-busy-daemon-gpu.log; desktop Wikipedia ≥ 24 fps on the owner's bench (T-527)) — PARTIAL 2026-10-19 (rounds 240-267, the express lane: each card's YAP commands read on a thread of the proxy's own, the daemon's frame lock (render/goanna/JihadFrameLock.h; the main thread lets go of it in the pump and the layer paint), an express thread that composes the layer tree as it stands and publishes, engine 0048 (the paint and the compose post apart, jobs from any thread). Met: the zoom under script (Wikipedia after load 10/10 within 160-225 ms, held off median 681 and three never framed; a page holding the main thread 600 ms in 700 within 45-91 ms, held off median 450); pans on that page 26-32 fps (2.4-3.8); Wikipedia's 2.2 s restyle under frames every ~45 ms. Open, and no longer the main thread's: desktop Wikipedia 19-21 fps, the compositor thread's ~45 ms a frame (T-517); double taps 40-449 ms, GPU draws of the other resolution's tiles (T-516). docs/rounds/t502-2026-10-17/express-r240-267.txt)
T-516	The zoom's first frame on every gesture: a zoom out after a zoom in's repaint draws the new resolution's tiles at ~0.55, 20-35 ms a draw on the Adreno (a 179 ms composite, round 263); found and removed, card double taps both ways within 250 ms in runs of ten	cavekit-gpu-rendering.md	R6	T-515	yes	M	DONE 2026-10-26 (the owner chose both levers after each was measured alone: the first frame at a new zoom reads the card's view alone (engine 0067) and the daemon brings the second core online on the card's zoom or pan; double taps by finger 20 of 20 within 250 ms (in 49-93 ms, out 0-152 ms), each lever alone 40 of 40; the readback's slowness was one contended core with the second unplugged after a pause, not the GPU's clock (measured read-only first)) — earlier: PARTIAL 2026-10-19 (rounds 270-280: the slow frames were not draws: the first readback after a pause ran 100-390 ms (30-40 usual) with the CPU swizzle 20-122 ms (7 usual): the clocks after two idle seconds (the CPU floor released, the second core unplugged, the GPU idled down) and the engine's other threads. Kept: the YAP reader raises the clock hold on a scroll, zoom or smart-zoom request (JIHAD_CLOCK_HOLD_MS), the compositor and express threads at nice -6 (JIHAD_GPU_FRAME_NICE), the clocks on the timeline. Double taps (12 a run): before 4-8 over 250 ms (max 479), after 0-2 (261-323; median 158). Tried and dropped: an 8 s hold (no change: the GPU still idles), a warm-up frame on the smart-zoom request (its own slow frame queued ahead of the zoom's: worse). Open: the GPU's ramp after a pause)
T-517	Desktop Wikipedia's pan frame under 41 ms: the compositor thread's ~45 ms a frame (3-7 ms of CPU a composite, the column passes, the readback's wait for the GPU) broken down and cut, whichever lane asks; ≥ 24 fps in the scripted pan; the same floor under a large animated area: the 480 px WebGL cube ≥ 24 fps on the card (T-511's open part; measured option: the adapter invalidating only the changed strip, +12 %, round 304)	cavekit-gpu-rendering.md	R3, R8	T-515	yes	L	DONE 2026-10-26 (desktop Wikipedia's scripted pan 24.6 and 24.2 fps on the owner's bench, a mean frame of 40.7 and 41.3 ms; the 480 px WebGL cube 25.3-28.0 fps on the card (T-527: the dirty rect through the trailer, engine 0063)) — PARTIAL 2026-10-19 (rounds 314-332, docs/rounds/t502-2026-10-17/floor-r314-320.txt. Kept: no cache clean of the frame rows on write-through segments; frame windows (a frame moves within its segment instead of 3.7 MB of rows being moved: render/browserserver/Src/JihadFrameWindow.h, the adapter's BrowserOffscreen, engine 0053), with both lanes keeping every slot's stale rows and the express lane building on its own slot when that measures faster. The long test page 30.5-36.4 fps (median gap 26 ms, was 29). Open: Wikipedia 19-21 fps: its two see-through sticky panels are composed again every frame in 3-4 chunks of the band surface, ~5 ms of CPU a pass; a rule keeping see-through sticky columns once proven still did not apply there (round 333: the contents panel changes as the page scrolls) and was reverted; next: one tree traversal for all of a frame's regions; rounds 342-348: engine 0055 (a band frame's regions in one composite) and 0056 (tiles outside the clip skipped before their lock): chunked frames compose 10.6 ms where they took 22.5; Wikipedia's 10 s pan 23.4-23.9 fps (was 20.0-21.7), typical frame 35-38 ms; the gaps left are the page's own work in the compositor's queue and on the other core; 2026-10-19 evening (rounds 349-357, floor-r314-320.txt): the page's layer transactions on the compositor thread timed (tile uploads), every cut after 0055/0056 measured and recorded in dead-ends.md; BLOCKED with T-505)

Tier 86 — Content blocking: ads and trackers (2026-10-15; added to the site, not a regeneration)

The owner: "build in an ad and tracker blocker if possible (following license terms)". Kit: cavekit-content-blocking.md R1–R7. The lists: EasyList + EasyPrivacy, taken under GPLv3+ (the kit's table says what was rejected and why).

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-485	The lists in the tree: third_party/easylist/{easylist.txt,easyprivacy.txt,README.md} pinned (upstream ! Version: kept, sha256 in make-device-bundle.sh, bundled raw under blocklists/); NOTICE, CREDITS.md, about:jihad, licenses/ name them under GPLv3+	cavekit-content-blocking.md	R2, R7	—	no	S	DONE (third_party/easylist/ snapshots 202610151429 with README, sha256-pinned and bundled raw under blocklists/ by make-device-bundle.sh; NOTICE, CREDITS.md, about:jihad and the About pane name them under GPLv3+)
T-486	render/goanna/components/jihadContentBlocker.js: the compiler (host set, third-party set, exceptions, generic cosmetic selectors, skip counts), the profile cache, nsIContentPolicy in the content-policy category (top-level and same-host never blocked, eTLD for third-party), the user sheet, prefs jihad.blocking.*, the counter, the log lines; registered by make-device-bundle.sh, created at daemon start; test pages jihad-blocking.html, jihad-blocking-many.html	cavekit-content-blocking.md	R1, R2, R3	T-485	yes	L	DONE on all three (v5.1 packages: Enyo 14:20, Mochi 14:36, Mojo 14:53 on the fifth try; the http test page right on Mochi; MDN clean); memory +40 MB idle open (R6: the memory reporter figures being read; jemalloc is off in this build)
T-487	The updater: every 4 days from easylist.to over HTTPS, newer ! Version: only, compile and swap in place, keep-what-you-have on failure, jihad.blocking.lastUpdate; tools/host/count-blocklist.py for the independent count	cavekit-content-blocking.md	R5, R2	T-486	yes	M	DONE 2026-10-19 (forced update, network off, no repeat within 4 days or a day after a failure, on the Enyo card; two bugs found and fixed: EasyPrivacy refused as truncated on every update, the early look skipped and the next six hours away)
T-488	Settings → Privacy → "Content blocking": two switches, the read-only versions/counter line, the first button row ("Update the lists now") in preferences.js; effect-verified live in the three variants	cavekit-content-blocking.md	R4	T-486, T-487	yes	M	DONE on all three (v5.1 packages; Mochi's rows shot, Enyo and Mojo to shoot in the second run)
T-489	Device verification and numbers: MDN without adverts (before/after screenshots, the README's image), VmRSS delta, load done with the sheet on/off and with 200 blocked images, the exception and third-party cases, network-off update; the README's feature sentence with the credit	cavekit-content-blocking.md	R1, R3, R6, R7	T-488	yes	M	DONE as far as the device allows (MDN without adverts in the README; the http test page; memory: idle 34.8 MB with the lists on demand, +22 MB with a card (the 12 MB goal not met, accepted for this version); the GC criterion met (1,073 vs 1,101 ms); the 200-image timing not measured (the stalls are the pages' own scripts); the updater's forced update and network-off tests still pending)
T-490	Review of Tier 86 (the inspector; Codex when its credits return) and fixes; the licensing audit extended to the lists	cavekit-content-blocking.md	R7	T-485…T-489	no	M	DONE 2026-10-19 (round 26 in impl-review-findings-6.md, every finding dispositioned; F-6 fixed with the recomposed card view; F-21 fixed: count-blocklist.py has a character-level tokenizer sharing nothing with the compiler, and --cross finds no rule the two classify differently in the bundled lists or today's 202610191813; the component's fixes run on the bench in all three interfaces)

Tier 85 — The modern web: ledger and fixes (2026-10-15; added to the site, not a regeneration)

The owner: "showcase the modern web", then "dont just breeze past that. document and fix issues loading the modern web." Kit: cavekit-modern-web.md R1–R5.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-480	docs/MODERN-WEB.md: the ledger with every site of 2026-10-15 (the README round and the six-site survey), each row complete; how to get content-page console output on the device (JIHAD_JS_CONSOLE=all); linked from the README's Known issues	cavekit-modern-web.md	R1, R4	—	no	S	DONE (docs/MODERN-WEB.md: 19 rows, every one dispositioned; the console method with the log-rotation caveat; linked from the README's Known issues)
T-481	dom.max_script_run_time sized from the device (DuckDuckGo results as the reference; 60 s expected) in packaging/prefs/, the guard kept; render/goanna/test/jihad-busy.html still brings the dialog	cavekit-modern-web.md	R2	—	yes	S	DONE (device 2026-10-15: dom.max_script_run_time 60; The Verge renders in full, DuckDuckGo Lite renders, the endless script still brings the dialog — shots t481-verge, t481-lite, t481-busy; the pref in jihad-platform-prefs.js, set in the Mojo and Mochi profiles for the test)
T-482	Cloudflare: the challenges.cloudflare.com rule in the daemon's UA table (a Firefox ESR token, the Pale Moon finding cited), daemon rebuilt and pushed, w3.org tested; the row fixed or open with the console output	cavekit-modern-web.md	R3	—	yes	M	DONE as far as it goes (the challenges.cloudflare.com rule in EngineHost.cpp, daemon rebuilt and on the device; w3.org still stops — the console shows Turnstile running and failing on its own verdict; the ledger row is open with the next experiment, a current Firefox token for the site's own requests = moving the base UA, the owner's decision)
T-483	GitHub's deferred parts: the console with JIHAD_JS_CONSOLE=all on the device, the errors or the missing feature named, a UA-token or pref fix tried if one is indicated, the row dispositioned	cavekit-modern-web.md	R4	T-480	yes	M	DONE (console with JIHAD_JS_CONSOLE=all 12:16: ReferenceError: Turbo is not defined; ES-module loading, as Mastodon's import.meta error shows; upstream UXP, no configuration applies; the method and the log-rotation caveat in docs/MODERN-WEB.md)
T-491	Round 2: a pretend user's half hour on the device — the site list and the script in tools/device/ (browse-round.sh, browse-round-sites.txt), run in Mochi with the console bridge on all sources, every visit logged and shot; the "Round 2" section of docs/MODERN-WEB.md with every row dispositioned	cavekit-modern-web.md	R6	T-481, T-486	yes	L	DONE for the first run (2026-10-15 11:40–12:20 device time, Mochi, blocker v3: 30 visits, docs/rounds/round2-2026-10-15*; the 30 rows with cause and disposition in docs/MODERN-WEB.md; the scripts in tools/device/ and tools/host/run-browse-round.sh; the second run after the fixes is T-492's)
T-492	Round 2's fixes: every issue configuration or Jihad's own code can fix, fixed, packaged, pushed and re-run; the comparison in the ledger	cavekit-modern-web.md	R6	T-491	yes	L	DONE for this round (2026-10-16: the UA rows for Google/Bing/Etsy/YouTube kept, the rest removed; the GC threshold 64 + the v6–v9 blocker fixes (benchmark 5–6× faster, a full GC 1,073 ms with the lists); the daemon purges and collects when the last card leaves (428 → 170 MB); closed pages are released — the rest is the allocator, T-493 proposed; the NS_ERROR_NOT_INITIALIZED thrower unnamed (jQuery loads clean); Cloudflare/Startpage/eBay/old-Reddit and the four module apps open/upstream with causes)
T-493	The engine rebuilt with jemalloc (--enable-jemalloc in mozconfig.goanna-arm; hours) — the one lever left for resident memory after pages close (the allocator returns nothing and reports nothing without it); measured by the memory report and the round's RSS column	cavekit-content-blocking.md, cavekit-engine-update.md	R6	—	yes	L	DONE (2026-10-16: --enable-jemalloc; the allocator linked INTO the daemon with -rdynamic — on Linux UXP puts it in the program, the trap in docs/MODERN-WEB.md; daemon 11 4c318d49 on all three variants; after three heavy pages and the last close 149 MB resident, was 170; heap-allocated reports; memory.free_dirty_pages on)
T-494	The challenges (the owner 2026-10-16: "Try to fix the challenges blocking sites from loading including cloudflare and RECAPTCHA"; the base UA stays 52.9): Cloudflare's interstitial (Medium, Britannica, AP, npm, w3.org), Startpage's JS check, eBay; reCAPTCHA already past via google.com's row. Lead: the security error in Cloudflare's frame ("may not load data from https://example.org/"); a probe page and a console-on capture of every line in the challenge frames; then the fix	cavekit-modern-web.md	R3, R6	—	yes	L	DONE as a diagnosis (2026-10-16: the engine's load policy cleared; the challenge frames emit nothing; header/script consistency alone changed nothing; a whole-page current identity passes Cloudflare on w3.org, Britannica, npm, not Medium — the fix is T-495; reCAPTCHA already past via google.com's row; Startpage and eBay to retest under T-495)
T-495	The learned identity (the owner 2026-10-16: "52.9 by default but adjust based on the needs of the page"): components/jihadUserAgent.js — cf-mitigated: challenge on a top-level response learns the site, pins the docShell, reloads once; the header for the site's requests; prefs jihad.ua.*; Settings → General → Browser identity; registered by the bundle and created by the daemon; the challenges.cloudflare.com row out of the daemon's table	cavekit-modern-web.md	R3	T-494	yes	M	DONE (2026-10-16: w3.org and Medium learned, reloaded and rendered on the device; the list persists; Settings → General shows it; shipped in the 03:21 packages with daemon 11)
T-496	The identity read on the wire and the second signature: render/goanna/test/jihad-identity-server.py (a host-side site that fakes cf-mitigated: challenge and Startpage's /sp/captcha-block redirect, logs every request's header, and has pages that report navigator.userAgent); the component's _challengeOf with the redirect signature; the no-loop branch, the frames' header and the third-party frame measured; eBay and old Reddit shown not to be the identity's (curl from the host, both tokens); daemon 12's console line carries the column (source:line:column) for the NS_ERROR_NOT_INITIALIZED bundles	cavekit-modern-web.md	R3, R1	T-495	yes	M	DONE (2026-10-16 12:13–12:25 on Enyo: every box of R3 ticked but Startpage's search, which meets Anubis's proof of work; docs/MODERN-WEB.md T-496)
T-497	YouTube sped up (the owner 2026-10-16: "speed up youtube"): the desktop watch page measured against the mobile site on the host (13.7 MB vs 3.0 MB of script) and on the device (worst stall 24.4 s vs 5.3 s, RSS 274 vs 162 MB, the home feed in 18 s); jihad.ua.sites rules in components/jihadUserAgent.js (header + docShell pin), youtube.com → a mobile Firefox identity; the daemon's youtube.com row out; Settings → General "Sites with a set identity"	cavekit-modern-web.md	R3, R6	T-495	yes	M	DONE (2026-10-16; the ledger T-497; the tap on the mobile player is a finger test)
T-498	The web compatibility shims (the owner 2026-10-16: "fix the regular duckduckgo"): components/jihadWebCompat.js — a sandbox per document defines what the engine lacks (CSSStyleSheet.replace, popover, inert, Intl.ListFormat, reportError, checkVisibility, AbortSignal.any, URLSearchParams.size, Array.fromAsync) plus one agent sheet; jihad.compat.trace names the engine call that throws an XPCOM result to a page; the API probe page's second list; registered by the bundle, created by the daemon (daemon 13)	cavekit-modern-web.md	R1, R6	T-496	yes	M	DONE (2026-10-16: 16/16 on the shims' test page on the device; Etsy's 47 errors gone and its search styled; DuckDuckGo's results render — its open menu panel is T-499; all three variants carry the component)
T-499	The layout viewport follows a page's overflow (found 2026-10-16: DuckDuckGo's slide-out menu, parked at innerWidth + 4 = 772, is on screen because the page is 1000 wide and the card fit-zooms the whole width): an offscreen engine window as wide as the content once per load, the card's size kept apart from the engine's in the pan/floor arithmetic (BrowserPageGoanna.cpp emitGeometry/setScrollPosition, GoannaRenderPage.cpp), pref-gated, measured on DuckDuckGo, Wikipedia, BBC, YouTube mobile, with rotation	cavekit-modern-web.md	R6	T-498	yes	L	DONE 2026-10-19 (the card's own size resent keeps the layout viewport; the layout viewport is the document's, back to the card's size for the next one, decided once the document is parsed and never on about:blank; probe page with rotation, BBC, YouTube mobile, Wikipedia, DuckDuckGo on the Enyo card)
T-500	The stall (the owner 2026-10-16: "fix the stall"): a leaf-PC stall sampler in the daemon ($JIHAD_STALL_SAMPLER, tools/host/stall-resolve.py against the unstripped libxul) names what holds the thread — the pages' own JavaScript, compile, the collections; the clock found at 192–384 MHz under load (Palm's ondemandtcl ramps on a touch, up_threshold 95); the clock hold in JihadBrowserServer.cpp (the frequency floor at the top while the engine has work, released after 2 s idle; pref jihad.cpu.hold); never the governor (it wedged the kernel's cpufreq lock: a reboot)	cavekit-modern-web.md	R6	T-498	yes	L	DONE 2026-10-16 (commit 4ded8d72: the stall sampler and tools/host/stall-resolve.py; the clock hold in JihadBrowserServer.cpp, pref jihad.cpu.hold, never the governor; both in use in the GPU rounds of 2026-10-18)
T-501	H.264 and AAC: FFmpeg 4.4.5's decoders cross-built for the device (build/webos-oe/fetch-ffmpeg.sh, build-ffmpeg-arm.sh: libavcodec.so.58 + libavutil.so.56, LGPL v2.1+, decoders and parsers only), shipped by the bundle when built, NOTICE and licenses/runtime/; the engine's FFmpegRuntimeLinker loads them for <video>, <audio> and Media Source Extensions	cavekit-gre-widgets.md	R1	—	yes	M	DONE 2026-10-16, re-verified 2026-10-19 (commit 4ded8d72 built and played them; a device later updated only by push-engine-update.sh had lost the libraries and refused every MP4, MP3 and AAC; the push now carries them, and H.264, AAC and MP3 play to the end again on the device)
T-484	Wikipedia's layout decided and recorded; every fix of T-481…T-483 re-tested and its ledger row updated with a screenshot; the README hero retaken on sites that work	cavekit-modern-web.md	R1, R5	T-481, T-482, T-483	yes	S	DONE (Wikipedia: desktop kept; the README hero recomposed 14:52 — Internet Archive, BBC News, MDN without adverts, unthemed, Aero backdrop — in docs/screenshots/cardview.jpg)

Tier 84 — One repository builds Jihad on any Linux (2026-10-14; added to the site, not a regeneration)

The owner: "is everything for jihad browser now centralized to …/Jihad-Browser or is there still code out of tree for the build etc?", then "yes. and try to integrate some of the host tools into the readme instructions for building jihad. it should be as self contained to its own repo as possible while running on a wide variety of linux distros", and "use git submodules for build dependencies for automation". Kit: cavekit-device-build.md R11.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-471	Build scripts read only the repository: drop the sibling fallbacks (../toolchains, ../mochi-sampler, ../../webos-stacks) and the …/Jihad-Browser/… paths; build-webos inside the repository	cavekit-device-build.md	R11	—	no	S	DONE (adapter rebuilt byte-identical; Enyo and Mojo packages identical in content; Mochi differs only in BUNDLED-VERSIONS, which now names repository-relative paths instead of the builder's home)
T-472	Build dependencies as submodules: crosstool-NG, libpbnjson, yajl, npapi-headers, webos-sdk-redux, build-webos (the heavy and optional ones not checked out by default); Jihad's own files for them moved into the repository; scripts and manifests pointed at them; the adapter rebuilt byte-identical	cavekit-device-build.md	R11	T-471	no	L	DONE (61fe5974, f94af9c9, 9ff7d71a; both adapters, gcc 4.3.3 and gcc 9, rebuild byte-identical from the submodules; the Qt4 headers re-fetched to the pinned digest; the seven device libraries byte-identical over novacom; the sysroot assembled by arm-sysroot.recipe digest-identical to the one every build used; the toolchain rebuilt from the crosstool-NG submodule: same version, identical compiler output, byte-identical libc-2.23.so, ld-2.23.so and libstdc++, 24 of 5,652 files differ and all are build-time noise. The clean clone found 32 of the 187 Debian packages answering cached 404s from archive.debian.org's CDN; the fetchers now retry with a cache-buster, and all 187 came)
T-473	fetch-pdk.sh downloads the Palm SDK .deb from the Internet Archive by itself and checks its sha256	cavekit-device-build.md	R11, R3	—	no	S	DONE (e1ae67c0; the .deb downloaded from the Internet Archive, sha256 matched, the extracted tree identical to the one in use)
T-474	The adapter (32-bit PDK gcc) and the three packages (palm-package, Java) build inside the container; build-all-device.sh needs only git and podman on the host; the output equals the host-run build	cavekit-device-build.md	R11	T-472, T-473	no	M	DONE (00a82f35; through in-container.sh the adapter, the three packages and the daemon 06e75a68 are byte-identical to the host-run build and to the device; build-all-device.sh ran end to end through the containers. The run from a clean clone at another path (Jihad/.jihad-cleanclone-test, 2026-10-14): every input without the host's copies, the engine built, the daemon 06e75a68, the shims, the three impls and the three packages' contents byte-identical to this checkout's, once QT_NO_DEBUG (round 25) had taken the checkout path out of the impl)
T-475	novacom and novacomd from the webos-sdk-redux submodule with one script (tools/host/)	cavekit-device-build.md	R11	T-472	no	S	DONE (e3a5f4fc; the repository-built novacom lists the TouchPad, and dev-sh.sh resolves it first)
T-476	The scratchpad's device helpers and the Codex review runner into tools/, with no session paths	cavekit-device-build.md	R11	—	no	S	DONE (e0f38f07; the read-only device tools run on the TouchPad; set-prefs.sh run once with a pattern that matches nothing)
T-477	The README's "Building" section: host packages per distribution family, clean clone to three .ipks to the TouchPad; DEVICE-BUILD.md and TOOLCHAIN.md brought in line	cavekit-device-build.md	R11	T-474, T-475, T-476	no	M	DONE (5cace42e)
T-478	Outside the repository: the two reference clones (OpenSSL-legacyWebOS, squid-for-webos) from the scratchpad into ref-forks/; the Relay's two Squid scripts asked of the server agent, to keep in deploy/relay/	cavekit-device-build.md	R11	—	no	S	DONE (the two clones in ref-forks/, current; the Relay's two scripts received 2026-10-14 after the request was resent twice past the reboot, sha256 as on the server, in deploy/relay/ with a README; the admin's login name stays in their paths, the owner's call)
T-479	Review of Tier 84 (the inspector; Codex is out until Oct 18th) and fixes	cavekit-device-build.md	R11	T-471…T-478	no	M	DONE (the inspector: REQUEST CHANGES, 2 P1, 7 P2, 16 P3, round 25 in impl-review-findings-6.md; all fixed in 59040faf but the engine pin, which is the owner's push; the three packages rebuilt without the home path and pushed to all three variants, example.com loaded and painted in each; the clean clone at another path gives byte-identical shims, impls, daemon and package contents)

Tier 83 — Proxies and webOS's certificates (2026-10-13; added to the site, not a regeneration)

The owner: "add support for riverstone relay proxy, and webone, and self hosted proxies in jihad. also add support for system certificates including community patches if installed." Kit: cavekit-proxies-certs.md (decisions recorded there).

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-457	webOS's certificates: a component that reads webOS's stores at NSS start and trusts, in memory, the CA certificates not HP's, not removed by Mozilla, not Jihad's, not expired, not the Relay's; lists from HP's 3.0.5 doctor and CCADB with a rebuild script; the Settings switch and list	cavekit-proxies-certs.md	R4	—	yes	L	DONE (e223591f, 9394c08b; device, all three)
T-458	The Proxy section: No proxy (default) / webOS's setting / Riverstone Relay (its CA an option, off) / WebOne / My own proxy / PAC, read from and written to network.proxy.*, validated, the old row gone; the daemon reads webOS's setting over its Luna connection	cavekit-proxies-certs.md	R1, R2, R3	—	yes	L	DONE (typing a proxy by hand: the finger round)
T-464	Opening a CA certificate asks (a JS nsICertificateDialogs through the card's confirm), refusing the Relay's and non-CAs; the added CAs listed with Remove	cavekit-proxies-certs.md	R6	—	yes	M	DONE (device, Enyo; WebOne's attachment opened too)
T-459	A proxy's 407 names the proxy	cavekit-proxies-certs.md	R5	—	yes	S	DONE (Enyo, Mochi; Mojo's sheet names it)
T-460	The Relay's splice rule for the server agent (Squid browser acl on Jihad's UA, before the bump rules), sent over wormhole-rs	cavekit-proxies-certs.md	R2	—	no	S	DONE (applied by the server agent 17:07 UTC, squid.conf 31–33; Jihad's UA gets the site's issuer, others squid-proxy)
T-461	Device: each proxy mode through a logging PC proxy (three shells), webOS's setting, a login proxy, WebOne on the PC, the Relay (CA option off and on; after the rule); the stores' import (five newer roots, a Certificate Manager CA, the switch); a CA link	cavekit-proxies-certs.md	R1–R6	T-457, T-458, T-459, T-460, T-464	yes	M	DONE (the Relay's pass-through too: example.com's own Cloudflare certificate through it, Enyo, CA option off)
T-462	Review of Tier 83 (Codex when it has credits, the inspector before) and fixes	cavekit-proxies-certs.md	R1–R6	T-461	no	M	DONE (inspector round 23, 9394c08b; Codex after Oct 18th: T-466)
T-463	The server's round-22 reply: install.sh replaces an unedited site block in place (.new only for an edited one); DEPLOY.md says to check default_server on :443	cavekit-addons-site.md	R10	—	no	S	DONE (Riverstone-Addons c322904, 68caed1; scratch harness; sent with the Relay note)
T-465	[FOUND] After a pick in a card's drop-down that grows the page (Settings → Proxy → Riverstone Relay), the card keeps its old picture, its own list included, until the next touch; the daemon did paint (Enyo, Mochi). A page that grows by itself redraws fine (/media/internal/jihad-grow.html)	cavekit-offscreen-rendering.md	R8	—	yes	M	WITHDRAWN: not a redraw bug. Both sightings were screenshots taken before the pick ran (the shot at 30 s, the pick at about 30 s); replayed exactly, the 30 s shot shows the old state and a 40 s one the new, and 15 of 15 picks with time redrew within a second (Enyo, Mochi). card-selftest.sh run now stamps the shot's time and marks the hook's lines logged after it
T-466	Codex on Tier 83 (Jihad 5db7860c.., the site 057dbad..d9359ea) when its credits return (Oct 18th, 2026 11:59 AM), with round 22's	—	—	T-462	no	M	DONE by the inspector instead (the owner, 2026-10-13): round 24
T-467	[FOUND] JS components are read as Latin-1: non-ASCII in their strings came out garbled (the print status's ellipsis, the Persona importer's BOM strip); escaped. The install scripts leave a byte-identical adapter alone (a replaced one LunaSysMgr had mapped kept / read-write)	cavekit-device-build.md	R8	—	yes	S	DONE (e223591f)
T-468	The Relay hides users' addresses (the owner): Squid's forwarded_for delete, sent to the server agent with checks before and after; then http://httpbin.org/ip through the Relay on the device shows only the server	cavekit-proxies-certs.md	R2	T-461	no	S	DONE (applied 18:12 UTC, squid.conf line 103; the device's http through the Relay shows only the server; d9359ea live, smoke 43/43)
T-469	Review round 24's fixes (the inspector, for Codex): R4 roots only and never a certificate the SSL dialog installed; R6 only for a page (offer), site certificates refused; the [proxy] marker; webOS's last proxy kept; the gallery offers the bundled set; finger-sized link buttons; the kits brought up to date	cavekit-proxies-certs.md, cavekit-personas-ui.md, cavekit-chrome-look.md	R1, R4, R5, R6; R7, R9; R2	T-456, T-466	yes	M	DONE (cfbfbcf2; device, all three on daemon 06e75a68)
T-470	The site's round-24 fixes (71dd173, f9bcb79) to the server agent: install.sh's site-block rules, DEPLOY.md step 11, the MPL Personas' links, the panel redirect, 503 on a database outage, the feed's order and per_page, login timing	cavekit-addons-site.md	R5, R10	T-469	no	S	DONE (live 2026-10-14 10:26 UTC; install.sh kept the block as the repository's; the five MPL Personas updated at 1.0, pictures as they were; smoke 43/43 on the server and through Cloudflare; the feed 39 of 39 with Aurora Silk's MPL and credits links)

Tier 82 — Review round 22 and what the device checks found (2026-10-13; added to the site, not a regeneration)

Codex ran out at 22:32 on all four ranges (retry Oct 18th, 2026 11:59 AM); the inspector reviewed the three ranges no one had (impl-review-findings-6.md, round 22).

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-449	Round 22, Jihad: change before blur on keyboard-down; the focus flag drained first; only known setting types laid out; Unicode font names; odd sizes sorted in; card pictures swapped not redrawn, the queue self-restarting, small/failed headers remembered; Mojo's keyboard-away mark tied to its page	cavekit-chrome-look.md; cavekit-preferences-ui.md; cavekit-personas-ui.md	chrome-look R2, R4; prefs R9; personas R9	T-448	yes	M	DONE (device: Enyo, Mojo)
T-450	Round 22, engine tooling: update-engine.sh's guards and safe --finish/--abort, atomic pins, .gitmodules' branch a pin, the checker's ICU/loadURI/C++ checks, patch-queue --add's format, ENGINE.md §4.5/§6	cavekit-engine-update.md	R2, R3, R4, R7	T-448	no	M	DONE
T-451	R8's missing device checks: Mochi and Mojo (https, Palefill, a Persona, about:addons/config/settings), Enyo (about:addons, about:settings)	cavekit-engine-update.md	R8	T-448	yes	S	DONE
T-452	[FOUND] Daemon: the page's size before a page-made scroll (anchors and deep links at load stayed at the top)	cavekit-preferences-ui.md	R9	T-449	yes	S	DONE (Enyo, Mochi)
T-453	[FOUND] Mojo: a card opened on a page comes up without the keyboard	cavekit-chrome-look.md	R4	T-449	yes	S	DONE
T-454	[FOUND] Install scripts: the read-only restore tried for ~10 s (one push left / rw)	cavekit-device-build.md	R8	—	yes	S	DONE (BusyBox; three pushes from a read-only /)
T-455	Round 22, site: migration 002 retried and reported; the smoke test on the server itself and stricter on the feed; :443 answers only its name; the importer updates a Persona whose id is taken	cavekit-addons-site.md	R3, R4, R6, R10	—	no	M	DONE (Riverstone-Addons 057dbad, local copy; follow-up sent to the server agent)
T-456	Codex on round 22's four ranges when its credits return (Oct 18th, 2026 11:59 AM)	—	—	T-449, T-450, T-455	no	M	DONE by the inspector instead (the owner, 2026-10-13): round 24, impl-review-findings-6.md

Tier 81 — The engine on UXP RB_20261006; updates made easy (2026-10-12; added to the site, not a regeneration)

Owner: "Proceed with the UXP update, with a focus on making future updates easier, and optimizing the code where possible, while supporting modern web standards. target the latest release". Kit: cavekit-engine-update.md R2, R3, R4, R7, R8.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-442	Split the squash into one commit per patch; carry the 39 onto RB_20261006 (three-way); read the six overlapping files	cavekit-engine-update.md	R7, R8	—	no	M	DONE (no conflict; the NPAPI-GTK2 option dropped)
T-443	build/engine/update-engine.sh (start / --finish / --abort / --pins); UXP_TAG in the pins; the queue as git mails; the bundler reads the ICU name	cavekit-engine-update.md	R2, R7	T-442	no	M	DONE
T-444	render/goanna/EngineCalls.h: the 15 positional engine calls behind named wrappers	cavekit-engine-update.md	R3	—	no	S	DONE (builds with the daemon)
T-445	Release build in Thumb-2 (--enable-release --enable-gold --with-thumb=yes)	cavekit-engine-update.md	R8	T-442	no	S	DONE (libxul -35%, code -43%)
T-446	Settings: a switch for CSS :has() (off, as upstream releases ship it)	cavekit-preferences-ui.md	—	—	yes	S	DONE (code)
T-447	Device: full variant push of the new engine to all three; the §7 checklist; the modern-web page before (old engine: no adoptedStyleSheets, URL.parse, revert-layer, :has) and after	cavekit-engine-update.md	R8	T-445	yes	M	DONE (all three; postinst's read-write test fixed on the way)
T-448	Review (Codex 22:31, or the inspector) of the update and the tooling	—	—	T-447	no	M	DONE (Codex out at 22:32; the inspector — round 22, Tier 82)

Tier 80 — Fonts, as Pale Moon and Basilisk offer them; no Persona by default (2026-10-12; added to the site, not a regeneration)

Owner: "Add a setting for adjusting font like pale moon / basilisk have. ensure software and font licensing / redistribution terms are respected"; "no theme should be applied to any variant by default"; "you can apply themes if you need to for testing purposes just make sure you put the variants back to default afterwards". Kits: cavekit-preferences-ui.md R9, cavekit-personas-ui.md R9.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-438	Settings → Content → Fonts & Colors: fonts per script (proportional, serif, sans-serif, monospace, sizes, minimum), the engine's own list of installed fonts, samples, Restore defaults, a pane link	cavekit-preferences-ui.md	R9	—	yes	M	DONE (device: Enyo effect; all three drawn)
T-439	Licensing: no font shipped; the device's fonts used in place; own MPL-2.0 code	cavekit-preferences-ui.md	R9	T-438	no	S	DONE
T-440	No Persona worn by default; the test Personas taken off in all three	cavekit-personas-ui.md	R9	—	yes	S	DONE (device, all three)
T-441	Card pictures made in the background, one at a time (device: the per-request budget never finished them; decodes failed under cache pressure)	cavekit-personas-ui.md	R9	—	yes	S	DONE (device, Mojo: all 16)

Tier 79 — The owner's evening list and review round 20 (2026-10-12; added to the site, not a regeneration)

Owner: "add a select all to the navbar" (then: "it wasnt showing in the cut/copy menu upon selecting text in the urlbar"); "the scrolling on about:settings and about:personas lacks momentum and snappiness"; "some personas are still flashing like carbon light 2.0"; "when about:settings or about:personas is open the keyboard should be dismissed at first so the whole page is visible"; "extension settings dont appear correctly (cut off)"; "when the keyboard is closed in about:addons the cursor in the search panel is still there". Kits: cavekit-chrome-look.md R2–R4, cavekit-personas-ui.md R9.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-431	The address field gives up the focus when a Jihad page is opened (all three shells)	cavekit-chrome-look.md	R4	—	yes	S	DONE (device, all three)
T-432	The page's field gives up the focus when the keyboard goes down (daemon)	cavekit-chrome-look.md	R4	—	yes (finger: owner)	S	BUILT
T-433	Small card pictures for installed Personas	cavekit-personas-ui.md	R9	—	yes	M	DONE (device, Enyo)
T-434	An add-on's preferences fit the card	cavekit-chrome-look.md	R2	—	yes	S	DONE (device, Mochi)
T-435	Mochi: LunaSysMgr's gestures (Mojo.handleGesture) reach the adapter; flick velocities whole	cavekit-chrome-look.md	R3	—	yes (finger: owner)	S	BUILT
T-436	Select All in the URL bar's Cut/Copy bubble: not satisfiable (the system draws it); the owner chose select-on-tap in all three	cavekit-chrome-look.md	R4	—	yes	S	DONE (device, all three)
T-437	Inspector round 20: Jihad's 3 P2 + 8 P3 and the site's 2 P2 + 11 P3	—	—	—	no	M	DONE

Tier 78 — The server agent's work order, the gallery's flicker, each shell's buttons, Free Palestine (2026-10-12; added to the site, not a regeneration)

Owner: "The active persona seems to flicker in about:personas. Fix that. Then there is a prompt from a claude agent on the riverstone relay server that you need to consider" (~/Documents/PROMPT-PC.md); "buttons in mochi should be mochi style, buttons in enyo should be enyo style, buttons in mojo should be mojo style"; "draw a 'Free Palestine' persona for distribution with jihad browser". Answers: the flicker is while the page loads; bundle Free Palestine + Aurora Silk; every Jihad page takes the shell's buttons; MIT colour palettes allowed, credited. Kits: cavekit-personas-ui.md R8/R9, cavekit-persona-formats.md R4, cavekit-chrome-look.md (new), cavekit-addons-site.md. Site code in ../Riverstone-Addons (branch jihad).

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-418	about:personas: nothing moves or is redrawn while it loads	cavekit-personas-ui.md	R8	—	yes	S
T-419	Gallery: opens at once with the bundled two; the site searched in place (its full list stays on its page), no duplicates (owner: "bundled + site search only")	cavekit-personas-ui.md	R8	T-418	yes	M
T-430	Gallery: addons.mozilla.org searched through its public API (its pages show this engine a bot check); Install fetched from AMO only, asks first naming it	cavekit-personas-ui.md	R8	T-419	yes	M
T-420	Bundle Free Palestine (owner: "an artistic depiction of palestinian land", then "photo realistic": the CC BY 2.0 photograph "Village of Battir" by Labour Palestine with the flag and the words) + Aurora Silk; the rest to the site's catalogue, credits split	cavekit-personas-ui.md	R9	—	yes	M
T-421	Index colours as a JSON string; the site page's Wear keeps colours too	cavekit-persona-formats.md	R4	—	yes	S
T-422	Momentum on Jihad's pages: no scroll echo while the card streams positions (daemon); Mochi hands flicks to the adapter	cavekit-chrome-look.md	R3	—	yes (finger: owner)	M
T-423	Each shell's buttons on Settings, about:personas, about:addons	cavekit-chrome-look.md	R1	—	yes	M
T-424	about:addons in the webOS style, every function kept	cavekit-chrome-look.md	R2	T-423	yes	L
T-425	Site: kinds and colours (migration 001, importer, feed, page, update service), the catalogue's palettes	cavekit-addons-site.md	R5, R6	T-420	no	L
T-426	Site: discover pane in the Jihad skin; panel text; fail2ban include; DEPLOY.md	cavekit-addons-site.md	R7, R10	—	no	M
T-427	Device, against the local copy: every kind from the gallery and the page — previews, Wear asks first, listed in about:addons, colours applied; the refusal of a theme with neither picture nor colour	cavekit-persona-formats.md	R1, R4	T-419, T-421, T-425	yes	M
T-428	Pack the site and the catalogue, write the reply to the server agent, one wormhole code	cavekit-addons-site.md	R10	T-425, T-426, T-429	no	S
T-429	Review: Codex (credits back 5:29 PM) or the inspector, and fixes	—	—	T-418..T-427	no	M

Tier 77 — Settings: a clear way to Personas, extensions and plugins (2026-10-12; added to the site, not a regeneration)

Owner: "add a clear route to about:personas and extensions in browser settings on all 3 variants"; then no extension named as the row's subheading, a Plugins row with the Flash Player's details, and one button per row. Kit: cavekit-preferences-ui.md R8. One page (about:settings) serves all three shells, so one change covers them; each is checked on the device.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-415	Browser pane opens with "Personas, Extensions and Plugins", one button each: Browse Personas (about:personas), Manage Extensions and Manage Plugins (about:addons lists); what is worn, how many extensions, the Flash Player's version and state	cavekit-preferences-ui.md	R8	—	no	S	done
T-416	Device: the group and each button, from each shell's own Settings entry, in all three variants	cavekit-preferences-ui.md	R8	T-415	yes	S	done
T-417	Review (Codex; the inspector when it is out of credits) and fixes	cavekit-preferences-ui.md	R8	T-416	no	S	done (inspector round 19; Codex owed)

Tier 76 — Jihad's add-on site, addons.riverstonerelay.org (2026-10-11; added to the site, not a regeneration)

Owner, choosing among the Persona-index options: a Phoebus site on the owner's domain, the owner's Devuan/nginx/OpenRC server, the full feature set with Personas first, owner-only content, originals + vetted CC, a native feed in about:personas plus the site's own buttons, updates on demand, a Palm-era skin, built locally first. Kit: cavekit-addons-site.md. Site code: ../Riverstone-Addons (Phoebus fork). Nothing is deployed, pushed or put in DNS without the owner.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-400	Fork hygiene: README/NOTICE, licences, Moonchild skins/apps/mail/bans out of the tree	cavekit-addons-site.md	R1	—	no	M	done
T-401	PHP 8.2/8.4: error handler, E_STRICT, Smarty 4.5.8, SafeMySQL, librdf, mysqli, Persona preview fixes	cavekit-addons-site.md	R2	T-400	no	M	done
T-402	Local stack: rootless containers (nginx, PHP-FPM 8.2 and 8.4, MariaDB), schema, owner account, up/down	cavekit-addons-site.md	R9	T-401	no	M	done
T-403	Hardening: config out of the root, registration and special gone, panel path-only, CSRF, escaping, SQLi, downloads, headers, rate limit	cavekit-addons-site.md	R3, R8	T-402	no	L	done
T-404	The Jihad application: id, UA token, features, https URLs	cavekit-addons-site.md	R4	T-402	no	S	done
T-405	Personas: version/licence/source columns, importer from personas.json (sha256), safe page, AUS with version, previews at import	cavekit-addons-site.md	R5	T-403, T-404	no	M	done
T-406	Feed component ?component=personas (paged, UTF-8 search, escaped JSON, cache headers)	cavekit-addons-site.md	R6	T-405	no	M	done
T-407	Palm-era skin base/skin/jihad/, front page, error pages; TouchPad/desktop/phone layouts	cavekit-addons-site.md	R7	T-404	yes	L	done (on-device landscape: the owner's)
T-408	Test CA for the device, in a test copy of a profile, removed after; site reachable from the TouchPad	cavekit-addons-site.md	R9	T-402	yes	S	done
T-409	Browser prefs: getAddons/discover/update URLs, getMoreURL, galleryURL, the site's install permission, offline behaviour	cavekit-addons-site.md	R11	T-404	yes	S	done (getMoreURL unset by design)
T-410	about:personas: the site's Personas paged, search, More on the web	cavekit-addons-site.md	R12	T-406, T-409	yes	M	done (typed search: owner's finger test)
T-411	Updates on demand: Personas (Check for updates) and add-ons (about:addons) against the site	cavekit-addons-site.md	R13	T-405, T-409	yes	M	done
T-412	Codex review of the site and the browser half; fixes	cavekit-addons-site.md	R3	T-403..T-411	no	M	done by the inspector (Codex owed when credits return)
T-413	Deploy checklist and scripts for Devuan/nginx/OpenRC/PHP-FPM/MariaDB/certbot, backups	cavekit-addons-site.md	R10	T-412	no	M	done
T-414	[OWNER] Server go-ahead, DNS for addons.riverstonerelay.org, root steps, smoke test	cavekit-addons-site.md	R10	T-413	human	S	owner — BLOCKED 2026-10-26 on one step: a colour Persona worn from the live site in Pale Moon on a desktop (the site is live and current at f9bcb79, DNS and root steps done; the smoke test 43/43 from the PC on 2026-10-26; on the TouchPad Dracula was worn from the live site by finger and taken off)

Tier 75 — More Persona formats (2026-10-11; added to the site, not a regeneration)

Owner's choice after the format survey: colour-only themes, layered pictures, more colour slots. Kit: cavekit-persona-formats.md.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-390	Extras beside a Persona (profile file, cleaned, deleted with it); colour-only packages get a drawn picture	cavekit-persona-formats.md	R1, R4	—	yes	M	done
T-391	Layered pictures: every picture kept (bounded), placed and tiled; the shells and the pane draw them	cavekit-persona-formats.md	R2	T-390	yes	M	done
T-392	Colour slots in the shells: address field, menus, second bar, glyphs, start page; put back for classic Personas	cavekit-persona-formats.md	R3	T-390	yes	L	done
T-393	Travel: getTheme, channel limit order, Undo/Default/uninstall/editions, index colour slots	cavekit-persona-formats.md	R4	T-390..T-392	yes	S	done (Undo/Default/editions: code only)
T-394	Device verification in all three shells; screenshots; review	cavekit-persona-formats.md	R1–R4	T-390..T-393	yes	M	done

Tier 74 — Engine update readiness (2026-10-08; groundwork only)

Owner: "lay the groundwork for a uxp/goanna update by making our interaction with it more modular and maintainable by humans." Kit: cavekit-engine-update.md.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-380	Branch is the source of truth; patch-queue.sh --check/--add; repair the drift	cavekit-engine-update.md	R1	—	no	M	done
T-381	engine-version.env + check-engine-pins.sh	cavekit-engine-update.md	R2	—	no	S	done (generation open)
T-382	EngineCompat.h: every jihad_* export, startup report	cavekit-engine-update.md	R3	—	yes	S	done
T-383	check-engine-api.py + baseline	cavekit-engine-update.md	R4	T-382	no	M	done
T-384	docs/ENGINE.md; ENGINE-SOURCE.md brought up to date	cavekit-engine-update.md	R5	T-380..T-383	no	M	done
T-385	Wrappers for the positional engine calls	cavekit-engine-update.md	R3	T-382	yes	M	DONE 2026-10-18 (the wrappers landed with Tier 81 as render/goanna/EngineCalls.h; build/engine/check-engine-calls.sh now guards them: the wrapped methods are read from the header and any direct call in render/ fails it; it found one, test/embed_load.cpp's LoadURI, now engine_calls::LoadURIPlain, compiled in the desktop container; 19 call sites)
T-386	JS shim for engine modules and about:addons markup	cavekit-engine-update.md	R4b	—	yes	M	DONE 2026-10-18 (render/goanna/modules/JihadEngine.jsm, installed as resource://gre/modules/JihadEngine.jsm by the bundle, the desktop build and the engine-update push: LightweightThemeManager, AddonManager, PermissionsUtils and about:addons' ids, row anonids/classes, view controller and URLs; jihadPersonas.js and jihadBundledAddons.js go through it; check-engine-api.py reads its SURFACE block (one list) and fails a component importing an engine module around it; test/jihad-engine-shim-test.js passes; device (Enyo): about:addons shows "Personas" and "Uninstall", the Personas view lists "the 14 you have" and Default look in use, no errors)
T-387	Shared component install; derived push list; desktop tests for the uncovered paths	cavekit-engine-update.md	R6	T-380	no	L	DONE 2026-10-18 (build/install-components.sh for the bundle and the desktop dist (the bundle identical); engine-loose-files.py derives the push list (four engine files now go that did not); the desktop engine on RB_20261006 (clobbered; patch 0043 guards the Palm NPEvent code; build-goanna.sh stops on a failed mach step); personas 7/7 and media 18/18 desktop tests with negative controls; prefsui brought up to date, 38/38; the daemon builds for desktop again (the stall sampler's PC per architecture, jemalloc's purge weak, the clock-hold flag in EngineHost))

Tier 73 — Personas UI on about:addons (2026-10-08; added to the site, not a regeneration)

Owner: "make it user friendly to manage and install personas, using about:addons as a base." Kit: cavekit-personas-ui.md.

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort	Status
T-365	Rows show the Persona's art, author, version (1.0 default), In use first	cavekit-personas-ui.md	R1	—	yes	M	done 2026-10-08
T-366	Wear / Default look / Uninstall (picture deleted), finger-sized controls	cavekit-personas-ui.md	R2	T-365	yes	M	done (Default look's Wear untapped)
T-367	Undo bar for 30 s after any change (Try dropped: owner "duplicative")	cavekit-personas-ui.md	R3	T-366	yes	S	done
T-368	Install from file; Get Personas; additional_backgrounds; wallpaper fit	cavekit-personas-ui.md	R4	T-366	yes	M	done
T-369	Intro / empty state; Settings → Appearance summary + Manage Personas	cavekit-personas-ui.md	R5	T-365	yes	S	done (empty state code-only)
T-370	Device verification in all three shells; screenshots	cavekit-personas-ui.md	R6	T-365..T-369	yes	M	done but the owner's finger pass
T-371	Bundled Frutiger Aero set: drawn, shipped, offered once, editions replace	cavekit-personas-ui.md	R7	T-365	yes	M	done

Tier 71 — Tabbed browsing (GROUNDWORK ONLY, 2026-10-06; every task PENDING by the owner's word: "lay the groundwork … but don't implement it yet")

Kit: context/kits/cavekit-tabbed-browsing.md. Design: docs/TABS-GROUNDWORK.md (plan B: pages in the daemon, one connection per card, tab commands over the variant's own Luna service; the YAP contract untouched).

Task	Title	Cavekit	Requirement	blockedBy	Device	Effort
T-347	Preferences jihad.tabs.{enabled,openLinksIn} (openLinksIn default tab) as real engine prefs with rows on the settings page; read by the daemon.	cavekit-tabbed-browsing.md	tabs R5	—	yes	S
T-348	Daemon tab set per connection: TabSet/Tab in JihadBrowserServer, sink by pointer and attachBuffers/detachBuffers in BrowserPageGoanna, maybePaint no-op when unbound, docShell->SetIsActive on bind/unbind; tabs/new, tabs/switch, tabs/close, tabs/list on the own Luna service answered for the focused card only; the tabs notification; the switch sequence (sentinel, geometry, title/url, history, load state, frame); reap on disconnect.	cavekit-tabbed-browsing.md	tabs R2	T-347	yes	L
T-349	Recording sink for background tabs: title/url/load/history recorded and replayed on switch; engine dialogs from a background tab held in the daemon (the prompter waits on its pipe) and shown on switch; downloads/createPage attributed by docshell key.	cavekit-tabbed-browsing.md	tabs R3	T-348	yes	M
T-350	Tab strips in the three shells (Enyo from Atlas's AtlasTabStrip, Mochi inline strip, Mojo row above the command row), responsive and horizontally scrolling, no cap: taps → pageFocused(true) then the Luna call; "Open in New Tab" on the link long-press; window.open → new tab; strip invisible with one tab; "New Card" untouched.	cavekit-tabbed-browsing.md	tabs R1, R4	T-348	yes	L
T-351	Memory on the device: the pressure point with typical pages; the LRU unload step for background tabs (url/title/scroll/history kept, dimmed in the strip, reload on switch); twenty tabs without the daemon dying; idle rules with twenty idle tabs.	cavekit-tabbed-browsing.md	tabs R3	T-349, T-350	yes	M
T-352	Test hooks (T, tN, xN, ?) in the three hooks and card-selftest.sh; device verification of the kit's R6 list by hook, with frame dumps against colour-coded pages after each switch.	cavekit-tabbed-browsing.md	tabs R6	T-350	yes	M
T-353	Review (Codex adversarial, inspector as backup) of T-348–T-350; fixes.	cavekit-tabbed-browsing.md	tabs R2/R3	T-352	no	M
T-354	[OWNER] Finger: tab taps, close controls, "Open in New Tab", the feel of a switch.	cavekit-tabbed-browsing.md	tabs R6	T-353	yes	S

Summary

Tier	Tasks	Effort
Tier 0	15	7S / 8M
Tier 1	1	1S
Tier 2	1	1S
Tier 3	2	2S
Tier 4	1	1M
Tier 5	1	1M
Tier 6	2	2S
Tier 7	1	1S
Tier 8	2	2M
Tier 9	1	1M
Tier 10	1	1M
Tier 11	1	1S
Tier 12	1	1S
Tier 13	1	1M
Tier 14	1	1M
Tier 15	1	1M
Tier 16	1	1M
Tier 17	2	1S / 1M
Tier 18	1	1S
Tier 19	1	1M
Tier 20	1	1M
Tier 21	2	1S / 1M
Tier 22	2	2S
Tier 23	1	1M
Tier 24	1	1S
Tier 25	1	1M
Tier 26	1	1M
Tier 27	1	1M
Tier 28	1	1M
Tier 29	1	1M
Tier 30	1	1S
Tier 31	2	1S / 1M
Tier 32	1	1S
Tier 33	1	1M
Tier 34	2	2M
Tier 35	2	1S / 1M
Tier 36	2	2M
Tier 37	2	1S / 1M
Tier 38	1	1M
Tier 39	1	1M
Tier 40	1	1M
Tier 41	1	1S
Tier 42	1	1M
Tier 43	1	1M
Tier 44	1	1M
Tier 45	1	1S
Tier 46	1	1M
Tier 47	1	1M
Tier 48	1	1M
Tier 49	1	1S
Tier 50	1	1S
Tier 51	1	1M
Tier 52	1	1M
Tier 53 (BLOCKED-EXTERNAL)	13	10S / 3M

Total: 91 tasks, 54 tiers (41S / 50M / 0L); 51 device-lane tasks, 13 BLOCKED-EXTERNAL.

Device lane

The single order in which agent-run device tasks execute (each is blockedBy its predecessor). Priority order: housekeeping and the batch-1 push, <video>, the idle daemon, the blank-page storm, fresh install / Luna registration / Atlas, TLS, media controls, then the remaining partials and the CE 3.1.0 sweep, then batch 2.

T-201 -> T-219 -> T-220 -> T-221 -> T-222 -> T-224 -> T-225 -> T-227 -> T-228 -> T-229 -> T-230 -> T-231 -> T-232 -> T-233 -> T-234 -> T-235 -> T-237 -> T-238 -> T-239 -> T-240 -> T-242 -> T-244 -> T-245 -> T-246 -> T-247 -> T-248 -> T-249 -> T-250 -> T-251 -> T-252 -> T-254 -> T-255 -> T-256 -> T-258 -> T-260 -> T-262 -> T-264 -> T-265 -> T-266 -> T-267 -> T-268 -> T-269 -> T-270 -> T-271 -> T-272 -> T-273 -> T-274 -> T-275 -> T-276 -> T-277 -> T-293 -> T-278

Coverage Matrix

Cavekit	Req	Criterion (abbreviated)	Task(s)	Status
cavekit-engine-embedding.md	R3	No busy-wait/spin; CPU idle when no work is pending	T-203, T-205, T-225, T-228, T-229	COVERED
cavekit-device-build.md	R3	Every declared input obtainable from a clean clone	T-208, T-275, T-279	COVERED
cavekit-device-build.md	R4	Cert/dialog/download flows function with the device services	T-244, T-280	COVERED
cavekit-device-build.md	R4	Fresh install on CE 3.1.0 via the supported path, all three variants	T-201, T-231, T-232, T-233, T-234	COVERED
cavekit-device-build.md	R4	Luna service registered after a fresh install without a hand restart	T-206, T-232, T-233, T-234, T-236, T-277	COVERED
cavekit-device-build.md	R4	Coexistence with Atlas (identity, md5s, both serve, own connects)	T-230, T-231, T-235, T-273	COVERED
cavekit-device-build.md	R4	Sweep - this kit (R1-R5, R7-R10 on CE)	T-271, T-231, T-232, T-233, T-234, T-244, T-272, T-246, T-273	COVERED
cavekit-device-build.md	R4	Sweep - cavekit-offscreen-rendering.md	T-264, T-265	COVERED
cavekit-device-build.md	R4	Sweep - cavekit-gre-widgets.md (binding, degradation, date, richlistbox/tree)	T-221, T-262	COVERED
cavekit-device-build.md	R4	Sweep - cavekit-addons-extensions.md	T-248, T-249, T-245, T-246, T-273	COVERED
cavekit-device-build.md	R4	Sweep - cavekit-preferences-ui.md	T-250	COVERED
cavekit-device-build.md	R4	Sweep - cavekit-input-bridging.md	T-256, T-260, T-262	COVERED
cavekit-device-build.md	R4	Sweep - card front-ends (ui-shell, mochi-ui, mojo-ui)	T-233, T-266, T-267, T-268	COVERED
cavekit-device-build.md	R4	Sweep - daemon-side kits (services, ipc, engine, navigation)	T-269, T-270	COVERED
cavekit-device-build.md	R4	Sweep - human-only residue	T-280, T-281, T-282, T-283, T-284, T-285, T-286	COVERED
cavekit-device-build.md	R8	Daemon log bounded WHILE the daemon runs	T-204, T-227	COVERED
cavekit-device-build.md	R8	An idle daemon is idle	T-201, T-203, T-225	COVERED
cavekit-device-build.md	R9	Remaining cross-ABI surfaces enumerated (two device items)	T-207, T-245	COVERED
cavekit-device-build.md	R10	Whether the boost should extend beyond plugins (media)	T-224, T-226, T-277	COVERED
cavekit-input-bridging.md	R2	Tapping an editable raises the VKB on EVERY page; no scroll-off	T-258, T-261, T-282	COVERED
cavekit-input-bridging.md	R2a	Left/Right by code point; Home/End; Up/Down by line	T-214, T-255, T-282	COVERED
cavekit-input-bridging.md	R2a	Programmatic edit dispatches a bubbling DOM input event	T-209, T-255, T-257, T-277	COVERED
cavekit-input-bridging.md	R3	A pinch changes page zoom; a tap maps to a click	T-283	COVERED
cavekit-input-bridging.md	R7	Keys swallowed for content when a PLUGIN has focus	T-211, T-288	COVERED
cavekit-input-bridging.md	R7	Verified on device with an interactive plugin (no leak; chrome regains keys)	T-274, T-288	COVERED
cavekit-input-bridging.md	R7	VKB can be raised deliberately for a text-wanting plugin	T-212, T-263, T-289	COVERED
cavekit-gre-widgets.md	R1	Controls render into the offscreen surface (thumb off 0)	T-238	COVERED
cavekit-gre-widgets.md	R1	Play/pause operable by tap; state changes	T-222, T-285	COVERED
cavekit-gre-widgets.md	R1	Codec reality measured and written down	T-221	COVERED
cavekit-gre-widgets.md	R1	A	T-202, T-216, T-219, T-220	COVERED
cavekit-gre-widgets.md	R1	The in-player volume control works (input + output)	T-239, T-241, T-285	COVERED
cavekit-gre-widgets.md	R1	The scrubber tracks playback and seeks, on a clean install	T-238, T-285	COVERED
cavekit-gre-widgets.md	R1	Hardware volume keys: level changes AND HUD shown, or a named limit	T-240, T-286	COVERED
cavekit-gre-widgets.md	R3	platformHTMLBindings are REACHED for a focused text field	T-213, T-282	COVERED
cavekit-gre-widgets.md	R3	Hand-rolled keycode map reduced to the VKB's own codes	T-213, T-255, T-282	COVERED
cavekit-gre-widgets.md	R3	No regression to typing on device	T-214, T-255, T-282	COVERED
cavekit-gre-widgets.md	R5	Non-modal messages use the toast (three shells, add-on toast)	T-242, T-249	COVERED
cavekit-gre-widgets.md	R7	Every content-facing capability working or off with the reason	T-220, T-221, T-223	COVERED
cavekit-browser-services.md	R5	Negotiated TLS version CAPTURED from a real page load	T-237	COVERED
cavekit-browser-services.md	R5	Platform cert store path still works on CE 3.1.0	T-244	COVERED
cavekit-mochi-ui.md	R2	Parity checklist vs ../app [human-review]	T-210, T-291	COVERED
cavekit-mojo-ui.md	R7	Mojo chrome-owned settings are editable	T-254, T-281	COVERED
cavekit-addons-extensions.md	R3	Mismatched targetApplication rejected with a clear reason	T-249	COVERED
cavekit-addons-extensions.md	R7	Flash holds a SMOOTH sustained composite frame rate	T-247, T-287	COVERED
cavekit-addons-extensions.md	R7	Keyboard reaches a focused plugin and not the card chrome	T-274, T-288	COVERED
cavekit-offscreen-rendering.md	R3	Invalidations coalesced; no redundant full repaints	T-205, T-228	COVERED
cavekit-offscreen-rendering.md	R8	A genuinely blank page SETTLES	T-205, T-228	COVERED
cavekit-preferences-ui.md	R2	Every control maps to a real engine pref - no decorative toggles	T-250, T-290	COVERED
cavekit-preferences-ui.md	R2	The four frozen-contract settings present and correct	T-251, T-253, T-278	COVERED
cavekit-preferences-ui.md	R3	A change takes effect without a daemon restart, on a real page	T-251, T-278	COVERED
cavekit-preferences-ui.md	R5	Frozen YAP settings commands keep working	T-215, T-252	COVERED
cavekit-preferences-ui.md	R6	Controls hit-testable by touch at 1024x768	T-284	COVERED
cavekit-preferences-ui.md	R6	Keyboard entry works in any text field the page offers	T-250, T-284	COVERED
cavekit-ui-shell.md	R7	Hold on text selects the word; menu offers Copy / Select All / page actions	T-332, T-333	COVERED
cavekit-ui-shell.md	R7	Hold on a field: focus + caret; Paste / Select All, Copy / Cut once selected	T-332, T-333, T-345	COVERED
cavekit-ui-shell.md	R7	A press inside the selection keeps it (Select All → Copy copies the page)	T-333	COVERED
cavekit-ui-shell.md	R7	A page handling its own contextmenu gets no second menu	T-332	COVERED
cavekit-ui-shell.md	R7	Cut deletes only once the card has the text, whole; the card is told	T-333, T-345	COVERED
cavekit-ui-shell.md	R7	Inside the frozen contract (adapter's selectAll/copy/cut/insertStringAtCursor)	T-332	COVERED
cavekit-ui-shell.md	R7	A real finger on text / field / link in all three shells	T-341	COVERED (owner)
cavekit-ui-shell.md	R7	Fields inside iframes get the page menu (known gap, owner's call)	T-341	COVERED (owner)
cavekit-ui-shell.md	R8	pageCertificate answered to the own card only, in the reply	T-335	COVERED
cavekit-ui-shell.md	R8	Padlock opens the certificate view with fingerprint / connection / trust rows	T-335	COVERED
cavekit-ui-shell.md	R8	No padlock over a failed load	T-335, T-338, T-342	COVERED
cavekit-ui-shell.md	R8	Several cards: the answer is for the card that asked (pageFocused)	T-336	COVERED
cavekit-ui-shell.md	R8	Mojo padlock tap; Mochi padlock after a failed load	T-342, T-345	COVERED
cavekit-offscreen-rendering.md	R9	Region uses the whole segment; a page that fits is painted whole	T-334	COVERED
cavekit-offscreen-rendering.md	R9	Frames are composed; ≤ 768 new rows a frame; maintenance frames	T-334	COVERED
cavekit-offscreen-rendering.md	R9	Off-screen rows swept once loaded and still, with pauses	T-334	COVERED
cavekit-offscreen-rendering.md	R9	Every row of a composed frame is row-exact after jumps	T-334	COVERED
cavekit-offscreen-rendering.md	R9	Zoomed in, the region is as wide as the content up to two viewports	T-334, T-343	COVERED
cavekit-offscreen-rendering.md	R9	A real pan and pinch judged by eye	T-341	COVERED (owner)
cavekit-offscreen-rendering.md	R9	Wide regions seen on the device	T-343	COVERED
cavekit-addons-extensions.md	R9	Palefill 1.30 vendored byte for byte, licence and hash	T-337	COVERED
cavekit-addons-extensions.md	R9	Offered once per profile through the add-on manager	T-337	COVERED
cavekit-addons-extensions.md	R9	Installs enabled (AppCompat rule)	T-337	COVERED
cavekit-addons-extensions.md	R9	It does something (debug output on a fixed site)	T-337	COVERED
cavekit-addons-extensions.md	R9	Known and accepted: the settings window it cannot open; whole-response buffering	T-337	COVERED
cavekit-mojo-ui.md	R6	Long-press menu and padlock in the Mojo shell	T-344, T-345, T-341	COVERED
cavekit-addons-extensions.md	R10	Personas: site install with confirm, painted on all shells with ink glyphs, AMO static themes, preview gating, Appearance section; tabs themed	T-355 (tabs: T-350)	COVERED
cavekit-mojo-ui.md	R6	Upper-right menu options on the circles; reload in the url field; app menu; pinch	T-357, T-358, T-361, T-364	COVERED
cavekit-mochi-ui.md	R2	Upper-right and app menus = Atlas/Enyo	T-356	COVERED
cavekit-preferences-ui.md	R2	DuckDuckGo Lite default, HTML option	T-362	COVERED
cavekit-addons-extensions.md	R7	Flash pacing for games: one producer, content rate measured	T-359, T-360, T-364	COVERED (eye test pending)
cavekit-tabbed-browsing.md	R1	Tab strip in every shell: switch / new / close / long-press "Open in New Tab" / hidden with one tab	T-350	COVERED (pending)
cavekit-tabbed-browsing.md	R2	Tabs are daemon pages; YAP and the adapter untouched; Luna commands for the own card; switch sequence; state kept; set dies with the card	T-348	COVERED (pending)
cavekit-tabbed-browsing.md	R3	Background tabs quiet and bounded: held dialogs, attribution, cap, memory, idle	T-349, T-351	COVERED (pending)
cavekit-tabbed-browsing.md	R4	Tabs and cards coexist: New Card unchanged, two cards two sets, thumbnail, move to card	T-350, T-352	COVERED (pending)
cavekit-tabbed-browsing.md	R5	Preferences as real engine prefs on the settings page	T-347	COVERED (pending)
cavekit-tabbed-browsing.md	R6	Hooks and device evidence; the finger	T-352, T-354	COVERED (pending)

Coverage: 86/86 criteria (100%) (6 rows added 2026-10-06 for cavekit-tabbed-browsing.md, groundwork only: every task pending.) (27 rows added 2026-10-06 for ui-shell R7/R8, offscreen R9, addons R9 and mojo-ui R6; "(owner)" marks a criterion only a finger can close.)

Dependency Graph

graph LR
    T-201
    T-202
    T-203
    T-204
    T-205
    T-206
    T-207
    T-208
    T-209
    T-210
    T-211
    T-212
    T-213
    T-214
    T-215
    T-202 --> T-216
    T-203 --> T-217
    T-204 --> T-217
    T-205 --> T-217
    T-206 --> T-217
    T-213 --> T-217
    T-214 --> T-259
    T-215 --> T-217
    T-216 --> T-217
    T-211 --> T-218
    T-217 --> T-218
    T-201 --> T-219
    T-216 --> T-219
    T-217 --> T-225
    T-209 --> T-219
    T-219 --> T-220
    T-220 --> T-221
    T-221 --> T-222
    T-221 --> T-223
    T-222 --> T-224
    T-224 --> T-225
    T-224 --> T-226
    T-225 --> T-227
    T-227 --> T-228
    T-228 --> T-229
    T-229 --> T-230
    T-230 --> T-231
    T-218 --> T-231
    T-231 --> T-232
    T-232 --> T-233
    T-233 --> T-234
    T-234 --> T-235
    T-234 --> T-236
    T-235 --> T-237
    T-237 --> T-238
    T-238 --> T-239
    T-239 --> T-240
    T-239 --> T-241
    T-240 --> T-242
    T-226 --> T-243
    T-241 --> T-243
    T-242 --> T-244
    T-244 --> T-245
    T-207 --> T-245
    T-245 --> T-246
    T-246 --> T-247
    T-247 --> T-248
    T-248 --> T-249
    T-249 --> T-250
    T-250 --> T-251
    T-251 --> T-252
    T-215 --> T-252
    T-251 --> T-253
    T-252 --> T-254
    T-254 --> T-255
    T-209 --> T-255
    T-255 --> T-256
    T-255 --> T-257
    T-256 --> T-258
    T-243 --> T-259
    T-236 --> T-259
    T-257 --> T-259
    T-258 --> T-260
    T-258 --> T-261
    T-260 --> T-262
    T-259 --> T-263
    T-212 --> T-263
    T-261 --> T-263
    T-253 --> T-263
    T-262 --> T-264
    T-264 --> T-265
    T-265 --> T-266
    T-266 --> T-267
    T-267 --> T-268
    T-268 --> T-269
    T-269 --> T-270
    T-270 --> T-271
    T-271 --> T-272
    T-272 --> T-273
    T-273 --> T-274
    T-274 --> T-275
    T-275 --> T-276
    T-263 --> T-276
    T-276 --> T-277
    T-277 --> T-278
    T-253 --> T-278
    T-208 --> T-279
    T-244 --> T-280
    T-268 --> T-281
    T-254 --> T-281
    T-258 --> T-282
    T-255 --> T-282
    T-260 --> T-283
    T-250 --> T-284
    T-222 --> T-285
    T-238 --> T-285
    T-239 --> T-285
    T-240 --> T-286
    T-246 --> T-286
    T-247 --> T-287
    T-274 --> T-288
    T-218 --> T-288
    T-276 --> T-289
    T-250 --> T-290
    T-210 --> T-291