Ka0s Absorb Tracker

By
aDd1kTeD2Ka0s

Display your total absorb shield value in a clean, customizable bar interface.

Back to Files

1.11.0-release

File nameAbsorbTracker-1.11.0-release.zip

Uploader: aDd1kTeD2Ka0s
Uploaded: Sep 26, 2026
Downloads: 222
Size: 6.6 MB
Flavors: Retail
File ID: 8985131
Type: R
Release

Supported game versions

12.1.0

Download Install Via App

What's new

commit 3b7e1fd395f3a46a7dc78418da9854b390a07ac3
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sun Sep 27 03:07:50 2026 +0530

Release 1.11.0: version bump, release notes and the gated release run

Release gate (docs/automated-tests/20260927-030336/manifest.json, run with
--release 1.11.0 on 9b5a369, clean tree), all five conditions PASS:
- Lint: pass, 0 warnings / 0 errors in 67 files
- Tests: pass, 810 passed, 0 failed, 0 skipped of 810
- Perf: pass, 6 scenarios from tests/perf.lua
- Complexity: pass, 13377 NLOC / 1919 functions, avg CCN 1.7
- CCN <= 15: 0 warnings, max CCN 14

Files changed:
- AbsorbTracker.toc: ## Version 1.10.0 -> 1.11.0
- core/Namespace.lua: NS.version "1.10.0" -> "1.11.0"
- docs/module-map.md, docs/slash-dispatch.md, docs/smoke-tests.md:
version mentions rolled to 1.11.0
- README.md: new 1.11.0 row at the top of ## Version History
- docs/automated-tests/20260927-030336/: the release bundle, with its
ANALYSIS.md
- docs/automated-tests/RESULTS.md: the release row and refreshed watch-list
dispositions

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 9b5a369a95fe9a9bc8eb0fdc1ad448d75d55b8e2
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sun Sep 27 03:03:28 2026 +0530

README: Usage as the primary setup flow, and sync-docs

- Usage restructured around the four-step setup flow.
- How the bar works now opens on UnitGetTotalAbsorbs and secret values.
- sync-docs fixes in DEPENDENCIES.md and the ARCHITECTURE documentation map.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit b2b4bb72ba5bd6eb002f2b6dbbc53bacf5fbc705
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sun Sep 27 02:28:21 2026 +0530

README: credit the third-party font and icons the addon ships

Credits now names the debug console's JetBrains Mono (SIL OFL 1.1) and the Open Iconic ? and X on the bar's drag handle (MIT). Both ship inside the vendored LibKa0s payload with their license text,
and documentation-§1 item 12 keeps Credits for exactly this kind of external asset (a font, an icon
set), never a library list.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit fc1a82d6c144d0dfea055ca4397f50543f60309e
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sun Sep 27 00:15:29 2026 +0530

README: plain-language pass for readers

Rewrote the prose for the person reading it: long run-on paragraphs split by topic, em dashes
and reflexive bold removed, passive and subjectless sentences given someone doing the action,
and sentence length varied. Every fact, command, setting name, link, badge, heading and the
section order are unchanged, as is every section a test or the tooling cites verbatim.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 4a1178f5fb5659f37c872064b069a9bbf94f7348
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 23:06:49 2026 +0530

AT-ATS-SDR: Repoint the Sl colon-call citation in .luacheckrc to core/AbsorbTracker.lua:52

Review finding on 917014b: line 44 is blank; NS.Slash:Register() is at :52. Comment-only.
Gate: 810/0/0, luacheck 0/0 in 67 files.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 4a1178f5fb5659f37c872064b069a9bbf94f7348
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 23:06:49 2026 +0530

commit 917014b15fc38c45578c58a55dc9cf4a24b80dd2
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 23:02:59 2026 +0530

AT-ATS-SDR: Correct stale comment citations (owner-approved sync-docs follow-up)

- .luacheckrc:118 - degraded-arm stub `SlashLib:New(d)` is at settings/Slash.lua:603 (was :410)
- .luacheckrc:120 - `function lib:New(d)` is at libs/LibKa0s/Slash.lua:484 after the v1.62.0
re-vendor (was :375)
- .luacheckrc:120 - `SlashLib:New({...})` call is at settings/Slash.lua:661 (was :448)

Comment-only; no code change.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 58c04ec625fe46150d185568fe4483cc48f0f119
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 19:44:25 2026 +0530

AT-ATS-SD: Sync docs to the tree after the automated-tests sweep

- docs/module-map.md: the harness paragraph's suite list now matches
tests/run.lua's ordered list. It gains test_draghandle, test_events and
test_diagnostics, and names all four kit suites (test_prose,
test_diagnostics_contract, test_eol, test_layout_cap) instead of test_eol
alone.
- docs/scope.md: the same suite list gains test_events and test_diagnostics
and the three kit suites it left out.
- docs/automated-tests/RESULTS.md: "neighbour" -> "neighbor" in the
test_widgets.lua Disposition prose (localization-§5). The kit's prose gate
was red on it at 0479a09 (809/1/810). No number in the record changed.

Green gate after: lua tests/run.lua 810/0/810, luacheck 0/0 in 67 files,
docs/test-cases.md matches `lua tests/run.lua --list`.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 0479a0906dd3394b1b51e75af9f2401a7ef15c74
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 19:33:01 2026 +0530

AT-ATS-99: Final automated-tests sweep run (20260926-193105)

Closing run of the 2026-09-26 automated-tests sweep, measured at fb55f44 on a
clean tree. Verdict green.

- lint: pass, 0 warnings / 0 errors in 67 files (66 at 20260926-160433; the
new tests/test_panelpages.lua)
- tests: pass, 810/0/810 (unchanged; AT-ATS-01 moved 71 cases verbatim,
now 40 in test_helpers + 31 in test_panelpages)
- perf: pass, 6 scenarios, api/iter and bytes/iter identical in every one
- complexity: pass, 13377 NLOC (13366), 1919 functions (1920), avg NLOC 6.4,
avg CCN 1.7, max CCN 14, avg tokens 46.5, 0 warnings; band 3 -> 2 files,
0 over cap

What moved vs 20260926-160433: tests/test_helpers.lua left the 1000-1500
band (1447 -> 573 lines) after the AT-ATS-01 peel; test_ltrap.lua lost one
function with the LibKa0s v1.62.0 re-vendor. Nothing newly crossed.
RESULTS.md: refreshed the carried Disposition cells for test_slashcmds.lua
(1321) and test_widgets.lua (1073) to cite this run; rulings unchanged.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit fb55f4407efb01275711c0cdb8281e422e5a3900
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 18:57:44 2026 +0530

AT-ATS-01: Peel the rendered-page cases out of test_helpers into test_panelpages

tests/test_helpers.lua was 1447 lines, 53 short of the layout-cap. Its
rendered-page cases now live in a new tests/test_panelpages.lua. They cover
the per-unit Appearance page from settings/UnitPanel.lua (Unit picker,
chrome block, mirror controls, tab strip, the two-tier refresher), the
General page's Reset position button, and /at resetposition beside it. The
split follows the file's existing "The Appearance page" section boundary,
and no moved case reaches into a local defined above it. The cases moved
verbatim and in their original order. The runner loads the new suite
straight after test_helpers because its first case asserts that the
Appearance page has not rendered yet. The L-trap case stays in test_helpers
beside the library contract it pins.

- tests/test_helpers.lua: 1447 -> 573 lines, 71 -> 40 cases
- tests/test_panelpages.lua: new, 894 lines, 31 cases
- Suite total 810 before and 810 after, the same set of case names
(sorted PASS lines compared). docs/test-cases.md was regenerated.
- Comments and docs that pointed at test_helpers for the moved cases now
point at test_panelpages. The module-map, common-tasks and scope suite
lists, the ARCHITECTURE layout census (largest authored file is now
test_slashcmds at 1321) and the RESULTS.md watch-list row are updated.

Gate: 810 passed / 0 failed; lint 0 warnings / 0 errors in 67 files;
complexity 0 warnings; inventory in sync; no authored file over 1500 lines.

Traces: ATS-11, review finding AbsorbTracker-R-14.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 0aceec73619c37ce9b28b5692bc9087a78fcb877
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 18:50:13 2026 +0530

AT-ATS-RVR: Roll the module-map census rows for the LibKa0s v1.62.0 re-vendor

Brings docs/module-map.md in line with what AT-ATS-RV vendored:
- tests/_kit/ row: kit revision 31, 10271 lines, framework.lua 984 and
test_prose.lua 750, and names the new inventory.lua, prose_coverage.lua
and prose_selftests.lua.
- tests/test_ltrap.lua row: 252 lines, and the Options tripwire reads every
Options*.lua off LibKa0s.xml (ten files) rather than three.
- The Options section prose, the settings/ intro and the
settings/OptionsSetup.lua row list the ten Options files; OptionsSetup is
383 lines.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 5042c6b487be0f596f79778dfff52a543694243f
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 18:46:22 2026 +0530

AT-ATS-RV: Re-vendor LibKa0s v1.62.0 and read the Options tripwire off the XML

LibKa0s payload from the local v1.62.0 tag (ATS-20, ATS-21): Options 25 -> 26, OptionsWidgets
31 -> 32, OptionsTabs 5 -> 6, and four new files at minor 1 (OptionsRegistry, OptionsIds,
OptionsIdList, OptionsCombat); Options key 26.1.32.1.1.6.1.7.4.1. tests/_kit moves from kit
revision 27 to 31, which carries the None. line under an empty complexity watch-list table (ATS-20)
and the generated-file carve-out in the band table (ATS-21). CLAUDE.md provenance and
docs/testing.md's version-now line roll with it.

No member, descriptor field or row field changed, so the library-absent stub needs no new member;
its comment now names the id members' new home. tests/test_ltrap.lua's Options tripwire read a
hand-typed list of three files, which the code peeled into OptionsIds, OptionsIdList and
OptionsRegistry would have left unseen; it now reads every Options*.lua off the vendored
LibKa0s.xml (ten files), same case. The library census in module-map, ARCHITECTURE and performance
moves to twenty-seven files with the new load order. docs/test-cases.md regenerated, unchanged.
Bundle docs/revendor/2026-09-26-v1.62.0; no adoption in this item.

Gate: 810 passed, 0 failed (810 before); 0 warnings / 0 errors in 66 files; no function above
CCN 15.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 7f4fa5ac9f9db0fed84435ce21fa53eef2db8c8d
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 16:19:00 2026 +0530

AT-ATS-00: Record the 2026-09-26 automated-tests sweep run (20260926-160433)

The collection-wide sweep's bundle and RESULTS.md, as the sweep left them. The fix queue it
produced is Ka0sAddonsCommonTasks docs/2026-09-26-AUTOMATED_TESTS_SWEEP.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 1c27f7a175f96edd7d373d2f829d121ed8d5dde3
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 14:44:12 2026 +0530

NR-AT-01: Re-vendor LibKa0s v1.61.0 and stub NavRail

LibKa0s payload from the v1.61.0 tag: Options 24 -> 25, OptionsTabs 4 -> 5, the new OptionsNav minor 1 (Options key 25.31.5.7.4.1). tests/_kit is unchanged at kit revision 27, so it is not copied. CLAUDE.md provenance and docs/testing.md's version-now line roll with it. The library-absent stub gains a NavRail no-op for the Options surface-parity case; no page here draws a rail.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 1c27f7a175f96edd7d373d2f829d121ed8d5dde3
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 14:44:12 2026 +0530

commit 13f1e23525372e6a1c796bb82264accef10f0f74
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 09:32:09 2026 +0530

DR-AT-07: Standards refresh to v2.68.0, docs synced, AUD-01 self-check passes

The M4 item of the diagnostics rollout: the revendor-standards sweep
against WowAddonStandards v2.68.0 (local master 2854053), a sync-docs
pass and the AUD-01 self-check, done by hand in this repo only.

revendor-standards: nothing to change.
- Three-place reference intact: AbsorbTracker.toc:12 X-Standard carries
the repo URL; the README badge (README.md:6) is the bare, unlinked
image with _ not %20; CLAUDE.md "Standards compliance (read first)"
keeps stop-and-flag and the two-way classification. No version stamp.
- Retired notation: grep -rEn '§[0-9]+\.[0-9]' . (libs, _kit, audits,
reviews, automated-tests excluded) -> 0 hits.
- 77 distinct filename-§N citations range-checked against the section
files' own "### N." headings: 0 out of range, 0 unknown files
(debug-logging has 14 sections, so the new §14 citations resolve).
- The one documentation-§1 item citation (item 9, test_docs.lua:75 and
module-map.md:902) already names the new Reporting a bug item.
- Retired names (complexity.md, agent-context.md) appear only as
retired; the one "drop-in" hit (settings-panel.md:379) is about AGSMW.
- testing.md and automated-tests/README.md name both checkpoints.
- .gitattributes pins CRLF with the *.sh/*.py carve-outs; every edited
file keeps CRLF (CR count equals LF count).

sync-docs: one AUD-01 gap and three drifts, all closed here.
- AUD-01 step 3 reads "tests/test_disabled.lua dispatches both forms".
Its step-7 walk sent `diagnostics` bare but never `debug diagnostics`
(only the kit's shared contract case did). test_disabled.lua gains
"disabled 7: both diagnostics forms reach RunDiagnostics, each once,
with no refusal". It is a characterization case, since the route
already worked. Falsified by renaming runDebug's `diagnostics` branch
(settings/Slash.lua:367): the new case went red (with the contract
and diagnostics cases), then the file was restored.
- docs/test-cases.md regenerated (809 -> 810); README badge 810/810.
The README change is the badge number only, so the de-AI pass had no
prose to review.
- docs/module-map.md: test_disabled.lua's line count (608 -> 633) and
its step-7 summary name the new case; docs/common-tasks.md's suite row
says the same.
- docs/ARCHITECTURE.md's core/DebugLogSetup.lua row did not mention the
diagnostics descriptor fields or the stub's RunDiagnostics; it now
names `brandName`, the call-time `diagnostics` hook and the stub's
library-absent line (module-map.md already did).
- Checked and unchanged: NS.COMMANDS has 19 rows and ARCHITECTURE says
nineteen verbs; the Documentation map both ways (compat-layer.md is
Not applicable and absent, debug.md Present); every module-map line
count; kit revision 27 (Kit.VERSION); fifteen majors across
twenty-two files; the v1.60.0 provenance line.

AUD-01 (02_SPEC.md), run by hand on this commit, all steps pass
1. grep '"diagnostics"' settings/*.lua core/*.lua: one COMMANDS row
   (settings/Slash.lua:119); runDebug tests `diagnostics` before every
   other word (Slash.lua:367). The third hit is the stub's DebugVerb
   (core/DebugLogSetup.lua:79), not a COMMANDS row.
2. grep -rniE '"(diag|dump|dx)"' settings core modules: empty.
3. liveVerbs is built on SlashLib.LIVE_VERBS plus resetposition and
   profile (Slash.lua:699-704); test_disabled.lua now dispatches both
   forms while disabled.
4. Both forms call NS.DebugLog:RunDiagnostics (Slash.lua:356); no
   :Clear in modules/Diagnostics.lua.
5. The stub's RunDiagnostics prints the library-absent line and writes
   nothing (core/DebugLogSetup.lua:34-39, :71).
6. README ## Reporting a bug (README.md:137) sits between
   Troubleshooting and Issues and feature requests, with the three steps
   and the note verbatim for /at, no destination and no link.
Result: AUD-01 passes (vendor v1.60.0, so no #47/#90).

Gate: lua tests/run.lua 810 passed, 0 failed; luacheck 0 warnings / 0
errors in 66 files; lizard -C 15 -w clean (libs and tests/_kit
excluded); largest Lua file 1447 lines.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 13f1e23525372e6a1c796bb82264accef10f0f74
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 09:32:09 2026 +0530

commit d15ba41ec33f773fabf1a9cf081acf4d750d1811
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 09:25:39 2026 +0530

DR-AT-06R: Rewrap the enUS header comment DR-AT-06 left at 125 columns

DR-AT-06 rewrote the "What IS routed" paragraph at the head of
locales/enUS.lua but left one line at 125 columns, where the rest of
the file wraps at 100. Comment-only reflow; the words are unchanged and
the file stays at 73 lines, so module-map.md's count still holds.

Gate: lua tests/run.lua 809 passed, 0 failed, 0 skipped; luacheck 0
warnings / 0 errors in 66 files; lizard -C 15 -w clean (libs and
tests/_kit excluded); no authored Lua file over 1500 lines.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit d1cbdfb62a271a93f05e4446704441e9962ff801
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 09:23:08 2026 +0530

DR-AT-06: Put a close mark (X) on every bar's drag strip, the player bar included

The owner's ruling (DR-OW-03, X-04): AbsorbTracker adopts
LibKa0s-Widgets-1.0's DragHandle close mark on every strip, player,
target and focus, with no player-bar exception.

modules/Bar.lua hands each DragHandle an onClose, the `close` icon from
the Media seam and its own closeTooltip. A click runs closeBar(unit),
which writes the existing row `units.<unit>.enabled = false` through
NS.SetByPath, the same row General > Bars draws and `/at toggle <unit>`
flips. Its onChange therefore re-syncs the unit's event registrations
and the [Set] trace, the bus and an open panel fire as they do from the
checkbox. It never touches the addon-wide `enabled`, the shared `locked`
or the bar's position, so turning the bar back on puts it where it was.
One chat line names the way back ("Target bar hidden. Re-enable it on
General > Bars or with /at toggle target."), and the X's tooltip carries
the same sentence. No combat rule of its own: the strip only shows
unlocked, and the lock closes in combat. Four new locale keys, ASCII.

Tests (X-05), test_draghandle.lua 22 -> 29, written first and seen
failing (7 failed): the X is present on all three strips with art; the
spec's closeIcon is NS.Icon("close"); Reserve() is RESERVE + HELP_HIT +
CLOSE_GAP on every strip; a click writes exactly the one row through
the seam; position, the other bars and the addon-wide enable are left
alone; the chat line; the X's tooltip. Measure pins re-checked: the
narrow-bar case (test_draghandle.lua:182) compares ApplyWidth against
handle:Measure() rather than a literal, so it absorbs the 36px the X
adds and needed no change; no other test pins a strip width.

Docs: README usage paragraph (the X and the way back) and test badge
809; test-cases.md; module-map.md (the handle paragraph, the Bar,
enUS, MediaSetup and test_draghandle rows and line counts);
ARCHITECTURE.md and common-tasks.md rows; smoke-tests.md item 68(a)
and new item 68b (S6); the MediaSetup and enUS comments naming what
reads them.

Gate: lua tests/run.lua 809 passed, 0 failed, 0 skipped; luacheck 0
warnings / 0 errors in 66 files; lizard -C 15 -w clean (libs and
tests/_kit excluded); no authored Lua file over 1500 lines.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 9f6d52e4e51dd6e225eb676805c0c0d6a70557b5
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 09:18:21 2026 +0530

DR-AT-05R: debug.md's trace-tag table names every tag the console gets

Independent review of DR-AT-05. The new docs/debug.md says the trace
tags table lists the tags NS.Debug writes, but it left out two tags and
understated two others:

- `[Set]` said "one line per bulk copy or reset". LibKa0s-Schema-1.0's
logWrite also writes one `path = value` line for every single setting
write (libs/LibKa0s/Schema.lua:479), and core/AbsorbTracker.lua writes
`[Set]` for a profile copied or reset (:380-:388).
- `[Profile]` was missing: core/AbsorbTracker.lua:369 logs a profile
change.
- `[Launcher]` named only core/LauncherSetup.lua. LibKa0s-Launcher-1.0
logs through the same descriptor sink (Launcher.lua:239).
- `[Cfg]` was missing: LibKa0s-Options-1.0 logs the panel opening and a
register or open held back in combat (Options.lua:1319/:1393/:1399).

Everything else in the commit was checked against the code and holds:
the thirteen sections, their order and tags, the stood-down wording,
the caps (min(1200, 3000 - 100), 40 per list, 200-char wrap, two lines
reserved), the library-absent line, the chat line, the one-key L table,
the smoke section V steps, and no alias. The TDD red was reproduced in a
scratch worktree: with the old ARCHITECTURE.md row, "every Tier 2
documentation-map row agrees with docs/" fails.

Documentation only; the file stays CRLF.

Gate: lua tests/run.lua 802 passed, 0 failed, 0 skipped; luacheck 0
warnings / 0 errors in 66 files; lizard -C 15 -w clean (libs and
tests/_kit excluded); largest authored Lua file 1447 lines.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 0ddccec46ebbb1619ea43429d0cb286c2542e274
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 09:14:48 2026 +0530

DR-AT-05: docs/debug.md documents /at diagnostics; the Tier-2 row flips to Present

debug-logging-§14 "Documented" and documentation-§3's Tier-2 trigger
(standard v2.68.0, STD-20, STD-23): the diagnostics report is a debug
surface of the addon's own, so docs/debug.md now exists.

- docs/debug.md (new): both forms and no alias (`debug diag` toggles the
console, `/at diag` is unknown), the raw-append rule (ungated, flag
untouched, answers while disabled; `debug hold` is the one word that
refuses), the library-absent line, the one localized chat line, the
report's shape (library frame, then the thirteen sections of
modules/Diagnostics.lua in order with their tags and contents), the
stood-down wording, the caps (min(1200, 3000 - 100) = 1200 lines, 40
names per list, 200-char wrap), what it never reads or calls (no
writes, no arithmetic on a secret absorb, no protected API, the AceEvent
registry walked with pairs), the `debug events`/`debug hold` words, the
trace tags, and which to paste.
- ARCHITECTURE.md: the Tier-2 `debug.md` row reads Present, with the
trigger.
- module-map.md: the DebugLogSetup descriptor snippet gains brandName,
diagnostics and the one-key L; the instance surface gains
RunDiagnostics; the stub paragraph names its library-absent line and
points at debug.md.
- smoke-tests.md: new section V (S1-S5, S7, S8, S11 and the
library-absent arm: append, section order, ungated, both forms while
disabled, the long alias, no alias, combat and restricted content, Copy
clean, the 3000-line cap); step 102 no longer claims no descriptor gets
an L; a triage reference for the report.
- common-tasks.md: "Add a section to the diagnostics report" and a
see-also entry; slash-dispatch.md and scope.md point at debug.md.

Already current, checked and left alone: the buffer sites
(ARCHITECTURE:75, module-map :306/:362, smoke-tests :95 read 3000) and
the LibKa0s file count (ARCHITECTURE:49, module-map :757/:910,
performance.md:390 read twenty-two), rolled at DR-AT-01's sync-docs; the
verb counts (19) in slash-dispatch, module-map, schema and scope, rolled
at DR-AT-03.

TDD: tests/test_docs.lua's "every Tier 2 documentation-map row agrees
with docs/" went red when debug.md was added under the old Not applicable
row, and green once the row flipped. Documentation only; no case added,
so test-cases.md and the README badge stand at 802. Every edited file
CRLF.

Gate: lua tests/run.lua 802 passed, 0 failed, 0 skipped; luacheck 0
warnings / 0 errors in 66 files; lizard -C 15 -w clean (libs and
tests/_kit excluded); largest authored Lua file 1447 lines.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 0ede784da21a1f35da143c833e037e2d38d09b1b
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 09:07:36 2026 +0530

DR-AT-04: README gains ## Reporting a bug; Troubleshooting and FAQ point at it

documentation-§1 item 9 (standard v2.68.0, STD-17): a `## Reporting a bug`
section between `## Troubleshooting` and `## Issues and feature requests`,
the standard's text verbatim with `/at` as the slash. Step 3 reads "include
it with your bug report"; the section names no destination and carries no
GitHub link (owner ruling DR-OW-01).

STD-18: Troubleshooting gains the row "Something looks wrong and I want to
report it -> Follow Reporting a bug below." No existing row told players to
paste logs another way; the FAQ "How do I see debug logs?" answer now ends
with a pointer to the section for logs sent with a bug report.

TDD: a new case in tests/test_docs.lua pins the section body to the
standard's text with /at, its place between Troubleshooting and Issues and
feature requests, the absence of any link or "github", and the pointer row.
It was seen failing against the old README first. test-cases.md
regenerated (test_docs 4 -> 5, total 801 -> 802), README badge 802/802,
module-map row 49 and the common-tasks test table describe the new rule.

Verification (plan section 4, -04): the section diffs clean against the
STD-17 block with /<slash> -> /at; no GitHub link in it. De-AI pass
(humanize) on the added prose: no change warranted.

Gate: lua tests/run.lua 802 passed, 0 failed, 0 skipped; luacheck 0
warnings / 0 errors in 66 files; lizard -C 15 -w clean (libs and tests/_kit
excluded); largest authored Lua file 1447 lines; every edited file CRLF.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit a2c89b3f2cea0a4e0ac7d75d1b43a020f47aeeaf
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 09:02:59 2026 +0530

DR-AT-03R: Read the diagnostics media rung off LSM, not off the resolved path

Independent review of DR-AT-03 (33e302b). The media section decided its
fallback flag by comparing the resolved path with C.FALLBACK_*. The shipped
border and font names ("Blizzard Tooltip", "Friz Quadrata TT") resolve under
LSM to the very paths C.FALLBACK_BORDER and C.FALLBACK_FONT hold, and the
"Blizzard" statusbar resolves to C.FALLBACK_TEXTURE. So every default install
would report fallback=yes on the border and font rows even though LSM
resolved them, which points a maintainer at a broken media install that does
not exist.

The row now prints rung=lsm|lsm default|fallback, asked of LSM itself
(Fetch with noDefault, then without): lsm when LSM knows the stored name,
lsm default when LSM answered its own default for a name nobody registered
(a SharedMedia pack uninstalled), fallback when there is no LSM or no answer.
Both Fetch calls are pcall'd. Still read-only; no absorb value is involved.

TDD: the new case in tests/test_diagnostics.lua (a fake LSM mapping the
shipped names onto the fallback literals, and an unregistered bgTexture)
was red before the fix and is green after.

Docs: module-map rows 17a and 55a (line counts and wording), common-tasks
test table, test-cases.md regenerated (800 -> 801), README badge 801/801.

Gate: lua tests/run.lua 801 passed, 0 failed, 0 skipped; luacheck 0
warnings / 0 errors in 66 files; lizard -C 15 -w clean (libs and tests/_kit
excluded); no authored file over 1500 lines; every edited file CRLF.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 33e302b021e24fa6a14798862c7d46be466f52a2
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 08:56:52 2026 +0530

DR-AT-03: Add /at diagnostics on the LibKa0s helper (DX-AT), /at debug events folded in

The diagnostics dump of debug-logging-§14 (standard v2.68.0, DX-AT), built
on LibKa0s-DebugLog-1.0's DebugLogDiagnostics helper (v1.60.0). The library
writes the markers, the identity header, the per-section pcall, the cap and
the chat line; this addon supplies sections only.

- modules/Diagnostics.lua (new, TOC after Timer): thirteen read-only
sections. state (schema stored and code, profile, stored enabled against
the latch, test mode: none), lifecycle (holds, perf probe, bus), settings
(rows off their default as path = value (default); the six globals and
global.minimap.shown always; state.debugConsole), units (switch, mirror,
UnitExists, ShouldShowBar with NS.VisibilityReason and
NS.LastAppliedVisibility), frames (shown/visible, alpha against
GetBarAlpha, scale, size, live vs stored point, default-position flag,
handle), media (LSM name, resolved path, fallback flag), colors (class
token, class-color flags, resolved RGBA), absorb (gated by IsConcatSafe;
a secret prints <secret>), events (AceEvent set read off its registry with
pairs, each unit frame's two events; one stood-down line while down),
rejected events (/at debug events folded in; the topic word stays),
repaint (NS.IsRepaintPending, throttle window, hold remaining), counters
(NS.SessionCounters, since last combat start) and ui (settings panel,
launcher, missing libraries). No setter, publish, hold, timer, Show/Hide,
OpenOptionsPanel or Clear; no compare, conversion or arithmetic on an
absorb value. Loads on a library-less install too.
- core/DebugLogSetup.lua: the descriptor gains brandName
(NS.Constants.BRAND), diagnostics (a closure over NS.Diagnostics.Sections)
and L, a plain one-key table routing DIAG_WRITTEN through NS.L (never the
locale table itself).
- settings/Slash.lua: the diagnostics COMMANDS row straight after debug
(18 -> 19 verbs), its text through NS.L; runDebug tests `diagnostics`
first, before on/off, the topic words and the window toggle. No
diag/dump/dx alias: `debug diag` toggles the window like any unknown word.
The live set picks the verb up from Slash 16's LIVE_VERBS unchanged.
- locales/enUS.lua: the row's help text and the chat line.

Tests (STD-19, TDD: the new cases were red before the code):
tests/run.lua wires Kit.diagnostics, so the kit's test_diagnostics_contract
runs against /at's own dispatcher (both forms, any case, markers with the
brand, append, ungated, while disabled, no alias) in place of its declared
skip. New tests/test_diagnostics.lua (11 cases): the row and its locale key,
every DX-AT section for every unit with none failing and under 150 lines,
changed and always rows, a secret absorb, the stood-down wording, a report
that changes no registration/timer/hold/throttle/flag/panel/console,
rejected events folded in, a raising section costing one line, the capped
ending, the one localized chat line, and the library-absent arm through both
forms on a degraded load.

Falsification: renaming the `diagnostics` branch in runDebug turns four
cases red (three kit contract cases and the library-absent case); restored.

Verification (plan section 4): grep -rniE '"(diag|dump|dx)"' settings core
modules is empty; the kit contract dispatches both forms while disabled.

Docs: test-cases.md regenerated (783 -> 800), README badge 800/800, verb
counts 18 -> 19 (ARCHITECTURE, module-map, schema, scope, slash-dispatch,
smoke-tests step 8), slash-dispatch live-set row, verb table and debug
note, module-map rows (enUS, DebugLogSetup, Diagnostics, Slash, run.lua,
test_diagnostics), ARCHITECTURE module table, common-tasks test table,
testing.md luacheck file count (66), and the Slash.lua line citations in
slash-dispatch and smoke-tests. debug.md and the rest are DR-AT-05; the
README section is DR-AT-04.

Gate: lua tests/run.lua 800 passed, 0 failed; luacheck 0 warnings / 0
errors in 66 files; lizard -C 15 -w clean (libs and tests/_kit excluded);
no authored file over 1500 lines.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit a553c558c8adab3ff4219a85ab19297e382e54a1
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 04:13:28 2026 +0530

DR-AT-02: Read-only diagnostics seams for the visibility ladder, the repaint throttle and the combat counters

The diagnostics report (DX-AT, DR-AT-03) needs four pieces of state that
lived in file-local variables. Each is now published read-only, and none
of them writes anything:

- modules/Display.lua: NS.VisibilityReason(unit) wraps the existing
visibilityReason ladder narration; NS.LastAppliedVisibility(unit)
returns what ApplyVisibility last applied (nil before the first pass).
- modules/Timer.lua: NS.IsRepaintPending() asks whether a coalesced pass
is queued, without arming or canceling one.
- core/AbsorbTracker.lua: NS.SessionCounters() returns a fresh copy of
{ absorbEvents, repaints, lastAbsorb }, counted since the last combat
start while debug was on; lastAbsorb is the last non-secret read.

The modules load on a library-less install too, so the degraded arm is
the same code; a case loads tests/degraded_env.lua and checks all four
answer there. No LibKa0s stub changes, so test_surface_parity is
untouched.

Tests (TDD, red first): four cases in tests/test_visibility.lua, one in
tests/test_timer.lua. Suite 777 -> 782 passed of 783 (1 skipped, as
before); docs/test-cases.md regenerated, README badge and the
docs/module-map.md rows (13, 16, 17, 39, 40: exports and line counts)
updated. luacheck 0/0, lizard -C 15 clean.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit a53a4b291a5467d0d7202a98b54c3a817949f756
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 03:59:26 2026 +0530

docs: sync-docs pass for the LibKa0s v1.60.0 payload

Content sync only; no document created or restructured.

- LibKa0s file count twenty-one -> twenty-two (ARCHITECTURE, module-map,
performance), with DebugLogDiagnostics.lua in the LibKa0s.xml load order
and under LibKa0s-DebugLog-1.0 in module-map.
- The console buffer 1500 -> 3000 lines (DebugLog 14, lib.MAX_BUFFER):
ARCHITECTURE, module-map, and the smoke-test counter lines.
- Reserved verbs live while disabled twelve -> thirteen (Slash 16 adds
diagnostics): ARCHITECTURE, module-map's test_disabled row.
- module-map line counts for the seven files DR-AT-01 changed, kit
revision 26 -> 27, the kit's four declared suites, the new
test_diagnostics_contract suite, and test_debuglog's stub case.

Gate (ka0s-bounded): tests 777 passed, 0 failed, 1 skipped, 778 total;
luacheck 0/0 in 64 files; libs/LibKa0s matches tag v1.60.0 (diff -r and
--strip-trailing-cr both empty).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit a53a4b291a5467d0d7202a98b54c3a817949f756
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 03:59:26 2026 +0530

commit 8152b9dc5069f8266f8a9e2e9b514d4aa71a25b3
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 26 03:52:46 2026 +0530

DR-AT-01: Re-vendor LibKa0s v1.60.0; the DebugLog stub answers RunDiagnostics

Both payloads copied whole from tag v1.60.0 (commit bed0eb1), spanning
v1.59.0: WidgetsDragHandle 2 -> 3, DebugLog 13 -> 14, the new
DebugLogDiagnostics.lua at 1, Slash 15 -> 16, kit revision 26 -> 27.
The CLAUDE.md provenance line and docs/testing.md's restatement move to
v1.60.0 in the same commit.

- core/DebugLogSetup.lua: the library-absent stub gains RunDiagnostics
(the collection's library-absent line naming /at diagnostics, nothing
written, returns 0), BuildDiagnostics (an empty report) and DebugVerb.
The DebugLog parity case went red on the copy alone without them.
- tests/test_debuglog.lua: a case pins the stub's line, count and routing.
- tests/run.lua declares the kit's test_diagnostics_contract suite (one
declared skip until DR-AT-03 wires Kit.diagnostics).
- The liveVerbs builder already reads SlashLib.LIVE_VERBS, so diagnostics
is live while disabled with no code change; its comment and the two
spelled-out live sets (test_disabled step 7, test_slashcmds
LIVE_WHILE_DISABLED) roll to thirteen reserved verbs.
- docs/test-cases.md regenerated; README badge 777/778.
- docs/revendor/2026-09-26-v1.58.0-v1.60.0/: delta, candidates,
decisions and summary. The diagnostics helper and Slash 16 are adopted
in DR-AT-03 and the DragHandle close mark in DR-AT-06; nothing was
declined and no issue was filed.

Gate (ka0s-bounded): luacheck 0/0 in 64 files; tests 777 passed,
0 failed, 1 skipped, 778 total; lizard no function above CCN 15;
largest authored file 1447 lines.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0137w2iyvwbPYwaYca4pi2Bz

commit 257d210fe69a92e85fe2628be1863edf1ced9c9b
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Sep 25 00:13:53 2026 +0530

M6-AT: address review — retire the stale left-click lock/refusal claims
What: comment and doc corrections only; no behaviour change.
- docs/slash-dispatch.md: the /at lock|unlock row no longer calls the
launcher's left click a writer of `locked` (retired rung (b)); the
menu's Locked entry is the launcher writer, left click writes nothing.
- settings/Slash.lua: the degraded stub's DisabledLine rationale and the
Sl:DisabledLine doc comment now name runHold (`/at debug hold`) as the
caller, not the launcher's refused left click, which no longer exists.
- docs/module-map.md: the settings/Slash.lua row says the same; the
settings/Schema.lua row's line count moves 550 -> 551.
- Beyond the review, the same stale claim in settings/Schema.lua's
WRITE_THROUGH comment and in tests/test_slashcmds.lua's refused-unlock
case comment.

Why: LibKa0s-Launcher minor 4 made left click open settings in either
state; these citations described the retired behaviour.

Departures: no new test; the changes are comments and docs only, and
the existing suites already pin left click writing nothing
(tests/test_launcher.lua) and no disabledLine being passed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 257d210fe69a92e85fe2628be1863edf1ced9c9b
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Sep 25 00:13:53 2026 +0530

commit 4fa2f0ffd640ce51b6c807d913b9943574600711
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Sep 25 00:09:31 2026 +0530

M6-AT: Re-vendor LibKa0s v1.58.0; launcher menu is Enabled and Locked

What: copy LibKa0s/ -> libs/LibKa0s/ and testkit/ -> tests/_kit/ whole
from the local tag v1.58.0 (tag 93cf3ad -> commit 34931c9) via git
archive; both diff -r checks are empty. Launcher moves minor 3 -> 4, the
kit stays at revision 26 byte-identical. CLAUDE.md's provenance line and
docs/testing.md's prose restatement roll to v1.58.0. The delta bundle is
docs/revendor/2026-09-25-v1.58.0/ (01_DELTA.md, 05_SUMMARY.md), shaped
like M5's v1.57.0 bundle.

core/LauncherSetup.lua adopts Launcher minor 4 (launcher-§2, standard
v2.67.0): left-click opens settings, right-click opens the options
menu. It passes the two pairs this addon has, matching ADDONS.md's row
(Enabled · Locked): isEnabled/setEnabled and isLocked/toggleLock. Each
toggle runs the slash verb's own NS.COMMANDS handler at click time
(/at enable|disable, /at lock|unlock picked from the stored lock), so
the SetByPath seam, the stand-down latch, the combat unlock refusal and
the echo are the verb's. No test-mode pair (options-ui-§15 exemption)
and no window pair (no primary window). onClick, leftClickLabel and
disabledLine are removed; version and isLocked stay. The unread
"Lock / unlock" locale key goes with leftClickLabel.

Why: the owner's M6 ruling. The copy alone reddened seven cases that
pinned minor 3's buttons and hints, so copy and adoption share this
commit rather than landing a red one.

Tests (written red first): tests/mock_menu.lua, a host fake of
MenuUtil modeled on the library's repo-local one, installed per case.
test_launcher pins left-click -> panel with no write, the menu's title
and exactly Enabled / Locked, per-open state reads, each entry running
the /at verb's handler (spied on NS.COMMANDS), the disabled gray with
the note, the no-MenuUtil fallback and no retired field in the setup
file; tooltip cases re-pin the fixed hints. test_disabled step 8 now
pins the unGated left click and the grayed Locked writing nothing.

Docs: README minimap paragraph and badge (776), launcher.md (two
buttons, options menu), ARCHITECTURE.md, settings-panel.md launcher
row, smoke-tests.md section U, module-map/scope/schema/slash-dispatch/
common-tasks mentions, test-cases.md regenerated.

Gates: lint 0/0 in 64 files, tests 776 passed 0 failed, complexity 0
warnings, largest authored file tests/test_helpers.lua at 1447 lines.
Departures: one commit instead of copy-then-adopt (the copy is red on
its own); disabledLine is dropped too, as launcher-§5 retires it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit a45c97510ffef684952b02fb53b5833c38383a55
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 23:11:28 2026 +0530

M5-AT: Feed the launcher's status tooltip: version, lock, left label

What: core/LauncherSetup.lua passes the Launcher minor-3 descriptor
fields the library's always-on status tooltip reads (launcher-§1, the
owner's M5 ruling): `version` (the TOC's ## Version via NS.Meta),
`isLocked` (the `locked` setting the left click toggles) and
`leftClickLabel` (NS.L["Lock / unlock"], rung (b) as ADDONS.md records
it). isEnabled / disabledLine were already passed and now also feed the
Enabled line and the disabled hint (`Left-click: disabled — /at enable`).
locales/enUS.lua gains the one key.

Not passed, on purpose: `isTestMode` (no Test mode here; options-ui-§15's
exemption makes the lock the preview), `slash` (the hint's command is
read out of NS.Slash:DisabledLine()), and `onTooltipShow` (the addon
never had one and has no lines of its own to append).

Tests first: six new cases in tests/test_launcher.lua drive the object's
OnTooltipShow against a recording tooltip and pin the whole five-line
tooltip, the TOC version and its absence, Locked read on every show with
its colors, no Test mode line, the disabled state, and the locale route.
All six were red on the re-vendored library before the descriptor change.

Docs: launcher.md gains a status-tooltip section; ARCHITECTURE.md's
Launcher paragraph and documentation-map row, settings-panel.md's
Minimap button row and README's minimap paragraph mention it.
docs/test-cases.md regenerated and the README badge moved to 772.

Gates: lint 0/0 in 63 files, tests 772 passed 0 failed, complexity 0
warnings, largest authored file tests/test_helpers.lua at 1447 lines.
Departures: `version` is NS.Meta("Version") rather than NS.Version(), so
an unreadable manifest draws the label alone instead of "v?".

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 310112cfd96b6a98401a090afb26998156f59494
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 23:07:09 2026 +0530

M5-AT: Re-vendor LibKa0s v1.57.0 and record its delta bundle

Copy LibKa0s/ -> libs/LibKa0s/ and testkit/ -> tests/_kit/ whole from the
local tag v1.57.0 (tag d03e836 -> commit aa37bc9) via git archive. Both
diff -r checks against the tag are empty and the runner stays executable.
CLAUDE.md's provenance line moves v1.56.0 -> v1.57.0 in the same commit,
as does the prose restatement of the vendored tag in docs/testing.md.

One library file moves a minor: Launcher 2 -> 3, which makes the LDB
object's OnTooltipShow always the library's status tooltip (launcher-§1,
standard v2.66.0). No file is added or removed, and the kit stays at
revision 26 with identical bytes, so only Launcher.lua changes on disk.

The bundle docs/revendor/2026-09-24-v1.57.0/ (01_DELTA.md, 05_SUMMARY.md)
follows the RV-AT bundle's shape. No contract blocker: this addon never
passed onTooltipShow and no test read the object's hook. The descriptor
adoption (version, leftClickLabel, isLocked) is the next M5-AT commit.

Gates on the copy: lint 0/0 in 63 files, tests 766 passed 0 failed,
complexity 0 warnings, largest authored file tests/test_helpers.lua at
1447 lines. No deviations.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 8a7818a2a38789c5c04330d91f3cf2318ac33931
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 13:26:17 2026 +0530

AT-26: Revendor standards reference against v2.65.0 (no drift found)

Ran the /wow-addon:revendor-standards sweep against WowAddonStandards
v2.65.0 (2026-09-23). The gate was met: the GitHub default branch serves
STANDARDS.md naming v2.65.0, and origin/master is fdcde97. The section
files were read verbatim from that commit because raw.githubusercontent
fetches timed out.

Findings: none. No file changed.
- Three-place reference (documentation-§6): TOC `## X-Standard:` carries
the repo URL, the README badge is the bare unlinked image with `_`
escapes, and CLAUDE.md "Standards compliance (read first)" keeps the
stop-and-flag directive and the two-way classification.
- 3b notation sweep: `grep -rEn '§[0-9]+\.[0-9]' . --exclude-dir=.git
--exclude-dir=libs --exclude-dir=_kit --exclude-dir=audits
--exclude-dir=reviews --exclude-dir=automated-tests` gives 0 hits.
The 126 hits left in frozen audits/reviews/automated-tests bundles
stay there on purpose. All 1102 filename-§N references resolve inside
the fetched section ranges. The only non-matching token is CLAUDE.md's
canonical `filename-§N` placeholder. No doc set is stated as a bare
count, and nothing uses the retired "drop-in" label.
- 3c: every CLAUDE.md pointer exists. docs/ has the canonical trio, the
five record docs and the six Tier 1 docs. No agent-context.md,
complexity.md, file-index.md or conventions.md.
- 3d: the standard has no vendored quirks-catalogue section, so there
is nothing to vendor.
- 3f: testing.md and automated-tests/README.md already name both the
commit and release checkpoints. The RESULTS.md lead-in is generated by
the runner and was not swept.
- 3g: .gitattributes pins `* text=auto eol=crlf` with the *.sh carve-out
and binaries marked. 0 tracked files disagree with the pin. No
renormalization was done.

Deviation: the item text and verify step expect `v2.65.0` in CLAUDE.md,
README.md and AbsorbTracker.toc. documentation-§6's canonical forms
carry no version: X-Standard is the URL, the badge is fixed, and the
CLAUDE block is unversioned. The revendor-standards rules also forbid
writing a standard version into the reference, and the fetched standard
wins. So nothing was stamped, and that verify grep correctly finds no
hit in those three files. The follow-up code MUSTs that v2.65.0 adds
(the minimap.shown path, /at debug hold) belong to their own items and
are not part of this reference roll.

Gates: luacheck 0/0; tests 766 passed, 0 failed; 0 functions above
CCN 15; the largest authored .lua file is 1447 lines.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 31ce2b0cc8c2661f0b5aa9b13104c7d554d3e757
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 11:31:38 2026 +0530

AT-DOCS: address review — drop the retired `test` verb from schema, scope and smoke docs

AT-11 retired `/at test` and NS.COMMANDS (settings/Slash.lua) now has 18
rows. The sync-docs pass fixed slash-dispatch.md but missed three live docs:

- docs/schema.md: the non-key/value verb list no longer names `test`,
and the count is 18 verbs, not 19.
- docs/scope.md: "the 19 verbs" becomes "the 18 verbs".
- docs/smoke-tests.md step 8: lists the 18 verbs, says `/at test` now
answers unknown command, and quotes the `debug` help text as the code
has it ("`on`/`off` logging, `events`, `hold <value> [secs]`").

Docs only. luacheck clean, 766/766 tests pass, no function above CCN 15.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit a99cfe79ad04e7183fa229eec96750874a079398
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 11:25:19 2026 +0530

AT-DOCS: Record the 20260924-112145 automated-test bundle

The kit revision 26 runner (all four suites, through ka0s-bounded) ran
over e84b6d8, clean tree, on the remediation branch:

- lint pass, 0/0 in 63 files
- tests pass, 766/0/766 (docs/test-cases.md and the README badge agree)
- perf pass, 6 scenarios
- complexity pass, max CCN 14, 0 functions above 15, 3 files in the
1000-1500 band, none over the cap

The runner rolled RESULTS.md and widened it once with the Commit and
Tree columns, carrying the eleven earlier rows forward as unknown. This
bundle replaces 20260916-184524 (1080857 on master, 74 commits back) as
the current record (AbsorbTracker-A-20). ANALYSIS.md follows the
AUTOMATED_TESTS.md template. appearancePass fell from 384.2 to 97.8
bytes/iter. A per-commit replay of the offline perf script pins the drop
on 7c43ab4, where the unit label and its font plumbing left
UpdateBarAppearance.

Watch-list Disposition cells, the only hand-edited column:
- tests/test_helpers.lua 1447: peel next by helper group, carried
forward (AbsorbTracker-R-14). History updated with AT-19's +31.
- tests/test_slashcmds.lua 1318: back under the cap, watch, re-check at
1400, next seam the /at profile cases.
- tests/test_widgets.lua 1073: newly in the band after AT-15, accepted,
watch, re-check at 1300.

Deviations: the item expected test_slashcmds.lua at about 1200 after
AT-11. It measures 1318, because AT-02/03/04/09 added slash cases. The
item named two band files and the run found three: test_widgets.lua is
new in the band and has been given a disposition. Not a release run:
no version bump, no Version History entry, no tag.

Findings: AbsorbTracker-A-20, AbsorbTracker-R-14

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit e84b6d895b6d788d6ed51bbf9308978f61691880
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 11:21:35 2026 +0530

AT-DOCS: Sync the docs to the v1.56.0 payload and the M3 changes

The sync-docs pass over README.md, CLAUDE.md, DEPENDENCIES.md and docs/
after RV-AT and AT-01..AT-25. Most of the ripple landed with each item;
this commit fixes what was still stale:

- docs/module-map.md: the tests/_kit/ row still described kit revision
25 (7326 lines). It now reads revision 26 (7820), with current per-file
sizes and the three files revision 26 added: asserts.lua (peeled out
of framework.lua), mock_events.lua and prose_lists.lua. The
tests/test_envsetup.lua row is 120 lines, not 104, and names the
dead-rung case AT-16 added.
- docs/slash-dispatch.md: eighteen verbs, not nineteen (AT-11 moved
`test` under `debug hold`). `debug` now parses a value and a duration
under `hold` and dispatches through DEBUG_VERBS, so the sub-verb-tree
intro no longer says only `profile` carries a table.
- docs/testing.md: the top-level-ignore row reaches 63 linted files, not
54. The re-vendor paragraph names v1.56.0, not v1.35.0, as the tag
CLAUDE.md claims.
- Stale file:line citations re-derived against HEAD: ARCHITECTURE.md
events-frames-taint-§8 row, the UnitGetTotalAbsorbs sites
core/AbsorbTracker.lua:199/:279 -> :225/:305. smoke-tests.md
AbbreviateNumbers sites Display.lua:412 -> :427 and
AbsorbTracker.lua:204/:206/:282 -> :230/:232/:308, UnitClass sites
Data.lua:227 -> :205 and CoreSetup.lua:59 -> :60, and
test_display.lua:813 -> :766. settings-panel.md, test_data.lua:192 ->
:200 for the deferred-closure pin.

Verified unchanged and correct: the CLAUDE.md v1.56.0 provenance line,
the README Tests badge (766/766) and docs/test-cases.md (its diff
against `tests/run.lua --list` is empty), the eighteen NS.COMMANDS rows
against ARCHITECTURE.md and slash-dispatch.md's verb table, and every
other module-map LOC cell against wc -l. The largest authored file is
tests/test_helpers.lua at 1447, and nothing is over 1500.

Comment-citation check over the addon's own source. These were reported
and not corrected, since the check corrects a code comment only on
confirmation:
- tests/test_envsetup.lua:19 cites tests/_kit/mock_base.lua:183 for
the C_AddOns note, which now sits at :1110.
- tests/test_perf.lua:572 cites core/PerfSetup.lua:33 for the
descriptor, which now opens at :35.
- tests/test_schema.lua:400-402 cites OptionsCompose.lua:240/:284/:313
and :182-186, now :324/:368/:397 and :248-255, and
tests/test_data.lua:192, now :200.

The automated-test bundle (AbsorbTracker-A-20) follows in its own
commit so that its manifest records this clean commit.

Findings: AbsorbTracker-A-20, AbsorbTracker-R-14, AbsorbTracker-A-19,
AbsorbTracker-A-27 (item AT-DOCS; the bundle and issue store follow)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 4cd2ffae8f0bce846a2a11c25c4bc31e28391419
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 11:14:55 2026 +0530

AT-25: Record the lapsed re-vendors v1.16.0..v1.54.2 in one span bundle

Adds docs/revendor/2026-09-23-v1.16.0-v1.54.2/ as the consolidated span
bundle that audit-review-history sanctions for a lapsed span. It holds
01_DELTA.md and 05_SUMMARY.md only. Line 1 of 01_DELTA.md follows the
WS-01 grammar and names every tag the bundle covers. The body maps each
tag to its vendoring commit (git log -- libs/LibKa0s tests/_kit, with
the tag read from CLAUDE.md's provenance line at that commit).
05_SUMMARY.md gives one line per tag: carried by sweep, or the adoption
commit's sha and subject.

Resolves AbsorbTracker-A-01.

Deviations from the item text:
- The span is 27 tags, v1.16.0 -> v1.54.2, not the 24 tags
v1.18.0..v1.53.0 ending at v1.55.0. The item says the list must be
whatever the amended AUDIT.md step-4 script prints, and that script
also reports v1.16.0 (the "$horizon 00:00" bound picks up the
horizon's own morning) plus v1.43.0 and v1.54.2 (the kit-only
re-vendors it now walks). The folder is named for the real first and
last tags. v1.55.0 already has its own bundle, so it is left out.
The head -1 | grep | wc -l check therefore counts 29 (27 plus the two
range ends), not 26.
- No frozen bundle's base is misstated, so no correction note was
needed beyond the one about the count. docs/ARCHITECTURE.md and
CLAUDE.md do not list revendor bundles one by one (ARCHITECTURE names
the store once), so neither needed an edit.

Verify: the amended step-4 check prints 0 unrecorded (37 vendored, 37
recorded). luacheck 0/0 in 63 files; tests/run.lua 766 passed,
0 failed; the complexity scan shows no warnings.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 5176809a82192caf519456bd38cccd078200b1be
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 11:04:58 2026 +0530

AT-24: Spill the ARCHITECTURE hub back under documentation-§3's rule

docs/ARCHITECTURE.md was 841 lines (788 at the audit) with Settings
Schema at 123, Message Bus at 73 and Event Subscriptions at 89, all past
the ~60-line spill MUST. It is now 400 lines and no section exceeds 55.
Everything moved, nothing was deleted; each spilled section keeps a
summary and one link.

- Settings Schema (123 -> 29): the hub keeps the row counts, the
runtime name, the NS.SetByPath seam, the no-registry statement and one
sentence per named non-setting state (owner and writers, as
architecture-§5 requires in this section). The runtime binding, the
bulk bracket, the no-registry argument and both named states' full
writer lists moved into docs/schema.md as new sections; the validator
paragraph was already there and was not copied again.
- Message Bus (73 -> 23): the receiver rule and the catalog table stay
(per the audit's C5 design). The Bus-record/stand-down paragraph, the
fan-out, what is not on the bus and the AceDB callback paragraph moved
to a new docs/message-bus.md. Its Conditional row flips from Not
applicable to Present, as the spill target of the hub's ~60-line rule
(the more-than-ten-message trigger is not met: five messages).
documentation-§3 names message-bus.md as this section's spill target,
so no upstream ruling was needed.
- The disabled-state essay (88 -> 7) moved whole to a new Tier 3
docs/lifecycle.md, registered in a new "### Addon-specific
(documentation-§3, Tier 3)" table after "### Verification and record".
- Event Subscriptions (89 -> 45): the bullets are condensed to a summary
with one link to data-flow.md; the facts data-flow.md lacked (the
latch gate at OnEnable, frames built once, the old two-frame packing,
rejected-event reporting, the degraded stub) moved there as a new
subsection. The carve-out subsection stays, as three docs link to it.
The stale "register the monospace font with LSM" OnInitialize step
(moved to core/MediaSetup.lua long ago) is dropped from the summary.

Links into the moved text were repointed (data-flow.md, profiles.md,
schema.md, module-map.md, performance.md). Tests: 766/766 green, case
count unchanged, so docs/test-cases.md is untouched; lint 0/0;
complexity clean. tests/test_docs.lua pins no moved heading, so it
needed no change.

Finding: AbsorbTracker-A-13 (audit AT-75).

Deviations from the item text:
- Three blocks were moved that the item did not name, because the
three it named left the hub at ~480 lines against its <=400 target:
the Launcher section to a new Tier 3 docs/launcher.md; the retired
register rows and "Recorded, but not deviations" to a new Tier 3
docs/recorded-decisions.md (the register keeps a one-paragraph
pointer); and the per-profile stamp's "in full" argument, which
duplicated profiles.md's "The v3 lift, and why the gate is
per-profile", is replaced by the register row citing that section.
Both new docs are registered in the Tier 3 table.
- Paragraphs that restated a topic doc were shortened to pointers: the
Overview's vararg paragraph (module-map.md), the Slash Commands
disabled-gate and no-test-verb paragraphs (slash-dispatch.md), the
Known Limitations English-only bullet (the register row) and the
agent-context.md paragraph (documentation-§3).
- The item cites :118-233 / :234-306 / :400-486 / :630; the file had
grown since the audit, so the moves followed the headings.
- README.md and CLAUDE.md list no doc that moved, so neither changed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 2e941998d1916d49b2e04ce199fa20aeb51f173b
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 10:53:16 2026 +0530

AT-23: Retire the per-unit frames register row as the §1 carve-out

The per-unit RegisterUnitEvent frames meet every condition of the
events-frames-taint-§1 unit-filter carve-out, so the register row that
called them a deviation reads as a deviation that is not one.

- docs/ARCHITECTURE.md: delete the events-frames-taint-§1 row from
"## Documented deviations" (3 rows remain: savedvariables-§1,
events-frames-taint-§8, localization-§1); delete "### The per-unit
event frames, in full"; add "### The per-unit frames are the
unit-filter carve-out" under "## Event Subscriptions", naming each
condition by symbol (one SetScript OnEvent plus RegisterUnitEvent
through NS.SafeRegisterUnitEvent; held at addon.__unitEventFrames;
UnregisterAllEvents in core/Lifecycle.lua StandDown; built once while
__unitEventFrames is nil, then reused). Reword the Event Subscriptions
bullet to cite the carve-out.
- core/AbsorbTracker.lua: comment only, same rewording.
- docs/data-flow.md, docs/module-map.md, docs/midnight-quirks.md: reword
to "the events-frames-taint-§1 unit-filter carve-out" and link the new
design note; midnight-quirks now says one frame per unit (player,
target, focus).

Resolves AbsorbTracker-A-11.

Deviations: the item's line cites had drifted. data-flow.md is :82 (not
:77); module-map.md is :430 and :810 (not :425/:807), and :900 carried
the same "documented events-frames-taint-§1 deviation" wording, so it
was reworded too; midnight-quirks.md's line is :124 (the :120 line is
about unknown event names and is correct as it stands). No "N rows"
sentence existed to fix, and no tests/test_docs.lua pin named the
removed heading. Frozen copies under docs/investigations and
docs/superpowers are untouched.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 146f2b496dde9ff04f7fc521d9c550face51839f
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 10:50:52 2026 +0530

AT-22: address review — PerfSetup rows name the latch, not suspend/resume

The core/PerfSetup.lua rows in docs/ARCHITECTURE.md and docs/module-map.md
(7b) still said the descriptor carries a `suspend`/`resume` pair, which
contradicted the same rows' latch wording and the code: the descriptor passes
`lifecycle = NS.lifecycle`, and StandDown/StandUp live in core/Lifecycle.lua.
Step 0 of NS.ShouldShowBar reads NS.IsStoodDown(); NS.Perf.suspended is a
view of the latch (R-07).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 5d8bde7ff5fc0333c5b3a67589e6eb341ef3bc1f
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 10:46:43 2026 +0530

AT-22: Sweep comment and doc drift against the code

One pass that corrects each stale comment or doc line to what the code
does. Comments and docs only; lint 0/0, 766/766 cases, count unchanged.

- settings/General.lua: the tab diagram draws [Minimap button] where
[Test mode] was; the header says seven rows plus the button pair and
why there is no Test mode row; the console-path comment sits on
DEBUG_CONSOLE_PATH again, below unlockGuard's own comment; the stale
384.5-byte figure gives way to tests/perf.lua's appearancePass; the
throttle note names NS.GetThrottleWindow().
- core/PerfSetup.lua, core/CoreSetup.lua, ARCHITECTURE.md, module-map.md:
the TOC order is CoreSetup -> Lifecycle -> PerfSetup; PerfSetup loads
after core/Lifecycle.lua, whose latch the descriptor carries.
- core/AbsorbTracker.lua: StandUp restores what StandDown tore down (the
Suspend()/Resume() wording is gone; AT-07 had not reworded it).
- defaults/Profile.lua: settings/Appearance.lua, not {Bar,Border,Font}.
- ARCHITECTURE.md and module-map.md: nine files bind addonName (named),
nineteen open `local _, NS`; .luacheckrc names the nine and drops
CoreSetup.
- ARCHITECTURE.md localization-§1 row: v1.42.0, not v1.41.0. Settled by
the vendor: Slash minor 14 (LibKa0s 8512fa1, first tagged v1.42.0) is
the one the refusal moved into, and 224a5df re-vendored v1.42.0.
- module-map.md load order: Lifecycle.lua and LauncherSetup.lua added
to the Core list; file-index line counts rolled for the three edited
Lua files.
- settings-panel.md: the General row names the minimap button.
- A-26: the documentation map names docs/automated-tests/<run>/ and
docs/perf-analysis/<run>/, and the RESULTS.md row says "generated,
except the watch list's Disposition column".
- A-18/A-25: two entries under "Recorded, but not deviations" -- no host
window so no close-button wrapper (standalone-windows), and the
nineteen `local _, NS` files with seven grandfathered self-naming
headers (documentation-§9). The intro now says four entries.

Findings: AbsorbTracker-R-07, AbsorbTracker-A-26, AbsorbTracker-A-18,
AbsorbTracker-A-25.

Deviations:
- The MINIMAP_PATH comment was already next to its constant (AT-12);
verified, nothing to do. The ARCHITECTURE.md size census already reads
1447 lines, which matches wc -l; left as is.
- module-map.md's file index already listed Lifecycle.lua and
LauncherSetup.lua; the gap was the "Load order" Core list, fixed there.
- "Recorded, but not deviations" is a bullet list, not a table, so the
two "rows" are bullets.
- The verify grep still hits core/Data.lua:329, which is the body of
NS.GetThrottleWindow() itself (live code, not a stale comment).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 72dad39e9f74fc3b1da0d04885b6521f256ae63b
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 10:39:24 2026 +0530

AT-21: Use canonical standard citations in the addon's own files

Addon half of AbsorbTracker-A-17: citations of the standard now use the
canonical section-file-§N / bare section-file form.

- localization-5 -> localization-§5 in tests/prose_waivers.lua (:3, :7),
tests/test_docs.lua (:194) and docs/module-map.md (:854).
- .pkgmeta :17, :20, :21: packaging.md:28 -> packaging (comments).
- lint.md -> lint in .luacheckrc (:9, :14, :77), docs/testing.md (:45
heading) and tests/test_lintconfig.lua (:1 comment and the three
failure-message strings at :132, :151, :262).

The kit's test_prose and the addon's test_docs accept the section sign in
these .lua comments (both files were already UTF-8); suite 766/0.

Deviations: the module-map hit is at :854, not :852; .luacheckrc's third
hit is at :77, not :78; the "every other lint.md hit" clause also reached
docs/testing.md and tests/test_lintconfig.lua (message strings, no test
asserts on them).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 42291e177203fa15a110da6fe7241b0f87bd204f
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 10:36:06 2026 +0530

AT-20: Annotate the load-bearing TOC lines and the conventional group

Comment lines only in AbsorbTracker.toc, each directly above its file
line. Five LOAD-BEARING notes say what each file publishes and who
captures it at file load: Namespace.lua (NS.version, into the perf
descriptor), Bus.lua (NS.NewBusTarget / NS.MSG, the guarded UNITS
subscription in core/AbsorbTracker.lua), CoreSetup.lua (NS.Print /
NS.Util.print and NS.LIBKA0S_MISSING), PerfSetup.lua (NS.Perf upvalue in
AbsorbTracker, Display and Timer) and Units.lua (NS.Units.DeepCopy taken
by Database.lua). One "Conventional order, not load-bearing" note covers
State.lua, Data.lua and Database.lua; it sits above State.lua and names
the other two, since the three lines are not adjacent. The Lifecycle
comment is unchanged. The file list itself is unchanged, so
test_loadorder stays green; CRLF preserved.

Findings: AbsorbTracker-A-05, AbsorbTracker-A-06, AbsorbTracker-A-07,
AbsorbTracker-A-08, AbsorbTracker-A-09.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit de4dac307299e8f72bd3d47404613287a1af0e3d
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 10:29:56 2026 +0530

AT-19: Hand three chat lines' parts to the printer; re-measure §8 row

Three pre-formatted chat lines outside the events-frames-taint-§8
register row's scope now hand their parts to NS.Print instead of
building the line first (the §8 SHOULD half):

- core/DebugLogSetup.lua (library-absent stub SetEnabled):
NS.Print("debug logging", on and ON or OFF)
- core/Lifecycle.lua (library-absent hold set PrintHolds):
NS.Print(addonName .. ":", holds or "no holds")
- settings/UnitPanel.lua (RenderUnitPanel failure report):
NS.Print("Unit panel render failed:", err) -- NS.Print SafeToStrings
each part, so the raw err loses nothing.

NS.Print joins parts with one space, so all three lines are
byte-identical to before. One output case per site pins that (none
existed): tests/test_debuglog.lua (degraded SetEnabled on/off acks),
tests/test_disabled.lua (degraded PrintHolds, empty and two holds),
tests/test_helpers.lua (a raise in the unit panel body). Each was
green on the old code and confirmed red under a changed separator
(a part carrying its own trailing space).

docs/ARCHITECTURE.md events-frames-taint-§8 register row: re-measured
with the audit's recorded grep (docs/audits/2026-09-23/03_EVIDENCE.md,
AT-78) at this commit: 19, all in settings/Slash.lua; Schema.lua has
none left. Ripples: README badge and docs/test-cases.md 763 -> 766,
module-map line counts for the three suites, and the layout-cap census
note (test_helpers.lua 1447 lines, still the largest, under the cap).

Finding: AbsorbTracker-A-16 (audit AT-78).

Deviations: the item expected the re-measured count near 17; it is 19
at this commit (AT-03/AT-04/AT-11 added Slash.lua lines), so 19 is
written. The row's recorded grep contains pipes, which would split a
Markdown table cell, so the row cites the evidence file for it
instead of inlining it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit c15d28c64b82ed9a9a93eae0e885622acab943a2
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 10:23:55 2026 +0530

AT-18: Read PARENT_TITLE from C.BRAND; drop schema printer's [AT] arm

settings/OptionsSetup.lua: PARENT_TITLE is now NS.Constants.BRAND (the
local name is kept). Constants loads well before OptionsSetup, so the
brand string has one literal again (AbsorbTracker-R-10).

settings/Schema.lua: chatPrint is `NS.Print(line)`, resolved at call time
so suites can still spy on it. The `elseif DEFAULT_CHAT_FRAME` arm with its
hand-typed '|cFF00FFFF[AT]|r' tag is deleted: CoreSetup defines NS.Print on
both arms before Schema loads, so it could never run, and it was a second
copy of NS.PREFIX that slash-commands-§4 forbids (AbsorbTracker-R-08).

Tests (red first, for the right reasons): test_optionssetup asserts the
canvas name equals NS.Constants.BRAND and that PARENT_TITLE's source reads
the constant; test_schema spies NS.Print across a ValidateSchema shape
error and asserts Schema.lua carries no "[AT]" literal and no
DEFAULT_CHAT_FRAME write. docs/test-cases.md regenerated (763), README
badge 763/763; ARCHITECTURE, settings-panel and module-map (brand source,
Schema.lua/test_schema.lua/test_optionssetup.lua line counts) rippled.

Deviations: the item cites Schema.lua:233-239 and test_optionssetup:297;
the code sits at Schema.lua:338-344 and the panel-name assert at :320.
The source-text asserts are added on top of the item's behavioural asserts
so the new tests are genuinely red before the change (the item's own
asserts pass on the old code because NS.Print is always defined).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 070d56aef74b34b3cc33029eb2620c0ce3b8652c
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 10:19:35 2026 +0530

AT-17: Prune the eight read_globals no authored file reads

Remove C_Timer, hooksecurefunc, CreateColor, PlaySound, strsplit,
strtrim, tinsert and tremove from .luacheckrc read_globals. No authored
file under core/, modules/, settings/, defaults/, locales/ or tests/
(excluding tests/_kit) reads any of them; the only C_Timer hit is the
comment at settings/OptionsSetup.lua:144, which stays. With them gone a
raw C_Timer or hooksecurefunc call lints red, which enforces
library-stack-§1's AceTimer rule. A short comment in the table says so.

Lint config only: luacheck stays 0/0, tests 762/0, lizard clean.

Finding: AbsorbTracker-R-09

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 44c6fdfea4080a6de4adab43ecb543124cf901c3
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 10:18:25 2026 +0530

AT-16: address review — module-map EnvSetup inventory row matches the two-rung ladder

The file-inventory row for core/EnvSetup.lua still described the deleted
pre-11.0 _G rung and called the file the only one naming a deprecated API.
It now says the file is the only one that reads TOC metadata, spells the
ladder as Env, then C_AddOns.GetAddOnMetadata, then nil, records why the
bare-global rung went, and carries the file's current 94-line count.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 34e21fa7f00427016c2916517dfea635b696f8d3
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 10:15:14 2026 +0530

AT-16: Delete the dead GetAddOnMetadata global rung from NS.Meta

core/EnvSetup.lua's library-absent ladder in NS.Meta fell back to the
pre-11.0 bare GetAddOnMetadata global after C_AddOns. No client the TOC's
## Interface line admits provides that global, so the rung could never
reach a working call. compat (WowAddonStandards v2.65.0) rules a dead
rung deleted rather than kept or shimmed, and its new applicability
condition means deleting it brings no core/Compat.lua back: this addon
calls no deprecated or version-variant client API outside LibKa0s's
majors.

- core/EnvSetup.lua: rung deleted; header and doc comment now describe
the ladder as Env, then C_AddOns, then nil.
- .luacheckrc: GetAddOnMetadata dropped from read_globals (unreferenced).
- tests/test_envsetup.lua: new degraded-env case, library and C_AddOns
absent with a global mock present, answers nil and never calls the
mock (red under the restored rung: got 9.9.9).
- docs/ARCHITECTURE.md compat-layer map row cites the applicability
condition; the metadata invariant, docs/module-map.md EnvSetup entry
and docs/scope.md now describe the two-rung ladder.
- docs/test-cases.md regenerated (762); README Tests badge 762/762 (it
was stale at 759 before this change).

Finding: AbsorbTracker-A-15 (audit AT-77).

Deviations: the item cites docs/ARCHITECTURE.md :626 for the
compat-layer row; it sits at :660. The header ladder line cited as :9
is in the "WHAT A DEGRADED INSTALL GETS" block (:33-38); :9 only named
the old Compat member, reworded so the verify grep stays empty.
docs/scope.md also described the global rung and was updated too.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 19761b88941b3efa67356748710d78560fa86a63
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 10:10:12 2026 +0530

AT-15: Decline RenderTabbedSchema opts; pin the Appearance strip

Evaluated LibKa0s v1.56.0's O.RenderTabbedSchema opts (OptionsTabs
minor 4: tabs, cfg, disabledFor, disabledNotice, chrome) for the
Appearance page's hand-built strip in settings/UnitPanel.lua
renderUnitPanelBody step 2, against the item's four adoption conditions:

(i) PageHeader before the call survives and ctx.__bannerHeight feeds
the strip's reservation: holds.
(ii) rows from rowsForPage(pageKey, ctx.unit): holds.
(iii) disabledFor(cfg) re-evaluated per render and tab click: holds.
(iv) skipRender rows stay out of the strip: FAILS. The library's
partition keys on row.group alone, so the mirror row
(skipRender, group "Link") would become a sixth tab on target
and focus.

Also found, outside the four: disabledFor draws the notice above rows
that are still drawn (disabled). This page draws the hint in their
place. And the library's tab click (ClearScroll plus a re-render of
itself) drops the page's two-tier mirror refresher. Declined with no
code change. Filed tusharsaxena/AbsorbTracker#32 (enhancement,
state:triaged, severity:low). It names condition (iv) and the re-check
trigger, which is the next Options minor.

Tests (tests/test_widgets.lua), the host pin of its own composition:
- For every unit in NS.Units.LIST and every tab: the TabStrip spec's
keys equal the schema's non-skipRender groups in declaration order.
Each tab draws exactly its group's rows. ctx.__bannerHeight > 0 and
ctx.chromeHeight >= ctx.__bannerHeight.
- target and focus, mirrored: every tab draws the hint and no
appearance rows.
Mutations confirmed red: partitionTabs keeping skipRender rows (sixth
tab), and renderUnitPanelBody ignoring IsMirrored (rows drawn).

A-04: the General page's chromeHeight > 0 assertion now carries a
comment. The strip is library-drawn (O.TabStrip via
O.RenderTabbedSchema), so the wrap-invariance pin belongs to LibKa0s's
suite and is not duplicated here (options-ui-§13 Testing, testing-§8).

Ripples: docs/test-cases.md regenerated (759 -> 761). The
docs/module-map.md test_widgets.lua row is now 1073 lines and names the
new pins.

Finding: AbsorbTracker-A-04.
Deviation: the item's mutation "disabledFor answers false" only applies
to an adoption. On this decline, the equivalent host mutation
(IsMirrored ignored) is the one confirmed red.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit a9de87e5bf8aee73e78659c67d4e49c2a2a87a7a
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 10:04:28 2026 +0530

AT-14: address review — module-map and comment ripple for the Perf and Lifecycle cases

dc477fc grew tests/test_surface_parity.lua to nine seams and gave
Kit.setSurfaceSource a fifth instance row, but left the prose describing
both behind.

- docs/module-map.md, tests/run.lua row: 139 -> 142 lines; the surface
source now maps five majors to instances (Lifecycle among them) plus
Bus and Schema to their library tables.
- docs/module-map.md, tests/test_surface_parity.lua row: 236 -> 286
lines; seven seams -> nine (adds Perf and Lifecycle); "the four
library-backed seams" -> every library-backed seam but Perf uses the
by-name form; notes the Perf case's reached-subset check.
- tests/run.lua comment: "all four ... stubs mirror an INSTANCE" -> five.
- tests/test_surface_parity.lua header: the same stale "four" and
"all three stubs" counts.

Comments and docs only. luacheck 0/0, 759/759 tests, lizard no
thresholds exceeded.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit dc477fc0cbc4b100af6401f8f6889f9eb701dbc2
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 10:01:57 2026 +0530

AT-14: Add stub-surface parity cases for Perf and Lifecycle

tests/test_surface_parity.lua now covers nine seams, not seven. Two new
cases gate the two degradation stubs that had no parity check:

- Perf: the member set the host reaches, derived with
grep -rohE '\bPerf[.:][A-Za-z_]+' core modules settings
({on, suspended, Note, OnCommand}), must be present on both NS.Perf
(live) and the degraded load's NS.Perf, and a function degraded
wherever it is a function live. The whole-surface by-name form is not
used: the live instance publishes far more than the four-member stub.
- Lifecycle: tests/run.lua registers NS.lifecycle as the live surface of
LibKa0s-Lifecycle-1.0 in Kit.setSurfaceSource, and the case uses the
by-name form. Measured at this commit the live instance publishes
nothing beyond the stub's nine members, so the ignore set is empty.

Proved by mutation: removing Note from core/PerfSetup.lua's stub turns
the Perf case red ("Note is missing degraded"), and removing Holds from
core/Lifecycle.lua's stub turns the Lifecycle case red ("Holds is
missing"). Both reverted. docs/test-cases.md regenerated; README badge
757 -> 759.

Resolves AbsorbTracker-A-10.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit c32be6917d48f8b8e7232c268e45efe3182cfd74
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 09:59:13 2026 +0530

AT-13: address review — module-map test_data.lua row to 544 lines

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 05c0be3ea5ca60fc068f9c3b450e9d54083b8b8e
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 09:57:34 2026 +0530

AT-13: Delete the test-only player aliases from the production namespace

modules/Bar.lua no longer publishes NS.bar / NS.statusBar / NS.valueText /
NS.backdropInfo. No production code read them; they survived only as a
handle for the test harness. With the four assignments and their comment
gone the suite went red at exactly the alias reads (39 cases), and every
read is now NS.bars.player / NS.bars.player.statusBar / .valueText /
.backdropInfo in tests/test_display.lua (46), tests/test_debughold.lua (3),
same meaning. The file header comment is updated to match.

tests/test_data.lua's "player alias points at the player frame" case
asserted the aliases themselves; rewritten as NS.bars.player it would
compare a value with itself, so it now asserts the four keys stay off NS
(indexed by string, so the verify grep stays empty). The case count is
unchanged at 757; docs/test-cases.md regenerated for the new case name.

Doc ripples: docs/module-map.md (the NS state list, the Bar API block, and
the Bar.lua row: 172 -> 159 lines, no aliases), docs/ARCHITECTURE.md's
Bar.lua row, and docs/midnight-quirks.md's SetBackdrop note, which still
named the single shared NS.backdropInfo table.

Findings: AbsorbTracker-R-13

Deviations: the test_data.lua alias case asserts the aliases' absence
instead of a literal rewrite (a rewrite would be tautological).
midnight-quirks.md also updated, not named in the item.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 6e4f2207f2e0bb3fd3250fcb74c1e7a20f70b361
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 09:52:29 2026 +0530

AT-12: Rename the minimap row's CLI path to global.minimap.shown

The minimap row's path now reads in the row's own sense (launcher-§3,
standard v2.65.0 / WS-06): C.MINIMAP_PATH = "global.minimap.shown".
Storage is unchanged: LibDBIcon's own db.global.minimap.hide, inverted by
the row's get/set in core/Data.lua. No `shown` key is ever stored
(anti-pattern #81), so there is no SavedVariables migration and no
schema-version bump; defaults/Profile.lua minimap.hide = false is as-is.

- core/Constants.lua: the constant's comment is moved directly above it,
after the BRAND block (it had been split from the constant by that
block), and rewritten: CLI name reads SHOWN, storage stays
minimap.hide, get/set invert, no `shown` key stored.
- settings/Schema.lua: defaultsRoot answers nil for a row that owns its
storage (type(row.get) == "function"), architecture-§5's closure-backed
exemption, so the validator does not report global.minimap.shown
missing from the defaults. Its comment and the resetExempt comment
name the new path. resetExempt / reachedByProfileReset key off the
constant and are unchanged.
- core/Data.lua and settings/OptionsSetup.lua comments name the new path.
- Docs: ARCHITECTURE.md (path lists, launcher visibility paragraph,
validator exemption, named-reset note), settings-panel.md,
slash-dispatch.md, schema.md (validator exemption), module-map.md
(path mentions, LOC cells for the touched files), smoke-tests.md U.6
(the CLI read and the old spelling). test-cases.md regenerated;
README Tests badge 752 -> 757.

Tests (red first, tests/test_launcher.lua): get global.minimap.shown
answers true over hide = false; set ... false stores hide = true and
MinimapShown() is false; get global.minimap.hide answers "Setting not
found"; ValidateSchema reports 0 errors / 0 missing; a legacy store
{ hide = true, minimapPos = 200 } reads shown = false, stays hidden,
keeps minimapPos, and never gains a `shown` key after a set.

Player-facing CLI rename, for the next Version History roll:
`/at get|set|reset global.minimap.hide` -> `global.minimap.shown`; the
old path now answers "Setting not found". A button hidden before the
upgrade stays hidden.

Findings: none listed on the item (plan item AT-12; R-07 split comment).

Deviations: the item cites defaultsRoot at settings/Schema.lua:404-409
and a :228-231 comment; the function and its doc comment actually sit
at :526-536, and that doc comment is the one updated. No literal
`resolved` count exists in tests/test_schema.lua, so nothing was
re-pinned there. ARCHITECTURE.md lives at docs/ARCHITECTURE.md. The
storage sentence in ARCHITECTURE.md was reworded ("the db.global.minimap
table's hide key") so the verify grep for global.minimap.hide is empty;
schema.md, module-map.md and smoke-tests.md were also updated as ripples.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit ae0c7ba949ab33406f28c976956711632e55ac3e
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 09:46:44 2026 +0530

AT-11: address review — reword the last `/at test` comment in test_debughold.lua

tests/test_debughold.lua:7 still spelled the retired verb as `/at test <value> [secs]`,
so the item's verify grep returned it; 0e089ae's body wrongly said only the README 1.8.0
Version History row remained. The comment now reads "the top-level `test <value> [secs]`
verb", as settings/Slash.lua already puts it. The verify grep now returns only README.md's
1.8.0 row, the declared deviation (release history is not rewritten).

Lint 0/0, tests 752/752, lizard -C 15 clean.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 0e089ae228759c3657131540f4d3467192b36b89
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 09:43:19 2026 +0530

AT-11: Move /at test to /at debug hold with a validated duration

This addon's unlocked view is its preview, and options-ui-§15 and
preview-mode say an addon in that shape ships no `test` verb
(anti-pattern #80). The one-shot value hold answers a question unlocking
does not, so it survives at `/at debug hold <value> [secs]`, where WS-06
(2) puts a kept value hold.

settings/Slash.lua
- The `test` row and runTest/runTestHold are gone. DEBUG_VERBS gains
`hold`, and runDebug now hands each sub-verb the rest of its line.
- runHold: parseHold refuses a missing or NaN value and any duration
that is not a number from 0.5 to 60 (default 5), printing HOLD_USAGE
("Usage: /at debug hold <value> [secs] — secs from 0.5 to 60, default
5") and holding nothing. It announces "Holding <v> on the bars for
<secs> s" with %g, so 2.5 reads as 2.5 rather than %d's "2".
- `debug` stays live while disabled, but a hold paints the bars, so
runHold checks the stored `enabled` itself and prints
Sl:DisabledLine(). on/off/events/bare are unchanged.
- The debug help row names `hold <value> [secs]`.

Tests: the /at test section of test_slashcmds.lua and its refused-while-
disabled case moved to the new tests/test_debughold.lua (run right after
test_perfcmds) and were retargeted at `debug hold`. Red-first additions:
negative, over-60, sub-0.5 and non-numeric durations print the usage and
never reach NS.HoldPreview (spied); 2.5 is announced as "2.5 s" and held
for 2.5; `test 1000` prints unknown command 'test'; with enabled=false,
`debug hold` prints the disabled line and holds nothing. test_disabled
step 7 gains the hold's own refusal (it has no literal 'test' entry; its
COMMANDS walk dropped `test` by itself). test_slashcmds.lua is now 1318
lines. docs/test-cases.md regenerated, and the badge is 752/752.

Comments and docs follow the move: General.lua, Lifecycle.lua,
Display.lua, Bar.lua, Timer.lua, test_display/test_timer case names,
README (the hold paragraph and troubleshooting row), slash-dispatch.md
(verb table, the refused list, hold's own gate, refreshed line cites),
smoke-tests.md (steps 33-34 and the disabled step, including a stale
"Test mode ticked" sentence), ARCHITECTURE.md's Slash Commands section,
module-map.md, settings-panel.md, common-tasks.md and scope.md.

Findings: AbsorbTracker-A-14, AbsorbTracker-A-24, AbsorbTracker-R-12.

Deviations:
- The README Version History row for 1.8.0 still says it added a
hold-time to `/at test`. That is a record of what 1.8.0 shipped, so it
was left alone, and the verify grep for '/at test' shows only that
row.
- ARCHITECTURE.md has no Slash Commands *table*. The section is prose,
and the prose was updated.
- lizard's Lua reader skips a function body that contains `[#t + 1]`, so
the old runTest never showed up in its output. parseHold matches two
words instead of building an args table, which lets lizard see runHold
(CCN 6).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 4e938b122204771eea67781be2b7d8ab77f9805d
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 09:33:07 2026 +0530

AT-10: address review — module-map LauncherSetup row: 155 LOC, library-owned disabled gate

The docs/module-map.md row for core/LauncherSetup.lua was left stale by
e35e134: the LOC column still read 169 (the file is now 155 lines), and the
row listed the descriptor fields without isEnabled / disabledLine and did not
say the rung-(b) disabled gate is now LibKa0s-Launcher-1.0 minor 2's, not the
host's. The row now names both fields and describes the library refusing the
left click before onClick runs, printing the Slash dispatcher's own line.

Docs only. luacheck 0/0, tests 746/746, lizard -C 15 clean.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit e35e134717fdb9fab947c81db07a67d17519d1b1
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 09:30:43 2026 +0530

AT-10: Move the launcher's disabled gate into Launcher minor 2's isEnabled

core/LauncherSetup.lua: the lib:New descriptor now passes
isEnabled (NS.GetSetting("enabled") ~= false) and disabledLine
(NS.Slash:DisabledLine()), both resolved at click time, so
LibKa0s-Launcher-1.0 minor 2 refuses a disabled addon's left click
before onClick runs (launcher-§2, slash-commands-§7). The hand-written
gate is gone from onClick and its comment block is cut to two lines
naming the descriptor fields. Right-click and openSettings unchanged.

Tests: tests/test_disabled.lua step 8 and its single-refusal-line
assertion stay green with the host gate deleted. Mutation-verified:
removing isEnabled from the descriptor turns step 8 red (and the new
case); isEnabled answering false turns the new case and the existing
left-click case red. New tests/test_launcher.lua case: one registered
object toggles `locked` while enabled, writes nothing while disabled,
and toggles again once re-enabled (never cached).

Docs: docs/ARCHITECTURE.md '## Launcher' and docs/slash-dispatch.md say
the gate is the library's. docs/test-cases.md regenerated and the README
badge moved 744 -> 746.

Deviation: the regenerated inventory also picks up the test AT-09's
review commit (0850145) added without regenerating ("degraded: the
library-absent line is keyed by its English text"), so the count moves
by two, not one.

Findings: none (item carries no finding ids).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 08501459d2d3b837825c35c4e736e2034cf85c08
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 09:26:48 2026 +0530

AT-09: address review — key the library-absent line by its English text; fix stale Slash.lua refs

- localization-§2: locales/enUS.lua keys the library-absent line by the
English sentence rather than the symbolic LIBRARY_ABSENT token, and
settings/Slash.lua's stub reads NS.L["%s is unavailable: the LibKa0s
library did not load."], so the metatable fallback yields English.
A new test pins the key shape (red before the fix).
- ARCHITECTURE.md localization-§1 row, module-map.md rows 1 and 19 and
slash-dispatch.md name the English-text key; module-map row 1's line
count 53 -> 54.
- Line refs to settings/Slash.lua moved by bdd4ecf's header: MirrorNote
:48 -> :58, NS.COMMANDS :61 -> :71. Older neighbouring drift fixed in
the same pass: Sl:Register :645 -> :713, runProfile :456 -> :524,
PROFILE_VERBS :408 -> :465 (slash-dispatch.md), the PROFILE_VERBS
dispatch :469 -> :537 (ARCHITECTURE.md) and AbbreviateNumbers
:349/:353 -> :392/:396 (smoke-tests.md).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit bdd4ecf5d8685ed76ad40ccfd31b4ecfc91053c7
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 09:21:52 2026 +0530

AT-09: Give the Slash stub the LK-18 shape: pinned disabled line, plain rows

settings/Slash.lua's library-absent stub now takes the shape
slash-commands-§1 (WS-02) and LK-18 prescribe:

- STUB_DISABLED_LINE_FORMAT carries LibKa0s-Slash-1.0's DISABLED_LINE_FORMAT
verbatim (v1.56.0 bytes, em dash as \226\128\148), published as
Sl.__STUB_DISABLED_LINE_FORMAT and pinned in tests/test_slashcmds.lua with
T.assertLibraryConstant. The stub's DisabledLine formats it with the brand
and "/at enable", so it equals the live line, gold escape included. The
comment that forbade the copy is rewritten.
- The stub's FormatRow copy is gone. PrintCmd falls back to a plain
"cmd desc" row when the library has no FormatRow; LandingRows builds
the same plain rows.
- The schema verbs (list/get/set/reset/resetall) print the library-absent
line through the locale: L["LIBRARY_ABSENT"] = "%s is unavailable: the
LibKa0s library did not load." added to locales/enUS.lua.

Red first: four new cases in the degraded/live parity section (the pin,
the exact library-absent line, no |c in degraded help, degraded vs live
DisabledLine) failed for the expected reasons, then went green. The kit's
assertLibraryConstant falls back to LibStub, so the pin resolves the
library table even though tests/run.lua registers the Slash instance.

Ripples: ARCHITECTURE.md localization-§1 register row records the routed
line; slash-dispatch.md's degraded section describes the new stub;
module-map.md rows (Slash.lua, enUS.lua, test_slashcmds.lua) and line
counts; smoke-tests.md item 99 walks /at help and /at list degraded;
docs/test-cases.md regenerated; README badge 744/744.

No finding ids (item has none). Deviations: none beyond the smoke-test
step added under item 99 and module-map line counts refreshed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 3ebbfc165d75d2a8a4b88ea11bbbd40e311c9805
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 09:14:54 2026 +0530

AT-08: Adopt Schema writeThrough; hollow the Options stub's composers

settings/Schema.lua declares WRITE_THROUGH = { "enabled", "locked" } and
hands it to SchemaLib:New as writeThrough (LibKa0s-Schema-1.0 minor 2,
options-ui-§1 route (a)), so both the live instance (a partial load with
Schema present and Options absent) and the host stub store those composed
paths without a row. The announce dispatches a written-through path to
the host's own reaction, NS.MasterReactions[path], which
settings/General.lua now publishes (its masterOnChange table): the enable
edge still moves the stand-down latch and an in-combat unlock is still
refused. The HostSchemaStub comment says the host writers that reach
composed rows now write through the declared list.

settings/OptionsSetup.lua's degradation stub drops ORDER_STEP,
composeBlock, the ColorPair/FontGroup/BorderGroup/BarGroup/MasterControls
host copies and Helpers.MASTER_GROUP (its only reader is General's
build(), which a library-less load never reaches). The five composers
answer {} (MasterControls also a no-op tail), per options-ui-§1 (v2.65.0)
and anti-pattern #73. The load-time set is still six members. The file
goes from 509 to 378 lines. tests/test_surface_parity.lua's Options
ignore list gains MASTER_GROUP with that reason.

Tests. Characterization first, green before and after, in
tests/test_optionssetup.lua with degradedWithDb: degraded /at disable
stores enabled=false and the latch is down, /at enable reverses it; an
in-combat /at unlock leaves locked=true and prints the refusal. The
existing lock/unlock and combat re-lock cases stay green. Mutations
checked: dropping 'enabled' from WRITE_THROUGH, or disabling the
announce's reaction dispatch, turns the new cases red. Re-pins:
tests/test_perf.lua's degraded-schema case becomes three figures (full
70, degraded 15, and a named gap of 55 attributed per tab: Master
controls 7, Bar 12, Background 6, Border 12, Text 18), plus degraded is a
subset of live. A stub composer emitting rows turns it red. The composer
case in tests/test_optionssetup.lua asserts each composer answers
#rows == 0 and MasterControls a function tail; its leaf-shape asserts
are gone.

Docs: schema.md gains a "With LibKa0s absent" section; settings-panel.md
degradation stub section; ARCHITECTURE.md Settings Schema, invariant and
OptionsSetup module row; module-map rows for Schema, OptionsSetup, General
and the three touched suites (line counts refreshed); common-tasks.md
suite row; smoke-tests.md step 99 walks the item's smoke; test-cases.md
regenerated, README badge 740/740.

Findings: AbsorbTracker-A-02, AbsorbTracker-A-03.

Deviations: (1) The item's test (c) names tests/test_perf.lua for the
combat re-lock case; it lives in tests/test_optionssetup.lua and stays
green there. (2) "the live and degraded builds veto exactly the same rows
from Reset All" compared the whole live veto set against the degraded
one, which the hollow composers make unequal by design. It now compares
over the rows both builds carry (the gap itself is pinned by the perf
case). (3) The item's docs list did not name smoke-tests.md or
common-tasks.md. Step 99 claimed a "complete" degraded /at list that the
Slash stub never gave, so it was rewritten to the item's smoke walk.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 94bed8cbdfab3efea09f2e7aebcf30b3af6022c3
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 09:06:55 2026 +0530

AT-07: address review — module-map and data-flow ripples for the SafeRegister helpers

5b6e2cf updated ARCHITECTURE.md's module-map rows but not docs/module-map.md
or docs/data-flow.md, so both still described the bare registration path
the item removed.

- docs/module-map.md: line counts re-synced (State 13, CoreSetup 147,
Lifecycle 211, DebugLogSetup 127, AbsorbTracker 406, Slash 706, run.lua
136, test_coresetup 166; Slash's 648 was already stale before AT-07).
State row names NS.State.rejectedEvents; CoreSetup row names the three
NS.SafeRegister* publications and the degraded one-rung bodies;
DebugLogSetup row names the [Init] rejected-events suffix; AbsorbTracker
row names the pcalled registrations and RegisterLifecycleEvents; Slash
row names the `debug events` sub-verb; run.lua row's suite count
corrected to twenty-nine plus the kit's three; test_coresetup row names
the degraded SafeRegister case; new row 54a for tests/test_events.lua.
- docs/data-flow.md: the OnEnable diagram shows NS.SafeRegisterEvent /
NS.SafeRegisterUnitEvent at all seven call sites, RegisterLifecycleEvents,
and the rejected-list record; the prose after it explains the pcall.

Docs only. luacheck 0/0, tests 738/738, inventory diff empty.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit 5b6e2cf5216bbb426ecf53749323f85219e9a4a1
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 09:03:20 2026 +0530

AT-07: Route every event registration through the SafeRegister helpers

The client raises on an unknown event name, and a block of bare
RegisterEvent calls lost every registration after the one that raised.
All seven call sites in core/AbsorbTracker.lua (the three lifecycle
events, now driven from a { event, method } array, the two per-unit
RegisterUnitEvent calls and the two swap events) now go through
NS.SafeRegisterEvent / NS.SafeRegisterUnitEvent, which core/CoreSetup.lua
binds to LibKa0s-Core minor 8's helpers. A refused name costs only
itself, is appended once to the new session list NS.State.rejectedEvents
and is logged under the Events debug tag.

- core/CoreSetup.lua: live arm publishes SafeRegisterEvent,
SafeRegisterUnitEvent and SafeRegisterEvents; the degraded arm carries
the one-rung bodies (pcall plus append-once, no front gate) per
docs/api/Core/version-8-docs.md 'Degradation'.
- settings/Slash.lua: /at debug gains an `events` sub-verb through a
DEBUG_VERBS table; the help row names it.
- core/DebugLogSetup.lua: the [Init] summary appends
', rejected events: <n>' when the list is non-empty.
- core/Lifecycle.lua: the ADDON_EVENTS comment names the helper.
- Tests: new tests/test_events.lua (before test_disabled) covers the
AceEvent half, the per-unit frame half (kit 26), the IsEventValid front
gate and /at debug events, and restores the full registration set.
test_coresetup.lua covers the degraded SafeRegisterEvent. Both were
confirmed red first (bare calls raise; a stub without pcall raises).
- Docs: midnight-quirks note, ARCHITECTURE Event Subscriptions bullet and
module-map rows, slash-dispatch rows; docs/test-cases.md regenerated,
Tests badge 738/738.

docs/ARCHITECTURE.md already had a '## Event Subscriptions' section, so
the item's content went in as a bullet there rather than a new heading.

Finding: AbsorbTracker-A-12

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HTuhyvd3dgxDekBFUvKcoZ

commit be520a50b26bf0e73927c2197e2b253549801359
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 08:04:35 2026 +0530

AT-06: Stamp schemaVersion per WS-03: floor 0, runner-owned, v2 per profile

Brings the migration runner in line with savedvariables-§1 as ruled at
WowAddonStandards v2.65.0 (WS-03).

- defaults/Profile.lua: NS.defaults.global.schemaVersion is now 0, the
pre-migration floor. The comment gives §1's reasons: AceDB's
removeDefaults strips a stored value equal to its default, and AceDB
backfills a declared default onto a legacy account that stored no
stamp; 0 has neither problem. The per-profile stamp (default 1, the
registered savedvariables-§1 deviation) is unchanged.
- core/Database.lua: NS.SCHEMA_VERSION = SCHEMA_STEPS[#SCHEMA_STEPS].to
(derived, no second literal) and the suite seam NS.__migrationSteps.
RunMigrations reads an absent stamp as 0 and walks the ladder in a new
file-local runLadder: each step runs under pcall; on failure it logs
"[Migrate] vX -> vY failed: <err>", prints one chat line ("Settings
upgrade stopped at vX -> vY; saved settings were left as they were")
and breaks, so the stamp stays at the last completed step and later
steps do not run. On success it logs and stamps step.to as before.
RunMigrations is CCN 6, runLadder CCN 4.
- The v2 step was profile-scoped but cleared only the active profile
under the account-wide stamp. It now sweeps every stored profile
through forEachProfile and logs a count line, like v4 and v5.

Tests (tests/test_database.lua, red first): the global default is 0
(red: got 1); NS.SCHEMA_VERSION equals the last step's `to` (red: seam
missing); v2 clears updateInterval from an inactive Raid profile (red:
Raid kept 0.2); an appended raising step leaves the stamp at
NS.SCHEMA_VERSION, RunMigrations does not raise, and one chat line names
the stop (red under: stamping before apply, or no pcall). The existing
nil-stamp case still reaches 5. The v1->v5 [Migrate] order test now
expects v2's own count line first (seven lines), and a stale comment
citing "reads the absent stamp as 1" is corrected.

Docs: ARCHITECTURE.md (module map row, savedvariables-§1 register row,
"The per-profile schema stamp, in full"), data-flow.md, profiles.md,
module-map.md (API block, defaults block, file rows and line counts),
smoke-tests.md step 51 (stale v4 -> v5, [Init] "schema v5" check, v2
sweep). docs/test-cases.md regenerated (732); README badge 728 -> 732.

Deviations: the item names docs/schema.md's migration section, but
schema.md has none (it covers the settings row schema); the migration
prose lives in docs/profiles.md and docs/data-flow.md, which were
updated instead. CLAUDE.md does not quote schemaVersion, so it is
unchanged. The failure Debug line keeps the ladder's existing arrow
glyph rather than an ASCII "->"; the chat line uses "->" as specified.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ldej1stXpdhQLvEJxKqube

commit 7e67f75082893d04a931c28a6377290b7d450b3c
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 07:57:29 2026 +0530

AT-05: Publish the bar passes once on a profile adopt's enable edge

adoptProfile (core/AbsorbTracker.lua) re-read the new profile's `enabled`
through NS.SyncEnabledHold, then called NS.lifecycle:Reevaluate() and
published UNITS, POSITION, APPEARANCE and REPAINT unconditionally. The
Reevaluate was a no-op (lifecycle:Set already runs the edge), and on an
off-to-on switch StandUp (core/Lifecycle.lua) had already published
POSITION, APPEARANCE and REPAINT, so every three-bar appearance pass ran
twice.

Now adoptProfile captures `wasDown` before SyncEnabledHold, drops the
Reevaluate, always sends UNITS, and sends POSITION/APPEARANCE/REPAINT only
when the latch did not just stand up. The comment says why. The
MigrateProfileToV3 call and RefreshOptionsPanel are unchanged.

Tests (tests/test_database.lua, characterization first): a same-state
switch sends APPEARANCE once; an off-to-on switch sent 2 before the fix
(observed red) and sends 1 now (red under: restoring the unconditional
publishes); an on-to-off switch leaves the latch down and no APPEARANCE
pass reaches a subscriber. Cleanup restores Default and enabled=true.

Ripples: docs/test-cases.md regenerated (728), README badge 725 -> 728,
docs/ARCHITECTURE.md and docs/module-map.md no longer describe the
Reevaluate call (module-map row 13 line count 374 -> 380, and its
OnProfileChanged summary now names UNITS and the stand-up skip).

Resolves: AbsorbTracker-R-11 (review F-010).

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ldej1stXpdhQLvEJxKqube

commit 53c1c7ea892336c1c7f459a469e998a690538126
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 07:52:57 2026 +0530

AT-04: Echo the stored value for lock, unlock and enable; fix the comment

settings/Slash.lua gains a file-local echoStored(path): it refreshes an
open options panel, reads the path back from the store, formats it
through its schema row and prints slash-commands-§5's `path = value`
shape via SlashLib.FormatKV. setEnabled, /at lock and /at unlock all use
it, so the fixed 'Bar locked'/'Bar unlocked' lines are gone. An in-combat
/at unlock, which the locked row's onChange refuses and writes back to
true, now ends on `locked = true` instead of claiming success, and the
Lock frame checkbox moves with either verb. Help texts say "bars".
(AbsorbTracker-R-03)

The /at enable|disable comment no longer describes the removed draw gate
(VISIBILITY then REPAINT, "gates ShouldShowBar's second rung and nothing
else ... no event registration changes"). It now says what happens: the
row's onChange is NS.SyncEnabledHold, which moves the latch's `disabled`
hold; StandDown unregisters the unit-event frames, the five AceEvent
registrations and the bus; the dispatcher, COMMANDS and the settings
registration stay live (slash-commands-§2/§7). (AbsorbTracker-R-05)

Tests (red first): the lock/unlock case asserts `locked = true/false`
and no 'Bar locked'; a new in-combat case asserts the refusal, a last
line ending `locked = true` and the store still true; a new case spies
NS.RefreshOptionsPanel and asserts one call per verb. Three stale test
comments that repeated the draw-gate claim were corrected too.
docs/test-cases.md regenerated, README badge 723 -> 725.

Docs: README quoted neither old line, so it only carries the badge. The
docs/slash-dispatch.md enable/disable row did not repeat the draw-gate
claim and is unchanged; its lock/unlock row now describes the echo.
docs/smoke-tests.md steps 14e and 30 name the echoed lines, including
`locked = true` after an in-combat /at unlock.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ldej1stXpdhQLvEJxKqube

commit 32b60e4137d73df5ade7a6d37a630183b0be30b3
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 07:49:34 2026 +0530

AT-03: address review — move smoke step 36a before 37 so the tester is on SmokeTest

Step 36a (/at profile new Default refused) says "you stay on SmokeTest",
but it sat after step 37, which ends on /at profile use Default. Moved it
between 36 and 37, where the current profile really is SmokeTest.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ldej1stXpdhQLvEJxKqube

commit 0e5478b3cdb6be8cb665d7fca671e1f75180957e
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 07:47:04 2026 +0530

AT-03: Refuse /at profile new on an existing name; pre-check copy/delete

`/at profile new <existing>` switched to that profile and reset it, then
printed "Created", so one typed command wiped a profile (R-01). `copy`
handed AceDB a missing or current name, which raises a raw Lua error in
the client, and `delete` of a missing name printed a deletion that never
happened (R-02).

settings/Slash.lua gains a file-local profileExists(db, name) walking
db:GetProfiles() for an exact, case-sensitive match. `new` refuses an
existing name, pointing at `/at profile use` and `/at profile reset`.
`copy` refuses the current profile ("Cannot copy a profile onto itself")
and an unknown one ("Profile '<name>' not found — /at profile list shows
them") before CopyProfile. `delete` keeps its current-profile refusal and
adds the same not-found line. No pcall. The misleading comment above
`new` is rewritten; PROFILE_HELP's "Create new profile with defaults" is
now true. This also meets the LK-03 consumer obligation: kit 26's AceDB
fake raises on these names, and no raw error reaches a suite or player.

Tests, red first in tests/test_slashcmds.lua's /at profile section:
new on an existing Keep (barWidth 333 survives, still on Default), copy
of a missing name, copy of the current name, delete of a missing name.
The existing "new creates a profile carrying the defaults" case seeded
an existing "Fresh" and relied on `new` resetting it; it now makes a
genuinely new profile from a changed current one and deletes it after.
The file is 1352 lines, under the 1400 split threshold.

docs/profiles.md and docs/smoke-tests.md (36a, 38a, 39) document the
three refusals. docs/test-cases.md regenerated (719 -> 723) and the
README Tests badge moved to match. The regeneration also restores the
section signs in three kit test names (localization-§5, line-endings-§5)
that the committed inventory had lost, so the file matches --list again.

Findings: AbsorbTracker-R-01, AbsorbTracker-R-02.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ldej1stXpdhQLvEJxKqube

commit e3d14b13905125351491f897c447487c2f664263
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 07:42:39 2026 +0530

AT-02: Make slash/profile tests assert real values and clean up per-unit

The slash and profile tests still addressed the flat barWidth key and the
removed bar/border pages, left over from before schema v3 made barWidth
per-unit. That gave nil==nil asserts and cleanups that did nothing, and
they leaked state into later suites (AbsorbTracker-R-04, testing-§12).

- tests/test_slashcmds.lua: every RestoreDefaults("bar") / ("border")
cleanup (:224-225, :371, :552, :574) is now RestoreDefaults("appearance").
VALID_PAGES is {general, appearance, profiles}, so the old calls reset
nothing and units.target.barWidth=300 from :218 leaked.
- /at profile new now seeds units.player.barWidth=456 and asserts
NS.unitDefaults.barWidth ("-- red under: new that skips
ResetProfileCounted"). /at profile reset does the same with 478 ("-- red
under: reset that switches instead of resetting").

Mutation evidence: with NS.ResetProfileCounted(db) deleted from Slash.lua's
`new`, the old :545-548 case stayed green. After the rewrite it goes red
(expected 200, got 456). A `reset` that switches profiles, or skips the
reset, turns the reset case red. Both mutations were reverted. The case
count is still 719, and no ordering dependency surfaced.

Deviations:
- The profile-new seed is written to an existing "Fresh" profile (switch,
seed, back to Default), not to Default as the item text says. A
brand-new AceDB profile starts from defaults whether or not `new` resets
it, so seeding Default stayed green under the mutation. `new` on an
existing name is exactly the path the reset exists for.
- tests/test_schema.lua (outside the item's "test_slashcmds.lua only"): the
"ApplyDefault is a no-op for a row with no default" case used
rawSet("barWidth"), which the item's own verify grep must not find. It
now targets units.player.barWidth and restores to NS.unitDefaults.barWidth.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ldej1stXpdhQLvEJxKqube

commit aa4c783f344efb1a82b703e1e1e74614306bcc0b
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 04:45:03 2026 +0530

RV-AT: Re-vendor LibKa0s v1.56.0 and record its delta bundle

Copy LibKa0s/ -> libs/LibKa0s/ and testkit/ -> tests/_kit/ whole from the
local tag v1.56.0 (tag 4622018 -> commit 514fc0a) via git archive. Both
diff -r checks against the tag are empty, and the runner stays executable.
CLAUDE.md's provenance line moves v1.55.0 -> v1.56.0 in the same commit.

Fifteen library files move a minor: Core 8, Lifecycle 2, Bus 2, Schema 2,
Item 2, Media 4, Widgets 10, DebugLog 13, Slash 15, Launcher 2, Options 24,
OptionsWidgets 31, OptionsTabs 4, OptionsScroll 4 and Perf 13. No file is
added or removed. The kit moves from revision 25 to 26 and adds asserts.lua,
mock_events.lua and prose_lists.lua.

The re-vendor bundle docs/revendor/2026-09-23-v1.56.0/ (01_DELTA.md and
05_SUMMARY.md) was written by hand, following the local wow-addon
revendor-libka0s procedure as amended by WA-01. Adoption decisions are
left to this addon's M3 items.

Green on the copy, so there are no reds to list: luacheck 0/0 in 61
files, tests 719 passed, 0 failed, lizard 0 warnings, and the largest
authored file has 1416 lines. AT-01 (460c51d) had already caught the
Schema stub up to minor 2.
docs/test-cases.md and the README test badge are not regenerated here; a
later item owns them. No finding ids. No deviations.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ldej1stXpdhQLvEJxKqube

commit 460c51d24d8392666c6b5652720012c50dc158d3
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Sep 24 04:39:55 2026 +0530

AT-01: Catch the Schema degradation stub up to Schema minor 2

settings/Schema.lua HostSchemaStub() now matches LibKa0s-Schema-1.0
minor 2 (LibKa0s v1.56.0, docs/api/Schema/version-2-docs.md "The
degradation stub"), so the re-vendor lands green:

- S.Set is split into writeRow -> prepareWrite (resolve, validate,
row.normalize, missing root) -> commit (store + onChange) -> announce.
prepareWrite keeps Set's refusal arity (false, err, why for a bad
value; false, err for a missing root). Every Schema.lua function is at
CCN 8 or below (S.Set was 14).
- writeThrough: New builds one synthetic { path, writeThrough = true }
row per listed path; a row-less listed path is stored raw (a copy), with
no validate/normalize/onChange, and announced with the synthetic row.
Any other row-less path is still refused. The descriptor is untouched
here (AT-08 adopts the list).
- SetMany(entries, opts): all-or-nothing and log-silent; the first
refusal answers false, err, why, index with nothing stored; then every
store, every onChange, then announceBatch(writes, rid) once or announce
per write.
- S.Get forwards the id to row.get(id); S.ApplyDefault(row, id) forwards
it to Set.
- The stub's doc comment cites version-2-docs.md and names SetMany and
writeThrough.

Tests: new section "the degradation stub, Schema minor 2" in
tests/test_schema.lua (9 cases, red first). docs/test-cases.md and the
README Tests badge regenerated (719). Green on the v1.55.0 payload and on
a v1.56.0 payload dry run (schema instance parity case passes).

Finding ids: none (plan item AT-01).

Deviation: the item has SetMany "commit in order" in phase 2; it runs
every store first and then every onChange, as the library and the v2
stub contract specify, so commit() is store()+react() and the batch
calls the two halves separately. Helpers checkValue, writeRow,
prepareEntry and announceAll were added to hold the CCN ceiling.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ldej1stXpdhQLvEJxKqube

commit f4b61a6d7f0bffa0e2c70220b84094df2f39f191
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 23 20:50:31 2026 +0530

Record the 2026-09-23 review and standards audit

The frozen docs/reviews/2026-09-23/ and docs/audits/2026-09-23/ bundles
from the collection-wide sweep. Findings are collated and planned in
Ka0sAddonsCommonTasks/docs/2026-09-23-REVIEW_AND_STANDARDS_AUDIT_REMEDIATION/.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ldej1stXpdhQLvEJxKqube

commit 98b4f729c98c58aa78e8b20dab8a7102f4f7ff9c
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 23 19:11:23 2026 +0530

Bring the module map in line with the v1.55.0 payload

The vendored LibKa0s now ships fifteen majors across twenty-one files,
not twelve across eighteen, and its XML loads Compat, Bus and Schema
after Env. The kit is revision 25, whose prose and layout-cap suites
took the US-English check out of tests/test_docs.lua; that row now
names the four rules the file still holds, and tests/prose_waivers.lua
gets the row it was missing. Four stale line counts are corrected.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TTuxA9sSHE1H7JLgJ4kYAx

commit 98b4f729c98c58aa78e8b20dab8a7102f4f7ff9c
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 23 19:11:23 2026 +0530

commit 7c43ab427e780c7a5612072d8af9ffaa58a7b5f2
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 23 18:55:49 2026 +0530

Give each unlocked bar the LibKa0s drag handle strip

While the bars are unlocked each one now gets the strip ConsumableMaster
shows over its macro bar: LibKa0s-Widgets-1.0's DragHandle, a dark strip
with a gold edge, the unit's name centered in gold and a help mark on
the right, anchored above the bar. It replaces bar.unitLabel, which is
gone with its font plumbing.

modules/Bar.lua builds one handle per bar (<frame>Handle, moveFrame the
bar). canDrag asks the lock, and onDragStop saves through the same local
savePosition the bar body's OnDragStop now uses. The strip's tooltip has
one body line read off the lock on each hover; the help mark has its own
tooltip with a /at lock footer. Every string goes through NS.L, and
enUS.lua lists the keys. The help art comes from NS.Icon("help"), the
seam's first call site. The bar body keeps its own drag.

UpdateBarAppearance shows the handle while unlocked, sized with
ApplyWidth(barWidth), and hides it while locked. The combat re-lock
hides it too. The default stack now leaves the strip's HEIGHT + GAP
between bars, so the player bar's strip does not sit over the target
bar. With the library absent there is no strip, the stack spacing is
unchanged, and nothing raises. That is the widget's documented
degradation.

tests/test_draghandle.lua covers the handle per bar and its spec, shown
only while unlocked, width, canDrag, the shared save and both tooltips,
plus the degraded load. The unit-label cases leave test_display.lua and
its stack test moves to the new spacing. The docs move with it:
ARCHITECTURE (Widgets now bound by name, the localization row), the
module map, a rewritten smoke item 68, the test inventory and the
badge (710).

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TTuxA9sSHE1H7JLgJ4kYAx

commit 6416d11da7043e7a738c51682aa47d98a839b4e9
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 23 18:47:44 2026 +0530

Pin the bar body's drag before the strip lands

A new suite, tests/test_draghandle.lua, pins the one way to move a bar
that has to survive the drag-handle change: every bar body is registered
for OnDragStart/OnDragStop, and a drop saves the position through
NS.Units.SetPosition to that bar's own unit and to no other. The
unlocked unit label's show/hide/text cases already sit in
tests/test_display.lua. Green on the current code; the badge and the
test-case inventory move to 695.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TTuxA9sSHE1H7JLgJ4kYAx

commit 1a410cb1f1ce79defa94d94e6a2e1484c7f6f621
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 23 16:54:31 2026 +0530

Make the stand-down prose match the guarded UNITS receiver

The degraded-UNITS test's falsification note said four registrations
come back without the guard; measured it is five (PLAYER_TARGET_CHANGED,
and UNIT_ABSORB_AMOUNT_CHANGED and UNIT_MAXHEALTH twice each).

ARCHITECTURE.md's Message Bus paragraph and "What stands down" bullet
said no module learns the latch exists, which stopped being true when
the UNITS receiver began asking NS.IsStoodDown(). Both now say no
receiver has to ask for its subscription to come down, and name the one
that asks anyway and why.

The v1.55.0 revendor summary names the guard by its commit (9e6a4a6)
and says what it does, in place of "a Phase 6 follow-up".

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TTuxA9sSHE1H7JLgJ4kYAx

commit 7b2aef334ac884e89421e36b23c6934ca8ad3ead
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 23 16:50:40 2026 +0530

Name the v1.55.0 re-vendor bundle for its tag

audit-review-history requires a re-vendor bundle folder shaped
<YYYY-MM-DD>-v<tag>. This run's bundle was written to a bare-dated
folder because the sweep's brief named the wrong path. The folder is
renamed to 2026-09-23-v1.55.0 and every reference follows it. The
bundle's contents are unchanged.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TTuxA9sSHE1H7JLgJ4kYAx

commit 9e6a4a627484f076c9a3c20398bd314bef3b9bde
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 23 15:44:15 2026 +0530

Keep the UNITS receiver from re-registering on a stood-down addon

With LibKa0s missing, core/Bus.lua's untracked-target stub cannot take
the UNITS subscription down, and its handler called SyncUnitEventFrames
without asking the latch. A publish while disabled put four per-unit
RegisterUnitEvent registrations and PLAYER_TARGET_CHANGED back on the
addon, against slash-commands-7. The receiver now returns while
NS.IsStoodDown(); StandUp re-syncs from current state, so nothing is
lost. The live build never reaches the line.

A degraded-load case in tests/test_bus.lua pins it (red before the
guard with five registrations back). Known Limitations and the
2026-09-23 revendor summary stop saying every handler answers to the
latch; the badge, test-case inventory and module-map rows move with it.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TTuxA9sSHE1H7JLgJ4kYAx

commit 368e82f0d70c205418081430d334ff32a54085d8
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 23 15:28:41 2026 +0530

Record the v1.55.0 adoption run in the revendor bundle

Steps 5-8 for LibKa0s v1.55.0: the candidates (Bus, Schema, Compat), the
decisions as they landed under the owner's delegation, the execution plan
with what actually ran, and the summary with every gate's numbers. Bus
and Schema were adopted (92c48d8, 0e72839); Compat was declined as never
and filed as #31.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TTuxA9sSHE1H7JLgJ4kYAx

commit 0e72839ff8e2d04caeb4ad486a030e5444dff237
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 23 15:26:44 2026 +0530

Adopt LibKa0s-Schema-1.0 as the settings schema runtime

settings/Schema.lua becomes the major's setup file: the library's instance
over the live NS.Schema array, or the write-completing, log-silent host
stub when LibKa0s is absent. The host's names are bound to its members
(SetByPath, FindSchemaRow, RegisterSchemaRows, ApplyDefault, Bulk,
ResolvePath/SetPath, the reset count trio, ValidateSchema), so no
production call site moved. The console and minimap rows carry their own
get/set, stamped by settings/General.lua; NS.GetSetting is the runtime's
read with the shipped defaults behind it, and NS.SetSetting is gone. The
minimap sweep exemption is the runtime's resetExempt.

Characterization first: the write's order (store, [Set] line, onChange)
and the three degraded writers (the lock verbs, the combat re-lock, Reset
All) landing in the store. Pinned with the change: an unknown path is
refused and a table value is stored as a copy (the API document's
adoption notes); the degraded Reset All's debug lines re-pin to "the
write landed, the line is absent"; the stub's lib and instance parity;
the minimap row surviving a sweep. Raw fixture writes in the suites go
through the harness's rawSet. tests/perf.lua stays under its ceilings.
Test inventory and README badge move to 691.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TTuxA9sSHE1H7JLgJ4kYAx

commit 92c48d8eed7728123f143392a85d8f5b85bc3933
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 23 15:11:29 2026 +0530

Adopt LibKa0s-Bus-1.0 for the bus record and the message catalog

core/Bus.lua's hand-written subscription register gives way to the
library's stand-down record: NS.NewBusTarget hands out tracked targets,
the latch calls NS.BusStandDown last and NS.BusStandUp first, and the five
subscription sites register on their targets directly. NS.MSG is declared
through Bus.Catalog, so an undeclared key now raises for a publisher too.
With LibKa0s absent the untracked-target stub keeps each receiver on a
private target and records nothing; Known Limitations says so.

Characterization first (the five bus pairs by name through a full
disable/enable, and each consumer reached once after it), then the stub's
parity by name, the degraded load, the strict read, and the two record
properties the old register lacked. The surface-source map gains the Bus
row. Test inventory and README badge move to 683.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TTuxA9sSHE1H7JLgJ4kYAx

commit d5e37d6b35478684274e2419a90dbf7679053cdc
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 23 14:57:30 2026 +0530

Move the README test badge to the 674 the inventory now reports

27ccce6 regenerated docs/test-cases.md to a 674 total but left the badge
at 646. documentation-§1 and testing-§5 require the badge to move in the
same change as the inventory, so this is a same-change miss landed as the
next commit rather than folded back.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TTuxA9sSHE1H7JLgJ4kYAx

commit 27ccce69206740dc12fd87e47e0bfb92c1dfc164
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 23 14:53:29 2026 +0530

Re-vendor LibKa0s v1.55.0

Both payloads copied whole from the tag: three new majors (Compat, Bus,
Schema) and their LibKa0s.xml rows, every existing file's minor unchanged,
and test-kit revision 25. The provenance line moves in the same commit.
The runner declares the kit's new layout-cap gate by its (basename,
directory) pair. No contract blockers; nothing adopted yet. The delta is
recorded in docs/revendor/2026-09-23/01_DELTA.md and the test-case
inventory is regenerated.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TTuxA9sSHE1H7JLgJ4kYAx

commit b6360c9a2f4e4a2c5e99f562e0debbc339a0da27
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 23 14:50:23 2026 +0530

Add the layout-§1 census heading under the deviations register

Nothing is over the 1500-line cap today; the census says so as a result,
ahead of the kit revision 25 gate that reads it.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TTuxA9sSHE1H7JLgJ4kYAx

commit 35f92ea987a62d2537ba7c367fe40892e34832b5
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 23 11:34:17 2026 +0530

Re-vendor the standard reference to v2.63.0: .gitattributes to the canonical body

The standards sweep against Ka0s WoW Addon Standard v2.63.0 found one item
to apply in this repo. .gitattributes now matches line-endings-§5's
client-bound canonical body byte-for-byte: the shebang carve-out comment is
the current wording and `*.py text eol=lf` joins `*.sh text eol=lf`. The pin
(`* text=auto eol=crlf`) and the binary block are unchanged, the repo tracks
no .py file, and the worktree straggler count is 0 before and after, so no
renormalization rides along.

Everything else already conforms: the three-place reference (TOC
X-Standard, bare README badge, CLAUDE.md "Standards compliance"), zero
retired dotted notation outside frozen bundles, every filename-§N citation
in range, and the gate statements in docs/testing.md and
docs/automated-tests/README.md already name both checkpoints.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TTuxA9sSHE1H7JLgJ4kYAx

commit 7cfb3472169f6462b3842e4d044f5462cebdef96
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 22 21:04:19 2026 +0530

Adopt the kit's US-English gate, and delete the copy this repo was keeping

LibKa0s v1.54.2 re-vendored, both payloads whole. The library bytes are identical to v1.53.0 --
kit revision 24 changes no library file -- so what actually arrives here is tests/_kit/test_prose.lua
and one line in tests/run.lua declaring it.

WHAT CAME OUT IS THE POINT. localization-5 says wire the kit's gate or your own and never both,
because two gates over one rule is two copies of a list each required to be carried WHOLE. So the
hand-written one goes: seven repositories in this collection had written that gate themselves under
three different filenames, which is why nothing could tell at a glance which repositories had one.

COVERAGE WENT UP ON THE SWAP, which is why this is not a lateral move. The hand-written gates
walked a glob of the addon's own source. The kit's reads the tracked set out of git, so the TOC,
the docs, the locale files and .luacheckrc are held to the same rule -- and its exclusions are a
named list rather than whatever a glob happened to miss. Everything the sweep corrects below was
invisible to the gate that was here yesterday.

WHERE A SPELLING IS NOT THIS REPO'S ENGLISH TO CORRECT it goes in tests/prose_waivers.lua, per FILE
and per WORD, with the reason beside it -- never per file alone, which would hide every other
British spelling in a file this repo edits often. A waiver is a debt where it stands in for work
that is owed, and it says so.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YRNj7EQPJHjuxLUH6jv4xk

commit 5e661644227fb528c21b20a991c31be3c9c49e0d
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 22 19:38:20 2026 +0530

Close out the documentation drift the sync pass reported but could not act on

The sync run surfaced every hit its own rules fenced it away from: comment citations, which need a
named replacement rather than a guessed one; count claims it could see but not settle; residual
British spellings; and one root doc that was genuinely owed. This is that list worked through, each
item either corrected against evidence in the tree or left alone with the reason recorded.

A comment naming a file that is gone is not automatically wrong. Most of them name the absent thing
ON PURPOSE -- "it used to be ours", "which is why that file is gone rather than emptied", "has not
existed since debug logging moved to LibKa0s" -- and correcting those would delete the explanation
they exist to give. They were read individually and left. What was corrected is the other kind: a
citation whose target moved, a count that no longer counts, a claim the cited line does not support.

Nothing that runs was touched. No version string moved, no export was deleted, no generated doc was
hand-edited, and the vendored payloads are untouched. Every correction was re-checked against the
code by a pass that did not write it, and the green gate was re-run from scratch afterwards.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YRNj7EQPJHjuxLUH6jv4xk

commit a5d3e22ac07c48238fefb63ac55cea3144b8cb6e
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 22 18:05:55 2026 +0530

Sync the docs to what the code actually does

A /wow-addon:sync-docs pass over the whole collection, each repo's drift found against its own
source and each correction carrying the evidence that settled it: a count re-counted, a file:line
re-grepped, a load order read back off the TOC, a doc-map row checked against the directory in both
directions.

Documentation only. No version string moved, nothing under libs/ or tests/_kit/ was touched, and no
comment was edited -- the comment-citation check ran and its hits are reported rather than applied,
because that correction needs a named replacement and there was nobody to name one. Dead exports and
missing Tier 1 docs are likewise surfaced, not acted on.

Every claim added or changed here was re-verified against the code by a second pass that had not
written it, and the green gate was re-run from scratch afterwards: several suites in this collection
read the docs, so a documentation edit can and does turn them red.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YRNj7EQPJHjuxLUH6jv4xk

commit ca833c4dc7d2c5ee060945f6a0438f8d65ea7113
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 22 17:11:29 2026 +0530

Re-vendor LibKa0s v1.53.0

OptionsWidgets minor 30 (LibKa0s-Options-1.0 23.30.3.7.3); every other file
unchanged from v1.52.0, kit at revision 23. Payload taken from the v1.53.0 tag
and verified identical, and the provenance line in CLAUDE.md moves with the
bytes rather than after them.

O.IdInput's Add button is now the height of the edit box beside it. It was
never mis-anchored -- AceGUI's labeled EditBox publishes alignoffset = 30 and
Flow honors it, so the two centers were one pixel apart. The button was 24 tall
against InputBoxTemplate's 20-tall border art, so it overhung 2px at each end,
and the overhangs do not read alike: the top one sits against the row's gold
caption and the bottom against empty dark, which is what reads as "higher".

Nothing to adopt: no new member, no new field.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YRNj7EQPJHjuxLUH6jv4xk

commit a39a540045a4dc22debb5cd85dd6543c901a0701
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 22 10:21:17 2026 +0530

Re-vendor LibKa0s v1.52.0

Both payloads copied whole from the v1.52.0 tag -- libs/LibKa0s/ and
tests/_kit/ -- and the CLAUDE.md provenance line rolled in the same
commit, from v1.51.0.

Minor 29 is mostly a fix to the library's own v1.51.0: the O.IdList help
mark reserved too little of the row for its absolute frame, so the column
fit correctly collapsed every list that adopted `help` to one column.
This addon adopts none of the help surface, so nothing here changes.

Gate: lint 0 warnings / 0 errors in 59 files; 646 passed, 0 failed, 0 skipped, 646 total

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YRNj7EQPJHjuxLUH6jv4xk

commit ee09708e9f599a79fe19363f89fba4b3745f8e72
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 22 02:23:29 2026 +0530

Re-vendor LibKa0s v1.51.0

Both payloads copied whole from the v1.51.0 tag -- libs/LibKa0s/ and
tests/_kit/ -- and the CLAUDE.md provenance line rolled in the same
commit, from v1.50.0.

Minor 28 adds three OPTIONAL O.IdList spec fields (entry.help,
spec.helpIcon, kind.suggestTag) which this addon does not use, and one
behaviour change that arrives without being asked for: every id-list row
lights under the cursor now, not only above one column.

Gate: lint 0 warnings / 0 errors in 59 files; 646 passed, 0 failed, 0 skipped, 646 total

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YRNj7EQPJHjuxLUH6jv4xk

commit 5ce09cb44048dcf310b3efd22a1fe7f447b0ec80
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Mon Sep 21 14:47:47 2026 +0530

Re-vendor LibKa0s v1.50.0

Both payloads copied whole from the v1.50.0 tag -- libs/LibKa0s/ and
tests/_kit/ -- and the CLAUDE.md provenance line rolled in the same
commit. The kit stays at revision 23, so tests/_kit/ is unchanged.

Carries v1.47.0 through v1.50.0 in one step, which is where this repo's
master had fallen behind. No surface this addon uses changed, so there is
nothing to adopt: v1.50.0's member manifest is identical to minor 26's,
and the majors this addon consumes are otherwise untouched.

Gate: lint 0/0 in 59 files, 646 tests 0 failed.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YRNj7EQPJHjuxLUH6jv4xk

commit 9d9dfde8f4fe993f351926fc7e32e19e03ecc5e9
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sun Sep 20 19:55:35 2026 +0530

Re-vendor LibKa0s v1.47.0

LibKa0s v1.47.0 is Options widgets minor 24: O.IdList gains an optional
`columns` spec field that packs entries row-major, default 1 and unchanged
behaviour. Both payloads moved whole -- libs/LibKa0s/ and tests/_kit/ -- and
CLAUDE.md's provenance line rolled in the same commit, which is what the
vendor-sync gate reads. Only libs/LibKa0s/OptionsWidgets.lua actually differs;
the kit stays at revision 23, so no case name moved and docs/test-cases.md
still counts 646.

The two-column option is DECLINED here, and there is nothing to decline it on:
this addon has no O.IdList / H.IdList call site at all. It tracks absorb totals
for player, target and focus -- there is no spell-id list, no whitelist and no
per-id store anywhere in settings/, defaults/, modules/ or core/. The only
mention of the member is the inert degraded-build stub in
settings/OptionsSetup.lua, carried because the owner asked for the LibKa0s
v1.35.0 render-time members on every consumer's stub, and the exemption
tests/test_surface_parity.lua already records for exactly that reason. A stub
takes no spec, so `columns` reaches nothing. No test was added: a pin on
row-major order needs a list to pin it to.

Green through the bounded runner: lua tests/run.lua 646/646, luacheck 0/0 over
59 files.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YRNj7EQPJHjuxLUH6jv4xk

commit 9d9dfde8f4fe993f351926fc7e32e19e03ecc5e9
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sun Sep 20 19:55:35 2026 +0530

commit 34a81208d16cc4c6aff3b5b3a97d9877ef0a2336
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sun Sep 20 02:36:39 2026 +0530

CLAUDE.md is the documentation-§2 stub; what only it said moves into docs/

Root CLAUDE.md had grown past the stub documentation-§2 describes: a docs-set
essay, working rules, and a long LibKa0s section. It is now the H1, the
adherence line, the canonical Standards compliance block, the read-the-docs
pointers, the green-gate line and the provenance line.

What was recorded nowhere else moved to its standard home:
- docs/ARCHITECTURE.md: the root-docs trio and the no-agent-context.md rule
(with the frozen-history note) under Documentation map; the seam-symmetry
invariant, with the load-completing OptionsSetup stub and loadDegraded().
- docs/testing.md: the provenance line as a gate input and its README history;
the whole-folder rule (Perf minor 12 needs Lifecycle); the never-edit rule
for libs/LibKa0s/ and tests/_kit/ and their exclude_files; CRLF on disk, the
two WSL mirror paths and the LF-to-CRLF sed.
- docs/common-tasks.md: Working conventions (terse replies, file_path:line,
why-only comments, no unrequested docs).

Content already in docs (the module list, the nine seams, NEEDS_WIDGETS, luac
-p, the test-case inventory and badge rule) was dropped from CLAUDE.md. The
module-map.md top-level-docs line and a tests/test_ltrap.lua comment that
pointed at the removed CLAUDE.md section now point at the new homes.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WkU3isSkvWPZkWVr51uZKo

commit cc6af608253f8764e52bf2b1e56d15cea236c24d
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sun Sep 20 02:33:25 2026 +0530

Comments follow the v1.46.1 combat lock and stop citing TOC line numbers

settings/Appearance.lua still said the library refuses to draw in combat and
closes the Settings window first. Since LibKa0s v1.46.1 a page shown in combat
is covered and draws nothing until PLAYER_REGEN_ENABLED, writes, Defaults and
tab switches are refused, and the window is never closed; the comment now says
so, matching the General.lua and Profiles.lua rewording in e510db6.

settings/OptionsSetup.lua, docs/common-tasks.md and docs/module-map.md cited
AbsorbTracker.toc :29 (widget.xml) and :65 (OptionsSetup.lua), which had
drifted to 31 and 75. They now name the TOC entries instead of line numbers.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WkU3isSkvWPZkWVr51uZKo

commit 22e97542aa3fabffde3be0a78358163f703ac697
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sun Sep 20 02:17:17 2026 +0530

Docs catch up with the Lifecycle and Launcher majors and the v5 ladder step

The seam paragraph in ARCHITECTURE.md still counted four setup files and
five descriptor-taking majors; Lifecycle and Launcher make it six and seven,
nine majors bound of the twelve vendored. module-map.md and performance.md
said eleven majors across sixteen files and left Lifecycle out of the load
order and OptionsTabs out of the Options file list.

Also: the Database row gains the { to = 5 } step, data-flow.md counts four
ladder rows, profiles.md gives the account-wide stamp's real default (1, not
4) and why, the kit row reads revision 23 with mock_record.lua and current
line counts, Bar.lua's row records the bars being created hidden, the
test_widgets row names the sidebar combat-lock case, and settings-panel.md
says combat lock where it still said refusal.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WkU3isSkvWPZkWVr51uZKo

commit e510db651d25f287a9c9c5d3442d2449dba85f57
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 19 21:52:27 2026 +0530

Re-vendor LibKa0s v1.46.1

LibKa0s v1.46.1 implements the standard v2.60.0 combat lock (options-ui-§2/§13,
anti-pattern #88). A settings page shown in combat is no longer closed: it is
covered (ctx.__combatCover), nothing renders, one gray COMBAT_LOCKED_NOTICE line
is printed per combat, and writes, Defaults and tab switches are refused until
PLAYER_REGEN_ENABLED, which draws the page from current state. The library
registers PLAYER_REGEN_* only while one of its pages is on screen
(lib.__shownPages). /at config still refuses in combat with COMBAT_REFUSED.
Kit stays at revision 23.

Host test changes:
- tests/test_widgets.lua: the sidebar combat test now asserts the new
behaviour for General and Appearance -- each page is covered, draws nothing
on a show in combat, the Settings window is not closed (no
SettingsPanel:Close, no HideUIPanel), and exactly one COMBAT_LOCKED_NOTICE
line is printed across both shows. It hides the pages afterwards.
- tests/test_disabled.lua: bringUp() closes every page an earlier suite left
on screen before taking its baseline, so the library's own PLAYER_REGEN_*
watch is not counted as an addon registration the stand-down must remove.

Docs and comments that described the old close-the-window refusal now describe
the combat lock (settings-panel.md, smoke-tests.md 13a, common-tasks.md,
module-map.md, settings/*.lua); docs/test-cases.md regenerated (646 cases).

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WkU3isSkvWPZkWVr51uZKo

commit e510db651d25f287a9c9c5d3442d2449dba85f57
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Sep 19 21:52:27 2026 +0530

This mod has no additional files