healer-man/PLAYTEST_NOTES.md

# Healer Man playtest notes

This file is the durable record of runtime behavior, regressions, and the
evidence required before calling a visual bug fixed. Update it whenever scene,
camera, input, physics, mob-rendering, or asset-loading code changes.

## 2026-09-27 Bite-Sized editor movement regression

- The original edit-mode action blocker also suppressed continuous movement and
  input snapshots, so WASD did nothing even after returning focus to the world.
  Movement/jump gating is now separate from combat-action gating. Panel controls
  and overlays still pause movement; Walk in scene or clicking the world returns
  keyboard focus while keeping the panel open.
- A focused headless Edge check with `BITE_SIZED_EDITOR_INPUT_ONLY=1` moved the
  player in all four WASD directions on the Stockades entrance stairs. Focusing
  a coordinate field stopped W movement, and clicking the world restored it.
  The character remained grounded after walking. An ability key started no cast
  in edit mode. Right drag changed yaw, pointer lock stayed null, and release
  stopped camera changes; no page errors occurred.
- Evidence: `playtest-artifacts/bite-sized-editor/editor-movement.png` and
  `movement-observations.json`. All 33 focused input tests, the production build,
  and release candidate check passed. The earlier authoring playtest verified
  input suppression and camera drag but missed actual walking; its script now
  includes this movement regression check.

## 2026-09-27 Bite-Sized spot editor

- `scripts/playtest-bite-sized-editor.cjs` used the real app in headless Edge
  with an isolated temporary API database and GM account. It created a named
  Stockades spot, added an imported Forsaken Abbey Carrion Spider, duplicated
  and removed an individual placement, captured anchors, and published through
  the GM API. The scene showed textured imported mobs and selectable placement
  rings; the collapsible scrolling panel kept its publish controls accessible.
- WoW Stockades and RoM Forsaken Abbey editing loaded grounded players, normal
  dungeon geometry and textures, mob previews, and five party-seat markers.
  Evidence: `playtest-artifacts/bite-sized-editor/wow-editor.png`,
  `cross-dungeon-preview.png`, and `rom-editor.png`.
- In draft test runs, the after-mobs boss stayed inactive until the trash died;
  the present-from-start boss was active immediately. Injected lethal combat
  damage exercised encounter completion, and the editor displayed Test complete
  with an untimed run. This verifies activation/completion, not encounter balance
  or a manually played full clear. Evidence: `boss-after-mobs.png` and
  `boss-present.png` in the same directory.
- Focusing the spot-name field prevented W movement. After collapsing the panel,
  right drag changed camera yaw, pointer lock stayed null, and releasing the
  button stopped camera changes. No mouse-capture prompt or page error occurred.
- Randomized runs entered both the Stockades and Forsaken Abbey parent maps with
  ready assets and pinned creature assignments. Evidence:
  `randomized-stormwind-stockade.png`, `randomized-forsaken-abbey.png`, and
  `observations.json`. These are local isolated runs; no production layouts were
  published by this check.

## 2026-09-27 Shadowfiend summon visibility

- The dedicated `scripts/combat-effects/shadowfiend-preview.html` page cast the
  level-80 priest spell and displayed `Shadowfiend summon active`. The live
  `CombatSummonPopulation` renderer loaded creature 19668's textured client
  FleshBeast model, with its feet near the purple ground marker after placement
  adjustment. Evidence: `playtest-artifacts/shadowfiend/summon-preview.png`.
- The isolated model preview also loaded the same graph as one mesh layer. The
  combat test verified a fiend-owned damage event and 5% maximum-mana return
  from the successful hit. Full dungeon entry was not part of this check.
- After moving model preparation into the Priest gameplay transition, the
  preview waited for the same shared media cache before casting. In a fresh
  Edge session, the graph and texture requests completed before the cast
  (`15165.5 ms` and `15251.5 ms` versus cast at `15265.5 ms`); no page errors or
  frame gap over 50 ms occurred after the cast. The source model remained visible
  in `playtest-artifacts/shadowfiend/warmup-check.png`. This isolated preview does
  not establish the full dungeon frame rate.

## 2026-09-26 RuneWaker dungeon minimap and M-key map

- Created a fresh RoM Rogue, downloaded Forsaken Abbey, entered as damage, and
  waited for the boss preview to finish. The minimap displayed the extracted
  client map; moving backward changed the live player-marker coordinates.
- Pressing M opened the full map with the same artwork, player heading, entrance,
  and eight numbered encounters. Pressing M again closed it. No browser page
  errors were reported.
- Inspected `artifacts/playtest/runewaker-maps/forsaken-minimap.png` and
  `forsaken-full-map.png`. The map was readable and did not cover the normal
  playfield until opened. Navigation floorplans were checked by generated-image
  inspection and catalog tests, not by live dungeon entry.
- Validation: 24 focused map/UI tests, TypeScript, production build, content
  publication, and release candidate check passed.

## 2026-09-15 Reported Warder Stilgiss and Zereketh issues

- Reproduced the two submitted production reports in isolated Boss Testing
  sessions. Warder Stilgiss's imported Frostbolt and Frost Nova were damage-only,
  and Zereketh's extrapolated level-two health was low enough for the opening
  party volley to trigger his native rising death effect immediately.
- Warder now applies an eight-second 40% movement slow with Frostbolt and an
  eight-second movement-only root with Frost Nova. Both retain their original
  damage, create visible magic debuffs, and apply to the player/party movement
  paths without blocking actions while slowed or rooted.
- Zereketh now enters a level-one Boss Testing room at 540 health instead of
  using the tiny extrapolated native value. The browser scene retained him at
  540/540 after its three-second settle window; the native death effect appeared
  only after the playtest deliberately dealt 540 lethal damage.
- Evidence: `playtest-artifacts/reported-boss-fixes-2026-09-15/results.json`,
  `warder-stilgiss-rooted.png`, `zereketh-living.png`, and
  `zereketh-death-350ms.png`. The desktop scene showed the frost impact/root
  presentation and a living Zereketh, with no browser runtime errors.
- Validation: the complete 1,762-test Vitest suite passed (six skipped), along
  with TypeScript, the production asset audits/build/bundle verification, and
  the isolated two-boss browser playtest.

## 2026-09-14 New-character PC bars and missing-spell borders

- A temporary browser fixture used the real desktop bars, ability controls,
  spellbook, combat store, and character UI activation with memory-only saves.
  Fresh Priest and Mage profiles populated their PC bars automatically. The
  Mage started with Frostbolt, Fire Blast, Frost Nova, and Arcane Intellect.
- Clearing Lesser Heal gave its spellbook entry and ability-picker entry a
  visible light blue border, including the icon border. Both computed border
  colors were `rgb(146, 213, 255)`. Assigned and locked spells kept their usual
  borders. Reassigning Lesser Heal removed the missing-spell border.
- Switching to the fresh Mage and back retained the Priest's cleared slot.
  Screenshots of both missing-spell menus were inspected during the playtest.
  The temporary fixture was removed afterward.
- Validation: 62 focused tests, TypeScript, production build including bundle
  verification, and the release-candidate check passed. Full-game verification
  was limited by a persistent opening-expedition loading screen in both local
  development and production; menu visuals were verified in the fixture.

## 2026-09-14 Ability hover tooltips

- Entered Stormwind Stockade with the existing offline level-17 Priest through
  the normal roster, dungeon, and role screens. Hovering Smite on the controller
  bar displayed its description, Mana cost, cast time, cooldown, range, and
  target. Leaving the button removed the tooltip.
- Opened Ability controls and its ability picker. Hovering Shadow Word: Pain
  displayed its own description without assigning it. Flash Heal was readable
  by hover and keyboard focus despite its level-20 lock. Switching from keyboard
  inspection to another hovered ability left exactly one tooltip visible.
- Checked the picker at the default 650 x 490 viewport and at 1280 x 800.
  Tooltip DOM bounds remained inside the viewport. The in-app screenshot
  scales unusually with a viewport override; reset the override afterward.
- Enabled desktop bars in an unsaved UI-layout draft. Hovering the leftmost
  Smite button showed the tooltip above the bar, clamped to the left edge and
  outside the HUD's transformed layout container. Discarded the layout draft.
- A temporary Flash Heal assignment during input testing was restored to the
  original Smite assignment. No character or template was created or deleted.
- Screenshot evidence: `playtest-artifacts/ability-tooltips/desktop-bar.png`
  and `playtest-artifacts/ability-tooltips/locked-picker.png`.
- Automated verification: all 64 UI tests passed, including nine tooltip
  content, unavailable-control, and viewport-placement regression tests.
  TypeScript, the production build, and the release candidate check passed.

## 2026-09-10 Personal UI templates and desktop action bars

- Opened Wailing Caverns through the normal offline roster, dungeon, and role
  selection screens, then opened Pause > Edit UI & templates.
- Verified five bars containing 12 buttons each, including 2 x 6 and 3 x 4
  arrangements. Default stacking accounts for each bar's height. The editor
  can collapse or dock on either side so drag handles remain reachable.
- Clicked Renew and pressed Shift+Q in keybind edit mode. Reassigned Shift+Q
  to bar 5/button 12 and verified the previous keybind cleared. Assigned Renew
  to that button through the ability picker; after saving, Shift+Q cast it and
  displayed its cooldown. No spell fired during key capture.
- Dragged the minimap, hid party frames, saved/renamed/overwrote PC slot 1,
  loaded it, and round-tripped all layouts through export text and import.
  The imported JSON retained the minimap position, hidden party, bar geometry,
  ability assignment, key chord, and template name. Restoring a hidden widget
  works, and Cancel discards that visibility edit.
- At a 390 x 844 viewport, the editor remained within the viewport and the
  Save UI button was visible and 36 CSS pixels high. The in-app browser's
  screenshot preview scales the page unusually at this override; DOM bounds
  confirmed the editor was 310 CSS pixels wide. This was responsive browser
  verification, not an Android hardware playtest.
- Evidence: `playtest-artifacts/ui-layout-desktop.png` and
  `playtest-artifacts/ui-layout-phone-viewport.png`. Restored the offline UI
  to defaults and removed the temporary template via import after testing.
- Automated coverage includes migration to six slots, owner isolation,
  cloud upload/hydration into a separate device cache with a mocked endpoint,
  malformed imports, key conflicts, and input precedence. No production
  account credentials were used for the cloud round-trip test.
- `npm run build` and the focused 67-test suite passed. The development server
  was restarted after an existing watcher hit EBUSY on generated web-bootstrap
  assets during a concurrent build; no runtime UI error was logged in the
  subsequent playtest.

## 2026-08-18 GM Manastorm administration - COMPLETE

- Server-authorized `Phenom` access, revisioned SQLite configuration, safe
  all-Needs-tested defaults, public/offline caching, optimistic conflicts, and
  denial/revision logging were exercised through the production HTTP surface.
- The GM hub exposed all 406 deduplicated boss records, readiness and runnable
  state, pool filters, six global switches, dungeon entry, placement editing,
  and GM tests. A fresh revision correctly reported zero Ready bosses.
- Live Wailing Caverns authoring captured and nudged the shared group anchor,
  displayed derived party, boss, portal, and exact linked-mob data, and saved a
  three-mob / two-kill Chaotic Link stage. The desktop Save control was moved
  above the Tactical control after the playtest found their hit targets
  overlapping; a normal semantic click then published the next revision.
- Promoting an imported encounter directly to Ready initially exposed missing
  stage materialization at the server boundary. The readiness transition now
  persists its complete inherited stage. Cross-mode duplicates also merge
  their mode memberships, so a boss imported for both leveling and end-game is
  represented once and remains eligible in both modes.
- Live pool acceptance promoted Lady Anacondra and Burning Felguard, entered a
  level-1 public Manastorm, then marked the active Lady Anacondra record Not
  ready from a parallel GM client. The existing run retained Lady Anacondra and
  its two-stack Chaotic Link snapshot; after reload, the next run selected only
  Burning Felguard. This verified new-run publication and active-run
  immutability together.
- Desktop and Thor-preview passes rendered the DOM editor over the dungeon,
  retained camera/game separation while the panel was open, and exposed the GM
  pause-menu entry. Console inspection found no errors and only the existing
  third-party initialization deprecation warning.
- Verification passed: server **7/7**, full Vitest **121 files / 699 tests**,
  TypeScript, production build, and the KTX2 audit of 11,839 compressed textures
  with zero fallbacks. The build retained the existing informational
  large-chunk advisory.
- Evidence: `playtest-artifacts/2026-08-18-gm-manastorm-admin/desktop-admin-hub.png`,
  `desktop-world-editor-final.png`, and `thor-world-editor.png`.

## 2026-08-17 authentic combat VFX and audio - COMPLETE

- Player, party, and enemy combat actions now publish transient presentation
  phases without changing combat resolution. The scene consumes those events
  through bounded pooled effects (High: 64 systems / 8,192 particles; Low: 24
  systems / 2,048 particles with no effect lights), while Off skips visual
  spawning. Actor-group switches clear live systems and voices immediately.
- The generated source catalog covers 6,383 runtime spell/skill IDs: 5,144
  resolve to preserved WoW/Ascension/RuneWaker descriptors and the remaining
  1,239 are explicit school/delivery fallbacks. All 38 referenced source sound
  paths resolve to 32 deduplicated packaged clips (6,365,096 bytes); the audit
  reports no missing packaged assets.
- Owner settings migrated to local profile database v3 and remain independent
  of character gameplay settings. Options exposed High/Low/Off, independent
  Player/Party/Enemy visual switches, and independent audio mute/volume rows.
  The existing master volume also controls combat and Manastorm cues.
- Live Wailing Caverns verification entered through the offline roster as a
  Human Priest healer. The entrance fight visibly presented party melee/ranged
  effects and enemy actions. A High-quality Smite capture showed the player
  cast system alongside party effects; Low-quality Lesser Heal showed the
  reduced healing system; Off-quality Shadow Word: Pain showed combat state,
  aura/combat text, and damage continuing without spawned VFX. Defaults were
  restored to High after the pass.
- At a 390 x 844 mobile viewport, the Options dialog measured 362 px wide with
  345 px of scroll content and no document overflow. Player frame, party list,
  minimap, and action bar all remained inside the viewport. A discovered narrow
  Options text/control overflow was fixed by allowing grid labels to shrink and
  moving combat-audio controls onto a full-width narrow-screen row.
- Verification passed: combat-effects parser/coverage audit **4/4**, focused
  persistence/event/cue/UI suites **68/68**, full Vitest **118 files / 674
  tests**, TypeScript, production build, and Android web build. Both builds
  retained the existing informational large-chunk advisory; the browser console
  had no errors and one existing third-party initialization deprecation warning.
- Evidence: `playtest-artifacts/2026-08-17-combat-presentation/desktop-high-combat.png`,
  `desktop-player-smite.png`, `desktop-low-heal.png`, `desktop-off-combat.png`,
  `desktop-options.png`, `mobile-combat-controls.png`, and
  `mobile-high-combat.png`.
- Coverage limit: browser automation exercised Web Audio creation, group
  routing, mute, volume, and authentic clip loading but cannot judge perceived
  loudness or spatial balance; that final acoustic judgment still requires a
  listening pass on physical desktop/mobile speakers.

## 2026-08-17 dungeon proximity aggro and range overlay - COMPLETE

- Idle dungeon enemies now check the living player and party positions during
  the low-frequency combat tick. Crossing a mob's radius engages its complete
  linked pack through the normal threat path, anchors each member at its pull
  position, and seeds initial threat on the nearest actor who crossed the line.
- Normal enemies default to a 16-unit aggro radius and bosses to 20 units;
  creature definitions can override either value. Proximity uses full 3D
  distance so vertically overlapping dungeon floors do not pull one another,
  while the existing planar combat leash continues to support uneven terrain.
- Options now includes `Mob aggro ranges`, defaulting to Off and persisting with
  the rest of the account gameplay settings. When enabled, idle dungeon mobs
  render restrained orange world-space boundaries above uneven terrain. Rings
  disappear once their mob engages and are never rendered in Manastorm.
- Focused verification passed: `src/game/combatStore.test.ts` and
  `src/game/mobAi.test.ts`, **2 files / 60 tests**. `npx tsc -b --pretty false`
  passed, and `npx vite build` passed with 960 transformed modules.
- Live browser verification entered Wailing Caverns as a healer through the
  offline roster. With the party left in Defend and no player attack issued,
  the entrance pack automatically acquired and damaged Garrick. The Options
  dialog showed the new setting Off by default; matched scene captures then
  confirmed visible range rings when On and their removal when toggled Off.
- The browser console contained no errors and one existing third-party
  initialization deprecation warning. The local Vite host separately recorded
  the known optional `/content/manifest.json` proxy refusal because the content
  server on port 4173 was not running; packaged dungeon play was unaffected.

## 2026-08-17 Kresh cliff-pull aggro regression - COMPLETE

- Reproduced the reported reset with a deterministic combat-store regression:
  Kresh registers at one end of his long patrol, moves to a later patrol point,
  takes damage while the player is falling well above him, and advances combat
  after the player reaches the lower grotto.
- Roaming mobs now anchor their leash at their runtime position when a new pull
  starts instead of retaining the position where their model first mounted.
  Aggro leash eligibility uses planar dungeon distance, matching the existing
  mob-AI rule, while attack reach still includes vertical separation. A real
  horizontal leash break continues to evade and restore the encounter.
- Focused verification passed: `src/game/combatStore.test.ts` and
  `src/game/mobAi.test.ts`, **2 files / 58 tests**. `npm run build` also passed,
  including TypeScript, 956 transformed Vite modules, and the KTX2 audit of
  11,839 compressed textures with zero fallbacks.
- Production-preview smoke testing entered Wailing Caverns through the offline
  roster, rendered the full dungeon/HUD/party view, issued the Attack order,
  and observed live enemy damage on the tank. This checkout lacks the optional
  local playable-character and creature model packs, so the documented
  procedural fallbacks and their asset warnings were visible during this pass.
- Coverage limit: the exact cliff jump was validated by the deterministic
  combat-state regression rather than by manually traversing from the entrance
  to Kresh during the browser smoke test.

## 2026-08-16 responsive jump and natural teammate movement - COMPLETE

- Added the grounded single-jump path for Space and controller L3, including a
  `7.2 m/s` launch, `100 ms` coyote time, `120 ms` input buffering, landing
  lockout, upward-normal ground filtering, and blocked-input cancellation.
  Player physics and the latest grounded navigation anchor are now reported
  separately, so party breadcrumbs and recalls do not follow an airborne
  player position.
- Player locomotion now presents rising, falling, and landing phases. The live
  WoW human used its native jump pose while preserving the visible body and
  equipped weapon; RuneWaker and fallback clip selection is covered by the
  focused animation suite.
- Party members now derive stable movement personalities from member ID and
  living-party rank. The runtime applies personalized trail distance, lateral
  offset, comfort slack, speed, slow sway, stop/start hysteresis, combat
  bearing, preferred range, and ally separation only when the candidate path
  passes collision and floor-continuity checks. Breadcrumb centerline routing,
  Stop, stuck recovery, role decisions, and blocked-route reporting remain the
  fallback behavior.
- Focused verification passed 17 files / 117 tests:
  `playerJump`, `inputManager`, `inputMath`, `store`, `combatAnimation`,
  `partyMovement`, `partyNavigation`, `partyPathing`, `partyRuntime`, and the
  complete `src/avatar` suite. The first full-suite attempts exposed a
  pre-existing load-sensitive `30 s` timeout in the heavyweight RuneWaker
  animation audit; the isolated test passed in `6.5 s`. Its two asset-audit
  cases now use `60 s`, matching their actual full-suite workload more closely.
- Final `npm test`: **114 files / 656 tests passed**. Final `npm run build`:
  **passed**, including TypeScript, 954 transformed Vite modules, and the KTX2
  audit of 11,839 compressed textures with zero fallbacks.
- Production preview was exercised at `http://127.0.0.1:4174/`. At desktop
  `1440 x 900`, Space produced a visible jump from rest and while moving; the
  running capture shows the native raised-arm jump pose. Space pressed while
  paused did not launch on resume. Five rapid Space presses during one airborne
  cycle returned to a normal grounded pose rather than producing a double jump.
  The entrance smoke route covered its sloped floor and low cavern roof without
  visible clipping or a hidden avatar.
- In Defend, the companions settled at visibly different longitudinal and
  lateral offsets. In Attack, tank, ranged, and melee actors held distinct
  bearings; Brynn peeled to the right while the other roles retained separate
  safe positions. Party health and combat activity continued updating. Stop and
  Recall were then issued from the Thor companion display and the authoritative
  command state updated correctly.
- The browser Thor preview displayed both the `960 x 540` main surface and
  `620 x 540` companion surface. Entering as healer from the companion display,
  Attack/Stop/Recall dispatch, Space jumping on the main display, and
  right-mouse input with no pointer lock all worked. Runtime inspection found
  one visible canvas, no loading screen, and no scene-error overlay.
- Console capture contains no errors. It contains two occurrences of the
  project's existing third-party initialization deprecation warning, one per
  desktop/Thor runtime load. The Vite preview host separately logged the known
  optional `/content/manifest.json` proxy refusal because the auxiliary content
  server on port 4173 was not running; the packaged dungeon still loaded and
  played normally.
- Evidence is under `playtest-artifacts/2026-08-16-jump-party/`:
  `desktop-running-jump-preview.png`, `desktop-varied-follow-preview.png`,
  `desktop-varied-combat-preview.png`, `desktop-pause-jump-blocked-preview.png`,
  `thor-jump-preview.png`, `thor-varied-combat-preview.png`,
  `thor-repeated-jump-landing-preview.png`, and `preview-console.txt`.
- Coverage limits: no physical controller was connected, so L3 edge behavior
  and removal of L3 clear-target are automated-test verified rather than
  hardware verified. The entrance smoke route did not provide a deterministic
  walk-off ledge or narrow doorway fixture; exact coyote timing and lateral
  centerline fallback remain covered by their pure/navigation tests rather than
  claimed as live edge-case evidence.

## 2026-08-01 all-dungeon v5 animation acceptance pass - COMPLETE

- The Blackhorn Silencer and Giant Assassin failures demonstrated that valid
  files, finite tracks, expected clip names, and green Khronos validation do not
  prove a correct skeleton pose. The v4 exporter repair changed the reflected-Z
  quaternion conversion from `(x, y, -z, w)` to `(-x, -y, z, w)`. The v5 cache
  identity retains that repair and invalidates the actor cache for complete cast
  semantics: all exported cast-family clips must be preserved and reachable by
  the runtime selector.
- The authoritative scope is 30 dungeon packages, 436 entity templates, 268
  base actor configurations, and 322 runtime-referenced namespaced shipping
  variants. The audit and visual-review unit is the 322 variant paths, not the
  268 base configurations. Expected configuration is 321 native variants plus
  one pose-only Mantarick mapping (Hall of Survivors template `102452`, asset
  `judgement-light-pose`).
- `demon-stronghold-125` and `zurhidon-stronghold-124` are authoritative empty
  copy zones and require explicit zero-row review entries. Sardo Castle's Rune
  Warning Device (template `105758`, asset `rune-device-red`) is a native,
  stationary actor whose missing locomotion clip is an approved source-faithful
  exception; its other native animation families still require review.
- Required machine checks, run from `D:\Projects\HealerMan`, start with:

  ```powershell
  npm run runewaker:audit:animation-inventory
  npm run runewaker:audit:runtime-clips
  npm run runewaker:audit:animation-readiness
  npm run runewaker:audit:animations
  ```

  The exact PowerShell that derives all 322 GLB inputs from
  `runtime-clip-coverage.json`, runs Blender, generates contact sheets and review
  pages, and compiles the human disposition is documented in
  `scripts/runewaker-pipeline/POSE_AUDIT.md`.
- Blender numerically samples the bind pose plus the first, middle, and final
  frame of every action. This can miss a defect between samples and cannot prove
  semantics, timing, impact alignment, particles, audio, or encounter mechanics.
  Every variant contact sheet therefore requires human review; suspicious
  actors require `--render-mode all`, and every dungeon still requires an
  in-browser combat pass covering idle, applicable locomotion, configured
  attacks/casts, wound, death, transitions, and console/WebGL errors.
- **Rebuild result:** All 29 generic population recipes rebuilt green. Forsaken
  Abbey's separate nine-actor package completed Khronos validation with zero
  errors and zero warnings across all nine GLBs.
- **Inventory/runtime result:** The final inventory is green at 30 dungeons, 436
  entity templates, 268 base actor configurations, and 322 runtime-referenced
  variants with zero blockers. Runtime semantic near misses are zero. The only
  required-family exceptions are the documented pose-only Mantarick and the
  native stationary Rune Warning Device with no locomotion.
- **Pose result:** `artifacts/animation-audit/post-v5/summary.json` evaluated
  44,305 poses for all 322 variants: 282 numeric passes, 40 accepted-warning
  candidates, zero errors, zero failures, and 634 flagged samples. Those values
  exactly match the visually reviewed v4 batch totals, so the v5 cast-semantics
  rebuild introduced no aggregate numeric pose regression.
- **Offline browser result:** The packaged/offline catalog displayed all 30
  RuneWaker dungeons as Ready. Pasper's Shrine and Hall of Survivors both loaded
  their environments. The console contained only an unrelated library
  deprecation warning; no asset, skeleton, animation, WebGL, or dungeon-load
  error was observed.
- **Test boundary:** All 19 focused RuneWaker animation tests pass. Repo-wide
  build/test gates remain blocked by the unrelated unfinished RoM class catalog
  and missing Manastorm fixtures; those failures are not attributed to the v5
  animation work.

## 2026-08-01 Pasper goatman quaternion-basis repair

- Reproduced the severe animated-pose deformation on Giant Assassin (template
  `101039`), Giant Assassin Captain (`101045`), and Blackhorn Silencer
  (`101053`). The first two use
  `model/character/creature/goatman/act_fog_ferocity_03.ros`; the Silencer uses
  `model/character/creature/goatman/act_fog_loster_03.ros`.
- The source-to-HealerMan position conversion reflects Z. The exporter had
  incorrectly mapped rotations as `(x, y, -z, w)`. Applying the basis transform
  `R' = S R S`, with `S = diag(1, 1, -1)`, requires
  `(-x, -y, z, w)`. The corrected conversion is shared by skeleton bind
  transforms and every sampled animation frame.
- Rebuilt Pasper's nine actors and validated all nine shipping GLBs with zero
  Khronos errors and zero warnings. Then rebuilt all 29 generic dungeon actor
  packages plus the separately handled Forsaken Abbey package; every package
  completed its actor and Khronos validation stage green.
- Blender pose evidence is under
  `artifacts/animation-diagnostics/fog-ferocity-03-runtime-clips` and
  `artifacts/animation-diagnostics/fog-loster-03-runtime-clips`. The exact
  runtime-selected one-handed idle, run, `1h_attack01`, and `1h_attack02` clips
  retain coherent bodies, limbs, armor, and weapons at their first, middle, and
  final samples. The corresponding pre-fix directories preserve the malformed
  poses for comparison.
- Pasper's Shrine loaded in the local browser and the party entered combat. The
  browser console contained no model-load, skeleton, or animation exceptions.
  The automated camera remained at the entrance while the ordered party moved
  ahead, so no browser screenshot of these specific goatmen is claimed; the
  actor-specific Blender renders are the visual regression evidence.
- Added `blender/render-animated-actor.py` and documented the required visual
  pose gate. Clip presence, semantic-family coverage, finite tracks, and
  Khronos-clean files are not sufficient evidence of correct skeletal motion.

## 2026-07-31 Pasper's Shrine original skeletal-animation pilot

- Replaced the nine static Pasper combat packages with skinned GLBs generated
  from their original GR2 skeletons and ACT-referenced RAS controls.
- The actors expose 20-24 clips each. Generated entity data marks all 12 Pasper
  templates as native; combat selection covers attack variations, cast, wound,
  and death in addition to stand/idle, walk, and run.
- Khronos validation passed all nine shipping GLBs with zero errors and zero
  warnings. The focused animation/runtime/pipeline suites passed 19 tests. The
  final full run passed 513 of 518 tests; its five unrelated failing assertions
  cover missing Manastorm conversion inputs plus established Forsaken/Kalin
  availability state, and the unrelated Manastorm compound-import suite still
  has its pre-existing parser error. The production TypeScript/Vite build passed.
- The local Vite entry started at http://127.0.0.1:5173/. The in-app browser
  controller timed out during connection and after a clean reset, before a tab
  or screenshot could be obtained. No live visual result is claimed in this
  entry; an interactive Pasper combat pass remains the final visual gate.
- Reproduction and remaining-dungeon rollout instructions:
  scripts/runewaker-pipeline/ORIGINAL_ANIMATION_IMPORT.md.

## 2026-07-31 RuneWaker all-dungeon animation audit

- Three parallel read-only audits covered all 30 packaged RuneWaker instances.
- All 239 actor GLBs parsed successfully and contain zero native clips/skins.
- All 312 static-model entity definitions now select full procedural animation;
  132 dynamic-display entities use explicit procedural proxy bodies.
- Automated coverage opens every referenced GLB, resolves all 4,094 spawn
  references, and exercises finite idle, movement, attack/cast, wound, and
  terminal death transforms.
- The audit exposed and fixed the generalized generator's missing procedural
  mode and the runtime wrapper's missing wound-revision handoff.
- Focused command: npm run runewaker:audit:animations.
- Live Vite boot succeeded at http://127.0.0.1:4173/. The Codex in-app browser
  connection timed out twice before a tab could be selected, so no new WebGL
  screenshots were captured in this pass. Do not treat the automated results
  as screenshot evidence.
- Detailed matrix:
  scripts/runewaker-pipeline/RUNEWAKER_ANIMATION_AUDIT.md.

## Evidence rules

- `npm test`, `npm run build`, Android compilation, and APK signing prove logic
  or packaging only. They do **not** prove that the Three.js Canvas renders a
  dungeon before or after pointer lock.
- A pointer-lock fix is not verified without matching screenshots from the
  running game plus browser console/runtime checks.
- `assetStatus === "ready"` is not enough. `ProductionDungeon` latches that
  status after an initial frame; a later camera or frame-loop failure can still
  leave the HUD over the Canvas clear color.
- The Canvas clear color is `#07110d`. A nearly uniform dark-green playfield
  with a normal DOM HUD means the app shell is alive but the 3D view is not
  producing visible geometry.
- Record what was actually observed. Mark unavailable platforms and missing
  screenshots as unverified rather than inferring success from another test.

## Source and staging safety

- Installed game root: `C:\Users\phenom\Documents\Healer-Man`
- Current writable mirror:
  `C:\Users\phenom\Documents\wow335a\dungeon-export-work\HealerMan\gameplay-ui-staging`
- Treat `.healer-man-staging` and
  `dungeon-export-work\HealerMan\runtime-staging` as stale historical mirrors.
  They contain older pointer-input/camera files and must not be bulk-copied over
  the game.
- Before the pointer-lock transition patch on 2026-07-21, installed
  `PlayerRig.tsx` matched the gameplay UI mirror at
  SHA-256 `B918B24031DD7031F1EEBA75879E8291EABDE9AAB8A08223B9B392A105B4692A`.
  Installed `inputManager.ts` matched the pre-patch mirror at SHA-256
  `16D77BF075BD1355A23F81D019B8D34C7427026B113DBD257B4A7BE6FE53BF1F`.
- After the transition patch, the installed and staged hashes match:
  `PlayerRig.tsx` is
  `A2CD3D3951D99ECE05D99ACB3CAE53C412B075B01106DBAC0E37EC48B8AFC843` and
  `inputManager.ts` is
  `355A4AD7DE12AF777B4829428B32181F9511A8DF021102932036DAA643C3C716`.
- Compare source and destination hashes before and after every elevated copy.
  Copy only the files intentionally changed. The installed root currently has
  no Git history to rescue an accidental stale overwrite.

## Pointer-lock invariants that must remain intact

1. `GameScene.tsx` requests pointer lock on the Canvas itself.
2. `inputManager.ts` rejects non-finite mouse movement, clamps each event and
   frame, clears the buffer on focus changes, and suppresses the complete
   180 ms cursor-recentering burst after pointer lock begins.
3. `PlayerRig.tsx` repairs non-finite yaw, pitch, player translation, cast
   distance, and camera boom state before they reach Three.js or Rapier.
4. `ThirdPersonCamera` uses the default `useFrame` priority. A positive priority
   takes ownership of React Three Fiber's render loop; without an explicit
   `gl.render(scene, camera)` call it produces the exact clear-color-plus-HUD
   failure.
5. Errors thrown from a `useFrame` callback are not reliably surfaced by the
   DOM `SceneErrorBoundary`. Browser `pageerror` and console capture are part of
   the playtest, especially after adding new frame callbacks.

## Deterministic PC setup

Run the latest source, not an old Vite process or stale `dist` directory:

```powershell
npm run dev -- --host 127.0.0.1 --port 4173 --strictPort
```

Open `http://127.0.0.1:4173/` at `1440 x 900`. For a repeatable offline
character, run this in DevTools before reloading:

```js
const now = Date.now();
localStorage.setItem("healer-man.profile-database.v1", JSON.stringify({
  version: 1,
  accounts: [],
  rosters: {
    "offline-roster": [{
      id: "character-playtest",
      ownerId: "offline-roster",
      name: "Remedy",
      raceId: "human",
      classId: "priest",
      gender: "female",
      appearance: { skinColor: 0, face: 0, hairStyle: 0, hairColor: 0, feature: 0 },
      level: 10,
      experience: 0,
      talentRanks: {},
      location: "Wailing Caverns Entrance",
      createdAt: now,
      lastPlayedAt: now,
    }],
  },
}));
sessionStorage.setItem("healer-man.session.v1", JSON.stringify({
  kind: "offline",
  ownerId: "offline-roster",
  displayName: "Offline",
}));
location.reload();
```

## Required pointer-lock visual checklist

Store evidence under `playtest-artifacts/2026-07-21/` for this regression.

1. Select Remedy, click **Enter world**, and wait until `.loading-screen` is
   gone and `.scene-shell canvas` is visible.
2. Confirm textured dungeon geometry is visible, not only `#07110d`.
3. Record pre-lock state: `assetStatus` is `ready`, `pointerLocked` is `false`,
   and `playerPosition` plus `cameraYaw` are finite.
4. Capture the full window as `pointer-lock-before.png` and the Canvas crop as
   `pointer-lock-before-canvas.png`.
5. Click the center of `.scene-shell`. Confirm
   `document.pointerLockElement === document.querySelector(".scene-shell canvas")`.
6. Wait two animation frames and at least 500 ms. Confirm `pointerLocked` is
   `true`, player/camera values remain finite, the WebGL context exists, and
   `gl.isContextLost()` is `false`.
7. Capture the same views as `pointer-lock-after.png` and
   `pointer-lock-after-canvas.png`.
8. Move the mouse, then hold `W` for about 250 ms. Capture
   `pointer-lock-after-input.png`. This is required because a bad recenter event
   can fail on the first post-lock input rather than on the click itself.
9. Press Escape, open and close a menu, reacquire pointer lock, and confirm the
   dungeon remains visible.
10. Save console output as `pointer-lock-console.txt`. The pass condition is no
    `console.error`, `pageerror`, `pointerlockerror`, or `webglcontextlost`
    during the transition.

Visual pass criteria: all three full-window images show textured dungeon
geometry behind the HUD; the two post-lock images show a reticle; no loading or
scene-error overlay is present; the view changes after input without becoming
uniform dark green.

After the development-server pass, repeat against a fresh production build:

```powershell
npm run build
npm run preview -- --host 127.0.0.1 --port 4174 --strictPort
```

Use filenames ending in `-preview` for the production screenshots.

## Known evidence as of 2026-07-21

### Working or previously observed working

- Login/offline roster, character selection, and entering Wailing Caverns have
  been observed working by the user.
- Before this regression, the user could see and navigate the dungeon and give
  camera-clipping, pop-in, wall-material, mob, and performance feedback. This
  proves that the extracted dungeon and runtime can render on this PC.
- The current source contains pointer-delta sanitization tests and finite camera
  recovery guards.
- After the transition patch, `npm test` reported 109/109 tests passing and
  `npm run build` completed successfully. An Android APK had been produced by
  the preceding build, but has not yet been rebuilt from this patch.
- `scripts/probe-entrance.mjs` reported finite entrance camera geometry and a
  visible shell hit. That static probe does not exercise React, pointer lock, or
  the live frame loop.

### Not working or not yet verified

- User report: after entering Wailing Caverns, clicking the game to take focus
  changes the 3D view to dark green while the HUD remains visible.
- No current before/after screenshots have yet demonstrated that the latest
  source survives pointer lock. Until the required screenshots exist, this
  regression remains open regardless of tests/build status.
- Browser console and WebGL context-loss evidence for the failing transition
  has not yet been captured.
- The production preview answered HTTP 200 at `http://127.0.0.1:4174/`, but the
  visual-test browser was not attached to this session. No screenshot was
  captured, so the patch must not be called visually verified yet.
- The pointer-lock regression has not been visually rechecked in production
  preview, the Android WebView, or the AYN Thor top display.

## Regression log

### HM-2026-07-21-POINTERLOCK-CLEARCOLOR - PATCHED, VISUAL VERIFICATION REQUIRED

**Reported:** 2026-07-21

**Symptom:** The dungeon is visible after character selection. Clicking the game
window to take focus leaves a dark-green Canvas with the UI still visible.

**Expected:** Pointer lock activates mouse look and the textured dungeon remains
visible and navigable.

**Important history:** Two earlier failure classes can create this symptom:

- A positive `useFrame` render priority can disable React Three Fiber's
  automatic render call.
- A stale or malformed pointer-lock movement event can poison orbit/camera
  state with non-finite values.

The prior input guard discarded only one transition event. Browsers can emit a
burst of recentering events, so later stale deltas could still rotate the camera
away from the rendered cavern. The patch discards the full 180 ms transition
burst and reads `pointerLocked` imperatively in the player frame, avoiding a
Rapier subtree reconciliation at lock acquisition. This is the leading cause,
but it remains a hypothesis until the required screenshots and runtime evidence
are captured. Recent mob/combat work also added more high-frequency `useFrame`
paths; an exception or WebGL context loss can create the same visible symptom.

**Patch files:** `src/game/inputManager.ts`,
`src/game/inputManager.test.ts`, `src/scene/PlayerRig.tsx`.

**Automated evidence:** 17 test files and 109 tests passed. The TypeScript/Vite
production build completed with 655 modules transformed. These results validate
the patch logic and bundle, not the rendered pointer-lock transition.

**Visual evidence status:** blocked on an attached visual-test browser. The
local production preview is running and ready for the checklist above; no
screenshots exist yet.

**Fix evidence required:** `pointer-lock-before.png`,
`pointer-lock-after.png`, `pointer-lock-after-input.png`, their Canvas crops,
`pointer-lock-console.txt`, finite runtime state, and a live WebGL context. Add
the exact root cause, changed files, commands, and screenshot hashes here once
verified.

### HM-2026-07-22-CAVERN-WHITE-CONTEXT - PATCHED, PC DEV/PRODUCTION VISUAL PASS

**Reported:** 2026-07-22, PC development runtime

**Symptom:** Wailing Caverns renders correctly, then the screen turns white.

**Expected:** The textured cavern, player, mobs, and HUD remain visible after
the first rendered frame and while pointer lock is active.

**Source diagnosis:** No application DOM layer, loading state, error state, or
Canvas clear color is white. The development root was wrapped in React
`StrictMode`; the installed R3F Canvas teardown schedules a forced WebGL
context loss after 500 ms, making the development-only effect replay a strong
timing match. No live context-loss event was captured, so this remains the
leading code-path diagnosis rather than visual proof. Two independent scene
lifecycle gaps were also confirmed: player physics could start while the
dungeon collider was still suspended, and the prior ready signal used the
renderer-global frame count rather than a post-mount dungeon frame.

**Patch:** The game root now mounts once in development. `GameScene` listens
for WebGL context loss/restoration, releases pointer lock, pauses simulation,
and displays a dark DOM recovery surface instead of leaving a blank Canvas.
Three's automatic restoration preserves the current Rapier world. If automatic
recovery fails, the explicit restart returns to the entrance and remounts at
DPR 1 with antialiasing and shadows disabled. The player and mob assets continue
preloading behind the loading surface, but the entire physics world and mob
simulation remain paused until the current dungeon/collider instance has
completed a rendered frame. This prevents a fall-through race without causing
a late GPU-upload burst at the first visible cavern frame.

**Files changed:** `src/main.tsx`, `src/scene/GameScene.tsx`,
`src/scene/DungeonEnvironment.tsx`, `src/scene/ProductionDungeon.tsx`, and
`src/styles.css`.

**Automated evidence:** 23 test files and 136 tests passed. The TypeScript/Vite
production build completed with 662 modules transformed. The entrance probe
reported finite camera geometry, a 6.099 m camera-to-target distance, and no
camera-to-target collision. The creature audit validated all 16 GLBs. The live
development server returned HTTP 200 for the page and updated scene module and
served both the recovery UI and readiness gate.

**Visual evidence status (updated 2026-07-22):** Edge/Chromium was attached at
1440x900 through Playwright. A fresh offline character entered the development
runtime and the rebuilt production preview; both showed textured Wailing
Caverns after a 12-second load/stability wait and remained rendered throughout
party-command and pause-menu checks. The post-entry captures had no `pageerror`
or console error. The original post-load white frame was not reproduced.
Native pointer lock, Android WebView, and the AYN Thor displays remain separate
unverified surfaces and must not be inferred from this pass.

### HM-2026-07-26-AVATAR-EQUIPMENT-RIGHT-DRAG - FIXED, PC VISUAL PASS

**Reported:** Playable character models and weapon/gear presentation had
stopped appearing as expected. Desktop right-mouse camera look also triggered
the browser's jarring Pointer Lock capture transition.

**Root cause:** The July 25 authentic-equipment integration left
`starterWeapons.tsx` and its tests in place but disconnected
`StarterWeaponAttachments` and `ApproximateStarterEquipment` from the shared
`CharacterModel`. As a result, valid character rigs rendered bare in the roster
and dungeon. Authentic item GLBs were also loaded inside the character's
top-level Suspense/error boundary, so one slow or broken optional gear model
could temporarily replace its entire wearer. Character-manifest loading used
one all-or-nothing `Promise.all`, allowing one optional race pack failure to
prevent every otherwise-valid race from resolving.

The camera transition was intentional browser behavior caused by
`GameScene.tsx` calling `requestPointerLock()` on every right-button press. The
browser controls that transition; it cannot be styled away.

**Fix:** The shared character renderer once again portals class-appropriate
starter weapons into the animated hand bones and supplies approximate gear on
loading/error fallbacks. Authentic client item models replace starter gear only
for their occupied hand. Each optional item model now owns an independent
Suspense and `AvatarErrorBoundary`, with the matching starter item as its local
fallback, so it cannot hide the body or unrelated gear. Character packs load
independently and merge whichever valid manifests are available.

Desktop camera look now uses bounded ordinary cursor deltas only while the
right button is held. It ends on release, lost pointer capture, blur, overlay
opening, or teardown. The normal gameplay path contains no
`requestPointerLock()` call, so no browser mouse-capture transition is invoked.
Without Pointer Lock, camera rotation naturally stops at the physical browser
window edge; the player can release, reposition the cursor, and continue.

**Regression protection:** Root `AGENTS.md` now makes the character/equipment
boundaries and popup-free right-drag behavior mandatory. New tests pin the
shared renderer integration, independent manifest failure behavior, occupied
hand replacement, per-item loading/error isolation, and unlocked right-drag
deltas. Existing asset audits still verify every playable rig, hand bone,
native equipment GLB, and texture.

**Automated evidence:** Focused avatar/input/store/respawn coverage passed 11
files and 53 tests. The complete suite passed 81 files and 504 tests. TypeScript
and Vite production build passed with 750 modules transformed. Because this
workspace is iCloud-synced, Vite now retains old hashed output instead of
recursively emptying `dist`; this avoids iCloud restoring an asset during
cleanup and causing `ENOTEMPTY`.

**Visual evidence:** The live 1440x720 PC development runtime showed the female
Human Priest body and staff in the roster, then the player plus four distinct
party character bodies and equipped silhouettes in textured Wailing Caverns.
An ordinary right-click produced no Pointer Lock prompt or scene transition,
and no application console error, avatar warning, or WebGL context loss was
captured. The one console warning was the existing third-party initialization
deprecation. Evidence:

- `playtest-artifacts/2026-07-26-model-input/roster-character-staff.png`
- `playtest-artifacts/2026-07-26-model-input/dungeon-equipped-party.png`
- `playtest-artifacts/2026-07-26-model-input/right-click-no-popup.png`

The browser-control surface can issue an atomic right-click but cannot hold the
right button while moving. Continuous drag rotation is therefore covered by
the input integration test; final feel across long drags remains a short manual
PC check.

## Entry template

```text
ID / status:
Date and platform:
Source/build identifier:
Observed symptom:
Root cause:
Files changed:
Commands and results:
Screenshots and hashes:
Console/runtime evidence:
Remaining unverified platforms:
```

## Gameplay implementation log

### HM-2026-07-21-COMBAT-PARTY-TALENTS - IN PROGRESS

**Requested:** Replace generic talent text with the local 3.3.5 talent text and
make learned talents affect gameplay; make Psychic Scream fear nearby enemies
for eight seconds; respawn a defeated player at the dungeon entrance; link
pack aggro; add one tank and three DPS companions; and play attack/death
animations for creatures.

**Authoritative local-data finding:** The extracted `Spell.dbc` contains the
talent names and descriptions. The generator was reading the standard string
field offsets, but this client uses a custom 239-field schema with six fields
inserted before the localized string blocks. The missing names were a parser
bug, not missing client assets.

**Combat-pool finding:** The current health, mana, XP, and combat-power curves
are explicitly prototype curves. Mana classes currently use a 100-point
percentage-style pool; these numbers are not exact 3.3.5 character-sheet
values. The client export was checked for base HP/MP tables, but those
server-side character-stat tables are not present in the resolved MPQ view.
Replacing the prototype pools accurately will require the matching 3.3.5
server database (class-level and race/class-level stats), plus the eventual
gear/stat model; they must not be presented as authentic yet.

**Staged and focused-tested so far:** Psychic Scream uses an eight-yard radius
and eight-second fear; hostile control starts combat; pulling any authored
`pack:member` wakes its surviving packmates; and death exits pointer lock and
resets the encounter/player position to the entrance. Full installed-project
tests, build, screenshots, and animation/party verification remain required
before this entry can be marked complete.

**Creature-animation finding:** All 16 browser creature GLBs had previously
been reduced to Stand and Walk by `prepareAnimatedCreatures.mjs`; the source
files were not generally missing. Eleven source creature exports also contain
exact `AttackUnarmed (ID 16 variation 0)` and `Death (ID 1 variation 0)` clips,
which are now restored in the staged browser assets. The five composite
humanoids (`druid-of-the-fang`, `lady-anacondra`, `lord-cobrahn`, `lord-pythas`,
and `lord-serpentis`) were themselves exported with animation IDs 0 and 4 only,
so they use bounded procedural attack and one-shot terminal death poses.

**Animation/fear staged behavior:** Attack clips resolve while an engaged mob is
in its attacking phase. Death has priority, plays once, clamps on its last
frame, preserves the corpse's combat position, and stops both mixer and proxy
root updates once settled. Active fear suppresses attack visuals and steers the
mob away from the player on the dungeon plane; stun/root still lock movement,
slows still apply, and the flee position is clamped to the mob's home leash.
The existing 95-unit population activation gate remains in place.

**Animation automated evidence:** The Three.js GLTF loader audit parsed all 16
staged GLBs and found 11 exact combat-clip models plus five explicit fallbacks.
The staging TypeScript app/node configurations passed `--noEmit`; the complete
Vitest suite passed 22 files and 131 tests using the staging-safe config loader.
This is logic/asset evidence only. Installed-project build, combat screenshots,
and live attack/death/fear observation remain required.

**Talent-data result:** The corrected custom-schema parser regenerated all 829
talent nodes and 2,243 rank spells from this client's DBC files. Visible names
and rank descriptions contain no generic fallbacks or unresolved tooltip
tokens. The talent panel shows both the learned rank and the next rank; the
SpellRadius table resolves values such as Circle of Healing's 15-yard radius.
All exact raw rank-effect metadata remains available to runtime systems.

**Talent-runtime coverage boundary:** 460 of 829 nodes currently yield at least
one common scalar modifier (damage, healing, mitigation, pools, regeneration,
cost, crit/haste, cast/cooldown, attributes, armor, range, duration, or threat).
This is partial coverage, not proof that every conditional clause or proc is
simulated. The other 369 nodes require triggered-spell, aura/proc, summon,
weapon-event, form, or active-ability systems. There are 151 one-rank
active/spell-granting rows. Faithful generic activation also needs DBC cost,
cast-time, range, and trigger payload generation plus friendly/party/dual
targeting metadata; the baseline action set now distinguishes hostile, friendly,
and self targets.

**Implemented scalar behavior:** Learned scalar modifiers are cached and now
feed max health/resource (including Stamina/Intellect), regeneration, costs,
direct and periodic damage/healing, shields, incoming mitigation, cast time,
cooldowns/groups, haste/GCD, range/AoE, and status/shield duration. Focused
tests prove concrete rank changes for Improved Frostbolt, Improved Fire Blast,
Twin Disciplines, Divinity, Convection, and Fel Vitality (21/21 passed).
Conditional/proc/stance clauses are not statefully triggered yet; crit uses a
deterministic expected-value benefit; armor is approximated; threat, flat-point
cost reductions, added periodic ticks, and learned active spells remain outside
this scalar layer.

**Party staged behavior:** A stable character-seeded roster now creates exactly
four AI companions: one Protection Warrior, Blood Death Knight, or Protection
Paladin tank and three unique non-tank DPS specs. All four attack an engaged
enemy. While alive, the tank redirects 70% of incoming health damage and can be
downed. PC frames are clickable; Thor frames are touch-selectable; D-pad
Up/Down cycles living allies. Baseline direct heals and healing-over-time spells
now route to a selected living ally, with the player as the no-selection fallback.

**Screenshot status (superseded 2026-07-22):** The first browser query exposed
no in-app backend, but an installed Edge executable was later attached through
Playwright. Current party/world evidence is recorded in the entry below. Native
pointer-lock, Android, and Thor checks remain outstanding.

**Installed-project verification:** The integrated project passed 23 test files
and 136 tests. The creature audit loaded all 16 GLBs (11 real combat-clip sets,
five procedural fallbacks). TypeScript and Vite production build passed with
662 modules transformed, and the live preview returned HTTP 200 for the new
bundle. Vite reports existing large-chunk warnings; this is a startup/bundle
optimization item, not a failed build.

**Thor/Android artifact from the preceding combat build:** `npm run android:apk`
completed after a fresh Capacitor sync and 98 Gradle tasks. Artifact:
`android/app/build/outputs/apk/debug/healer-man-0.1.0-debug.apk`, 108,315,663
bytes, SHA-256
`77B2BB53BD21FAA799F5BE15FE722399FB03B33983512BE8647694C9B5C5845B`.
This artifact predates the party-world command changes below.

### HM-2026-07-22-PARTY-WORLD-COMMANDS - IMPLEMENTED, PC VISUAL PASS

**Requested:** Make the four roster allies visible in the world; add Attack,
Defend, and Stop orders; make Attack advance toward the next boss while clearing
adds, Defend respond only to engaged enemies, and Stop hold the current world
positions; add pause-menu party management with current and future talents;
keep every ally at the player's level.

**Implementation:** Party command and boss-objective state now lives in the
low-frequency party store. Four procedural world actors keep their positions in
a render-loop registry, follow formation in Defend, freeze in Stop, approach
active targets at melee/ranged distance, and use the shipped Wailing Caverns
navigation GLB plus deterministic A* routing in Attack. A soft player tether
keeps the existing player-centric mob activation/leash system coherent; Stop is
exempt so it remains a literal hold order. Navigation loading is isolated behind
its own Suspense and error boundary, so an optional route-helper failure cannot
blank or suspend the dungeon Canvas.

**Interface:** The PC HUD exposes Attack/Defend/Stop with F1/F2/F3 shortcuts and
pressed state. The pause hub opens Party Management, where the four allies,
roles, live health/status, current legal talent ranks, and next three automatic
level choices are visible. The Thor companion snapshot and tactical panel expose
the same commands and named boss objective. World nameplates are hidden while a
menu is open, fixed at a small screen size, and stacked beneath the HUD.

**Automated evidence:** The full suite passed 28 files and 155 tests. New focused
coverage includes command acquisition/holding/range, deterministic roster level
sync, world-position registry and movement helpers, navigation welding/A*, all
22 supported specialization plans at every level 1-80, F-key mapping, and
dual-screen command dispatch. The final TypeScript/Vite build transformed 669
modules and emitted the 1,329.06 kB navigation GLB. The existing large-chunk
warning remains informational.

**Visual evidence:** At 1440x900 in the live development runtime, all four
companions and the command bar were visible over textured Wailing Caverns.
Attack moved the group into the entrance encounter and party/player health plus
mob health changed without `pageerror` or console errors. Stop appeared as the
selected HUD order, and all four members showed `HOLDING` in Party Management.
The level-1 panel showed future level-10/11/12 choices; the seeded level-10
production panel showed the current rank and level-11/12/13 choices. A first
capture exposed oversized nameplates leaking over the modal; the nameplates were
then reduced, placed below the HUD, hidden while paused, and the corrected dev
and production panels were recaptured cleanly. Fresh production entry remained
textured after 12 seconds with a Canvas present and zero post-entry console or
page errors.

**Remaining verification boundary:** The entrance encounter and menu flow were
visually exercised, but a complete manual walk through every boss route was not.
Late-dungeon navigation islands, native pointer lock, Android/Thor rendering,
and the full boss-by-boss Attack traversal remain manual follow-up coverage.

### HM-2026-07-22-FRIENDLY-TARGETING - IMPLEMENTED, PC INTERACTION PASS

**Requested:** Let the player mouse-click party members in the world or party
frames and use healing/support abilities on the selected teammate.

**Implementation:** Living world companions now have generous transparent
capsule hit targets. Their click handler stops both R3F and native DOM bubbling,
so selecting an ally does not also capture the mouse. Hostile and friendly
selection are mutually exclusive across world clicks, keyboard/controller
cycling, clear-target input, and the Thor companion channel. Party selection is
idempotent and clears automatically if the member is downed or removed.

Holy Light, Flash Heal, Healing Wave, and Rejuvenation now use 40-yard friendly
targeting. With no ally selected they intentionally target the player. Casts
snapshot the chosen recipient, revalidate that member's life, runtime presence,
and range on completion, and never put party IDs into hostile-target state.
Direct heals and per-recipient HoTs update party health without changing the
self-healing behavior of hostile hybrid attacks such as Death Strike.

**Interface:** A selected teammate gets the existing gold world/party-frame
selection treatment plus a green target frame with live health. Friendly action
slots carry an `ALLY` badge and accessible target wording.

**Evidence:** At 1440x900 in Edge/Playwright, clicking Garrick's party frame
selected him and left pointer lock off. Flash Heal displayed its cast bar,
restored Garrick from 157/217 to 217/217, and spent mana. Clicking Ash's actual
3D body switched both the target frame and selected party frame to Ash while
pointer lock remained off. The textured dungeon Canvas and all four companions
remained visible; no page error or white-screen regression occurred. The full
suite passed 28 files and 169 tests with a 15-second allowance for the existing
performance-heavy talent-plan audit. TypeScript and the production Vite build
passed with 669 modules transformed.

### HM-2026-07-22-PARTY-AVATARS - IMPLEMENTED, PC VISUAL PASS

**Requested:** Replace the four lightweight companion pawns with actual
characters that have races, genders, and randomized appearances.

**Implementation:** Every generated party member now owns a flat, serializable
`raceId`, `gender`, and five-value appearance record. Identity uses a dedicated
per-member PRNG stream keyed by the persistent player-character ID, party slot,
class, and specialization. It is varied between player characters but stable
across rerenders, level synchronization, combat resets, and re-entry. The race
selection respects the project's packed race/class catalog.

Party actors now use the same `CharacterModel` pipeline as the player: shipped
race/gender GLBs, skeleton-safe clones, composed skin/face/hair/feature textures,
Death Knight eye treatment, and Stand/Run animation. Each ally retains its own
Suspense/error boundary and procedural fallback, so one bad avatar cannot blank
the dungeon Canvas. Measured metrics for all 20 race/gender rigs drive each
nameplate, target ring, and transparent click capsule; this covers the 1.17 m
female Gnome through the 3.19 m male Troll. Party frames, target UI, pause-menu
management, and the Thor snapshot now expose race/gender identity.

**Asset evidence:** A direct loader audit covered all 20 character GLBs and
5,072 referenced model/texture URLs with zero missing files. Every rig had
finite geometry and animation tracks, matching appearance counts, required
material tokens, and valid Stand/Walk/Run clips.

**Visual evidence:** At 1440x900 in Edge/Playwright, the initial seeded party
rendered distinct Draenei, Night Elf, and Dwarf characters with visible body,
skin, hair, and feature differences. A second deterministic extreme party
rendered a male Troll, male Draenei, male Human, and female Gnome together.
There were no avatar loader warnings, page errors, WebGL context-loss reports,
or white-screen recurrence. Clicking the real Troll mesh selected Seraphin,
updated the green friendly target frame, and left pointer lock off. Party
Management displayed all four race/gender identities and preserved talent data.

**Automated evidence:** TypeScript passed. The full suite passed 28 files and
171 tests, including deterministic identity, valid appearance ranges, reset
stability, combat routing, and dual-screen propagation.

**Remaining boundary:** The shipped playable rigs currently provide the base
body/underwear, facial customization, and locomotion clips, but not equipped
armor/weapons or skeletal attack/death clips. The existing transform-based
attack/down motion remains. Android/Thor GPU performance with five animated
character rigs remains a separate native-device verification pass.

### HM-2026-07-22-CONTROLLER-ACTION-BINDINGS - IMPLEMENTED, AUTOMATED/NATIVE PASS

**Requested:** Replace the four-slot action bar with a complete controller
layout: Triangle/Circle/Cross/Square in their physical diamond, R1/R2, D-pad
Left/Right, L1 and L2 modifier layers, and a player-configurable binding menu.

**Implementation:** The combat runtime now owns 24 stable action bindings:
eight physical controls across Base, L1, and L2. Browser gamepads, the Android
controller bridge, and keyboard 1-8 with Shift/Alt all resolve through the same
semantic binding model. Modifier press/release ordering prevents a fast
L1/L2 transition from casting on a stale layer; native triggers use fallback
axes plus press/release hysteresis. Android system Back is distinct from Circle,
so it pauses gameplay or backs out of menus without firing a Circle ability.

**Interface:** Only the active eight-slot layer is shown over the playfield.
The four face buttons form the controller diamond, with separate shoulder and
horizontal D-pad groups. The pause menu's Ability Controls panel exposes the
same diagram, spatial D-pad navigation, current-assignment focus, empty or
duplicate bindings, safe reset confirmations, and immediate per-character
persistence. Friendly abilities continue to target the selected ally or self;
D-pad Up/Down cycles living party members plus the player. The PC and Thor
companion surfaces share the active layer and semantic binding commands.

**Automated/native evidence:** The full suite passed 33 files and 212 tests.
Coverage includes all 24 defaults, partial/corrupt save normalization,
per-character isolation, custom chord casting, modifier transition ordering,
native duplicate suppression, system/menu Back routing, self targeting, and
dual-screen dispatch. TypeScript and the production Vite build passed with 672
modules transformed. Android `:app:compileDebugJavaWithJavac --no-daemon`
completed all 43 tasks successfully; only the existing SDK XML/flatDir warnings
remained. A fresh production preview on an isolated local port returned HTTP
200 for both `/` and `/?layout=thor-preview`, then the preview process was
stopped.

**Current content boundary:** Every class currently exposes four runtime
abilities. Defaults therefore fill Cross, Square, Triangle, and Circle while
R1, R2, D-pad Left/Right, L1, and L2 slots begin empty. Players can reuse those
four abilities anywhere now; distinct assignments for all 24 slots require the
separate future expansion of the class ability catalogs and unlock progression.
No screenshot-based visual pass was possible in this session because no
connected in-app browser backend was available; responsive and accessibility
layout received static review, and fresh production entry was HTTP-smoke-tested.

### HM-2026-07-22-INPUT-HANDEDNESS - FIXED, AUTOMATED PASS

**Reported:** Keyboard A/D strafing and horizontal pointer-lock mouse look moved
in the opposite direction from the rendered third-person camera.

**Cause and fix:** Raw keyboard, mouse, and gamepad inputs already used standard
positive-right values. The shared +Z camera-relative transform supplied the
opposite screen-right vector, and positive horizontal look was added to yaw
even though this camera convention turns right by decreasing yaw. The movement
basis now uses the rendered camera's actual right vector, and a shared yaw
helper applies positive mouse/right-stick look in the matching direction.
Vertical look is unchanged. The correction also fixes left-stick strafing and
right-stick horizontal look.

**Evidence:** Focused input coverage passed 15 tests. New regressions verify
A/D at zero and nonzero camera yaw plus both horizontal look directions. The
full suite passed 33 files and 214 tests, and the production build passed with
672 modules transformed. A manual pointer-lock feel check remains device-side
because no connected browser backend was available in this session.

### HM-2026-07-22-SPELL-ICONS-STARTER-WEAPONS - IMPLEMENTED, PC VISUAL PASS

**Requested:** Use each spell's real client icon, equip the player and every
party member with a starting weapon, and replace the companions' repeated
whole-body thrusting with an attack that reads as an actual strike.

**Spell-icon result:** All 40 runtime abilities were joined through their
`dbcSpellId` to the local 3.3.5.12340 `Spell.dbc` and `SpellIcon.dbc` data. The
existing basenames were exact, including several non-obvious stock mappings,
but every URL pointed at a nonexistent `/assets/ui/spells/*.webp` directory.
The 39 distinct authoritative BLPs are now decoded into shipped 64x64 PNGs,
and the catalog serves those files. `scripts/export-spell-icons.mjs` provides a
dependency-free reproducible exporter for DXT1, DXT3, and DXT5 BLP2 sources.
The catalog audit pins all 40 DBC-id mappings and fails when an icon is absent
or empty.

**Equipment result:** No standalone weapon models or playable-character attack
clips were present in the browser asset pack; every race/gender rig contains
only Stand, Walk, and Run. A shared procedural equipment layer now supplies
swords, maces, staves, daggers, bows, and shields. Warriors use sword/shield,
Paladins and Shamans use mace/shield, Rogues dual-wield daggers, Hunters use a
bow, Death Knights use a runeblade, and cloth/Druid classes use staves. Weapon
scale is estimated in avatar-local space. Attachment discovers hands from the
named finger pivots instead of numeric bone IDs, which differ by skeleton; the
same class loadout is also rendered on the procedural loading/error fallback.

**Attack-motion result:** Party `nextActionAt` changes now create one stable
presentation pulse per real hit. Tanks no longer animate continuously merely
because their status is `guarding`, and the party wrapper no longer pitches or
bobs during attacks. Each pulse layers a short windup/strike/recovery onto the
right shoulder after the locomotion mixer, while the equipped weapon layer
adds bounded main-hand/off-hand or bow motion and returns exactly to rest. The
player receives the same one-shot presentation after a successful completed
ability. Combat simulation timing remains outside the render components.

**Automated evidence:** The focused icon/equipment/attack/party suites passed
4 files and 25 tests. The equipment audit reconstructs every one of the 20
shipped playable GLBs and resolves both hands. The complete suite passed 35
files and 224 tests with the established 15-second allowance for the heavy
level-1-through-80 party-talent audit. TypeScript passed. A clean Vite
production build transformed 674 modules; the existing large-chunk warnings
remain informational.

**Visual evidence:** At 1440x900 in Edge/Chromium, the roster preview showed
the Priest's staff attached to her hand. Fresh dungeon entry showed textured
Wailing Caverns, equipped player/party silhouettes, and the real Smite, Shadow
Word: Pain, Flash Heal, and Psychic Scream art on the controller action bar.
The successful captures are
`playtest-artifacts/2026-07-22-weapons/roster-staff.png` and
`playtest-artifacts/2026-07-22-weapons/dungeon-equipped-party.png`; neither
shows a loading/error surface or the prior white-screen failure.

The follow-up 1280x720 Edge/Playwright pass issued the real F1 Attack shortcut
and captured 25 frames from the pre-command state through two defeated entrance
adds. The sequence shows discrete arm/weapon actions on damage events, upright
rest poses between pulses, and no repeated pelvis pitch. Weapons remain bound
to moving hands throughout the advance. Evidence is in
`playtest-artifacts/2026-07-22-weapons/attack-contact-sheet.jpg` and the
individual `attack-frame-*.jpg` files. No page error, scene-error overlay,
white-screen recurrence, or WebGL context loss appeared. The log contains one
generic 404 plus software-WebGL and initialization deprecation warnings; none
prevented the dungeon, party, spell art, equipment, or combat from rendering.

**Remaining visual boundary:** A tank can briefly cross very close to the
player camera while the party leaves formation, making its shield dominate the
view for several frames. That is a companion separation/formation issue, not a
return of the old attack thrust. Native pointer lock, Android/Thor scale and
performance, and every race/class equipment silhouette remain separate
device-side checks. Re-exported authentic item models and native attack/cast
clips would be a future visual-quality upgrade over the current procedural
starter gear.

### HM-2026-07-22-MAIN-MENU-DUNGEON-BROWSER - IMPLEMENTED, AUTOMATED PASS

**Requested:** Add a main menu after character selection with Dungeons,
Manastorms, Roguelike RPG, Trials, Options, and Player Profile. Only Dungeons
needs to work initially and must open a dungeon list.

**Result:** Confirming a roster character now opens a dedicated mode hub
without initializing or resetting gameplay. Dungeons opens a separate browser
backed by a manifest-derived catalog. Wailing Caverns is the sole available
entry because it is the only complete runtime currently shipped; selecting it
activates the character, resets at the entrance, and mounts the 3D game. The
other five requested mode cards are visible, disabled, and marked Coming soon.
Back navigation follows dungeon browser to main menu to character roster, and
the in-game pause exit now returns to the main menu. Narrow PC layouts retain a
visible Continue button instead of hiding the only pointer-accessible action.

**Controller and companion behavior:** A/Enter confirms, directional input
navigates, and B/Escape backs out on both new top-screen views. The companion
display has explicit main-menu and dungeon states, so these shell phases never
fall through to a stale tactical HUD. Its roster Continue, dungeon launch,
change-character, and return-to-main-menu commands mirror the authoritative
top-screen transitions.

**Evidence:** Focused shell/catalog/dual-display coverage passed 3 files and 16
tests. The complete suite passed 36 files and 232 tests. TypeScript and the
production Vite build passed with 677 modules transformed; existing large
chunk warnings remain informational. The local development entry returned HTTP
200. Screenshot automation could not connect to the in-app browser backend in
this Windows sandbox, so responsive desktop/compact layouts received static
review rather than a claimed visual pass.

### HM-2026-08-01-RUNEWAKER-ALL-DUNGEONS - ACTIVE FIDELITY PASS

**Reported:** Every RuneWaker dungeon unexpectedly displayed Coming Soon, and
the imported actors did not consistently retain their original skeletons,
animations, appearances, or attacks.

**Catalog result:** The generated WoW availability refresh had replaced the
whole availability document and removed every RuneWaker key. Availability
generation now preserves non-campaign keys, and packaged RuneWaker definitions
fall back to their validated offline metadata when a key is absent. The live
dungeon browser showed all 30 RuneWaker packages as Ready. Evidence:
`artifacts/playtests/rom-dungeon-catalog-ready.png`.

**Native animation result:** The RuneWaker exporter now selects the largest
body GR2 hierarchy instead of a one-bone effects hierarchy, samples the ACT/RAS
motion set at 30 Hz, and maps skin weights from original OBJ positions to the
vertices Blender reordered or duplicated at seams. A complete audit accounts
for every recipe-referenced actor and rejects invalid GLBs or missing required
animation families. Dynamic paperdoll appearances are exported as exact
image-specific assembled actors instead of generic proxy bodies; the portable
appearance snapshot is checked into the build inputs so normal/offline builds
never connect to SQL.

For this historical rollout, the contract was expressed at the base-actor level
as 267 native actors, one documented pose-only actor, and no combat proxy actors.
The newer authoritative acceptance scope is the 322 namespaced shipping
variants recorded in the in-progress entry above.
Runtime population bindings were 435 native,
one pose-only, and zero proxy. The former eight actorless “bosses” were source
FX/door controllers (six invisible soul-ball effects and two dungeon doors),
not missing creature models; they are now deferred mechanics objects rather
than fake combat objectives.

**Combat-data result:** The portable RuneWaker spell catalog contains the
source assignments used by all dungeon population recipes. The generator only
executes positively classified direct-HP-damage families; buffs, summons,
heals, hooks, and unknown effects remain evidence rather than guessed gameplay.
The current cross-dungeon audit counts 388 source-backed executable attacks and
zero evidence-only spells entering runtime combat. Every entity still retains
a HealerMan basic fallback, so this is not yet a claim of formula-exact boss
state machines.

**Browser result:** Forsaken Abbey loaded from packaged files, rendered its
native population, and entered party combat with damage and XP progression;
captures are `artifacts/playtests/forsaken-abbey-native-population.png` and
`artifacts/playtests/forsaken-abbey-combat.png`. Pasper's Shrine initially
stopped at Blackhorn Nisorn because the source position was 5.25 meters outside
the retained navigation surface. Objective routing now projects a combat
approach point within the same six-meter rule used by the audit. The party
advanced, fought, took damage, and gained XP without console errors. Evidence:
`artifacts/playtests/paspers-shrine-route-fixed-combat.png`; the original
failure is retained as `artifacts/playtests/paspers-shrine-route-blocked.png`.

**Open fidelity boundary:** The source audit still reports disconnected ordered
boss routes in several packages and evidence for doors, phases, summons,
patrols, timers, interactions, and loot that has not been translated into
tested encounter state machines. Those items must remain explicit blockers or
review work; they must not be represented as a complete original game until
implemented and playtested instance by instance.

**Final production/offline verification:** The production build completed with
all 93 packaged paperdoll actor copies present in `dist`. The temporary
`HealerMan_RW_Objects_Paperdoll` forensic database was dropped and the manual
`MSSQL$HEALERMAN_RW` service was stopped before the browser check. From the
packaged production preview, offline login opened the RoM catalog with 30
available, 30 Ready, and zero Coming Soon entries; Forsaken Abbey then launched
into the 3D runtime with its population present. Evidence:
`artifacts/playtests/rom-dungeon-catalog-production-final.png` and
`artifacts/playtests/forsaken-abbey-production-final.png`. No browser error was
reported during dungeon launch. The log retained two earlier WebGL context-lost
messages from page transitions and one non-blocking initialization deprecation
warning.

**Historical automated scope:** TypeScript and the production build passed for
that rollout. Its RuneWaker animation audit reported 267 native base actors, one
pose-only base actor, and no configured combat proxies; required idle, movement,
attack, wound, and death animation-family gates were satisfied under the earlier
gate. It is superseded by the completed v5 all-variant verdict above. The
gameplay audit covered 30
dungeons, 8,199 source placement rows, 4,086 runtime static spawns, 436 runtime
entities, 61 boss objectives, and 824 executable attacks (436 basic fallbacks
plus 388 source-backed attacks). The full test run passed 87 files and 522
tests; its remaining three failures are pre-existing Manastorm fixture/parser
failures outside this RuneWaker work. The previously stale Forsaken Abbey test
was fixed and passes separately.

**Remaining source-fidelity work:** Thirty-one paperdoll appearances still need
their original dye-mask colors composited, although their geometry, base
textures, skeletons, skinning, and clips are authentic. Twenty-one optional
source RAS references are absent or non-GR2, while every required combat family
still passes. The navigation audit records 19 unreachable distance-derived
objective legs across eight packages; there are no preserved source-authored
orders, so unsafe permanent off-mesh links were not invented. Original
coefficient formulas and the evidenced door, phase, summon, patrol, timer,
interaction, loot, and modular-room state machines remain the next per-instance
implementation pass.

### HM-2026-08-01-ARROW-CAMERA-LOOK - PC KEYBOARD CAMERA PASS

**Requested:** On PC in the browser, make the arrow keys turn the camera when
playing with mouse and keyboard instead of a controller.

**Result:** WASD remains the keyboard movement cluster. Left/right arrows now
apply continuous camera yaw, and up/down arrows apply continuous camera pitch.
The pause controls and live HUD hint now describe arrows as camera input rather
than movement input. Controller left-stick movement and right-stick camera look
remain unchanged.

**Live browser evidence:** Wailing Caverns was launched from the offline roster.
Repeated right-arrow input changed the minimap heading from 244 degrees to 242
degrees; repeated left-arrow input then changed it to 247 degrees. Up/down arrow
passes visibly changed the vertical camera angle. The exploration counter stayed
at 0 m throughout, confirming arrow input did not move the player, and the
heading stayed at 247 degrees after a 600 ms release check. The updated camera
hint rendered in the dungeon and the browser console contained no errors.

**Pointer-lock boundary:** The in-app playtest browser does not expose the
Pointer Lock API, so `document.pointerLockElement` is undefined there and a
live null assertion or held-right-button automation is not possible on that
surface. No capture prompt appeared. The unchanged right-drag path remains
covered by `inputManager.test.ts`, including ordinary drag deltas and a null
pointer-lock assertion.

**Automated evidence:** `src/game/inputManager.test.ts` and
`src/game/inputMath.test.ts` pass 19 tests. TypeScript and the production Vite
build pass with 914 modules transformed; the existing large-chunk warnings
remain informational.

### HM-2026-08-01-RUNEWAKER-PLAYERS - RACES, CLASSES, AND DUAL CLASS

**Result:** The WoW/RoM content split is live across character creation,
persistence, roster summaries, party generation, Class Hall, the spellbook,
and combat. RoM ships Human, Elf, and male-only Dwarf plus the ten released
classes. Level 10 permanently fixes a race-eligible secondary; Class Hall swaps
the ordered primary/secondary orientation and its independent action bar.
Shared XP, automatic ranks, primary stats plus 10% secondary stats, role and
resource changes, equipment compatibility, and directional elite unlocks are
covered by the application tests. Existing WoW profiles migrate to storage v2
without losing their progression or bindings.

**Generator/audit:** `npm run runewaker:players:refresh` completed from the
RuneWaker source tables, strings, elite Lua, image metadata, and `interface.fdb`.
The portable catalog records source SHA-256 hashes and contains 269 ordinary
skill families plus 660 elite milestones across all 66 valid ordered pairs.
All 929 visible skills have an executable or explicit non-combat contract;
zero are unresolved. All 929 icon references resolve to 918 unique native PNGs.
The independently loadable `runewaker-v1` manifest contains five native rigs
(Human male/female, Elf male/female, Dwarf male), starter attachments, and
native idle/walk/run/attack/cast clip families. Khronos validation reported
zero errors and zero warnings for all five GLBs.

**Live controller/browser QA:** From the production preview, keyboard/controller
menu navigation moved from the default WoW Human race to the RoM category and
confirmed it with Enter. The draft reset to Taborean Human and exposed the
correct six-class pool. Dwarf creation enforced male gender and exposed
Warrior, Rogue, Mage, Priest, Warlock, and Champion. A native Dwarf Champion
rendered in creator and roster. The level-1 Class Hall showed its level-10 lock;
level-10 selection, irreversible confirmation, swapping, and both orientations
are covered by automated controller/persistence tests.

The same RoM character could select both the 54-dungeon WoW category and the
30-dungeon RoM category. Wailing Caverns and Forsaken Abbey both launched with
the native character, a rage pool, RuneWaker spell icons, and a same-category
RoM party. The tactical menu now says **Skills**, and its RoM view reported
Primary `2 / 18`, Secondary General `0 / 0`, Elite `0 / 0`, and automatic
rank progression for the pre-secondary level-1 Champion. Browser diagnostics
contained no errors. Expected renderer teardown messages and one non-blocking
initialization deprecation warning remain.

**Evidence:**

- `playtest-artifacts/2026-08-01-rom-dual-class/rom-roster-preview.png`
- `playtest-artifacts/2026-08-01-rom-dual-class/rom-character-wow-dungeon-preview.png`
- `playtest-artifacts/2026-08-01-rom-dual-class/rom-character-rom-dungeon-preview.png`
- `scripts/runewaker-pipeline/snapshots/runewaker-player-skill-audit.json`

**Final verification:** `npx tsc -b --pretty false` passed; Vitest passed all
95 files and all 549 tests; `npm run build` passed with 921 modules transformed.
Vite's existing large-chunk advisory remains informational.

### HM-2026-08-01-DUNGEON-ASSET-EVICTION - GLTF AND GPU LIFECYCLE

**Result:** Regular dungeon sessions now own their environment, collision,
navigation, and creature GLB URLs. Switching dungeons or leaving the gameplay
runtime clears those URLs from Drei's global `useGLTF` cache. Scene teardown
also deterministically disposes dungeon-owned geometries, cloned materials,
textures, creature skeletons, and bone textures, with identity-based
deduplication so shared resources are released once. Playable-character and
equipment assets remain cached because they are reused across every dungeon.

**Live browser evidence:** A fresh production-preview renderer loaded Auchenai
Crypts at 797.7 MB private / 856.6 MB working set. Returning to the main menu
released the outgoing scene and settled at 555.8 MB / 615.1 MB. Blackrock
Depths then loaded at 1,033.2 MB / 1,098.6 MB. Chromium initially retained that
allocator reservation after the Blackrock teardown, but loading Auchenai again
settled at 800.0 MB / 855.6 MB, effectively matching the first Auchenai load
instead of stacking both dungeon footprints. Leaving Auchenai a second time
settled at 592.5 MB / 651.5 MB. The environment, five-character party, HUD, and
minimap rendered correctly throughout. Browser diagnostics contained no GLB,
WebGL, or runtime errors; only the existing non-blocking initialization
deprecation warning remained.

**Verification:** The focused dungeon ownership, session-switch, invalid-switch,
and shared-resource disposal tests passed all 11 cases. The full Vitest run
passed 98 files and 556 tests. TypeScript, the 54-dungeon campaign audit, the
five-player import verifier, and `npm run build` all passed; the production
build transformed 923 modules. Vite's existing large-chunk advisory remains
informational.

### HM-2026-08-01-ROM-ABILITY-PRESENTATION - NATIVE ICONS AND CLIPS

**Result:** Runtime RoM abilities now consume the native icon path already
recorded in `romPlayerCatalog.generated.json` instead of replacing every skill
icon with its class emblem. RoM combat animation events now select native
RuneWaker character clips for casting/continuous casting, friendly buffs, bow
channel/release, melee weapon variations and specials, wounds, and death. WoW
characters retain the existing AnimationData-ID selector.

**Live browser evidence:** The saved Dwarf Champion rendered in the roster and
entered Forsaken Abbey with its body and starter weapon silhouette visible. All
eight live action slots resolved to distinct `/assets/ui/rom-spells/<id>.png`
assets (including Heavy Bash `589471.png` and Electrocution `589472.png`). The
ability control was exercised against a selected Ooze; the game correctly
reported that the target was out of melee range. Browser diagnostics contained
no errors. Native clip reachability for release/cast/bow/melee/wound/death is
therefore recorded by the focused automated selector test rather than claimed
from the out-of-range screenshot.

**Evidence:**

- `artifacts/playtest/rom-roster-preview-2026-08-01.png`
- `artifacts/playtest/rom-dungeon-heavy-bash-2026-08-01.png`

**Verification:** `src/game/combatAnimation.test.ts` and
`src/game/romPlayerCatalog.test.ts` passed all 11 tests; `npm test -- --run
src/avatar` passed all 28 avatar tests; `npm run build` passed with 921 modules
transformed. Vite's existing large-chunk advisory remains informational.

### HM-2026-08-01-WOW-CREATURE-ANIMATION-MEMORY - RUNTIME CLIP POLICY

**Result:** Shipped WoW dungeon creature GLBs now contain only animation
families requested by `MobPopulation`: stand, walk, run, and the highest-priority
available attack, cast, wound, and death families. Every variation in each
selected combat family is retained. The complete source GLBs remain untouched
under `HealerMan-Storage/pipeline-work/dungeons`, and all five-player, Epoch,
and Wailing Caverns packaging paths now apply the same policy to future exports.

The migration optimized 1,111 current dungeon creature GLBs. Source inventory
fell from 73,191 clips to 8,102 runtime clips. Runtime payload fell from
4,855,010,592 bytes to 2,963,934,636 bytes, saving 1,891,075,956 bytes (39%).
The 146-clip Auchenai reference model retained nine gameplay clips and fell
from 13,824,072 bytes to 7,650,004 bytes. Generated manifests record the source
clip count, runtime count, policy version, checksum, and post-policy semantics.

**Live browser evidence:** A fresh production-preview session opened the saved
Human Priest and entered Auchenai Crypts as healer. The environment and all five
character rigs rendered correctly. After moving inside the 95 m population
activation radius, the browser resource inventory confirmed optimized creature
displays 17932, 17924, and 17942 were loaded. Browser diagnostics contained no
errors; the existing initialization deprecation warning remains non-blocking.
After creature activation, the renderer settled at about 930 MB private / 989
MB working set and peaked at about 1.03 GB private / 1.09 GB working set. The
earlier pre-change two-dungeon pass peaked around 1.15 GB private / 1.2 GB
working set, so these measurements are directional rather than an identical
scene sequence.

**Verification:** The policy's four focused Node tests passed. The campaign-wide
WoW content test passed all 54 dungeons, the five-player verifier remained green
with only its three documented source blockers, mob/combat animation tests
passed all 18 cases, and `npm run build` passed with 921 modules transformed.
Vite's existing large-chunk advisory remains informational.
# 2026-08-14 — local-first content delivery and compact Android client

- Build under test: fresh full production `dist` served by `server/server.mjs`
  at `http://127.0.0.1:4173`.
- Booted to the login screen and entered through **Play offline** without an
  account-server request.
- Roster preview: the saved level 8 human female priest rendered her complete
  body, hair, and starter staff silhouette. The procedural/error fallback did
  not replace or hide the character.
- Downloads screen: the two-column local-first layout rendered without
  clipping at the test viewport. Its local catalog request returned the
  expected 404 because no generated `content/` volume was mounted for this
  local smoke test; the screen retained the offline starter-content message.
- Dungeon smoke: entered Wailing Caverns as healer. The production cavern,
  player body/equipment silhouette, four companion bodies, HUD, minimap, and
  action bar were visible. No missing-GLB proxy or scene error appeared.
- Evidence reviewed during the browser run: login, roster-preview, Downloads,
  and in-dungeon screenshots. Browser console contained no errors; one existing
  third-party initialization deprecation warning was present.

# 2026-08-16 — CoA character-creation category

- Opened the offline roster and entered character creation in the live Vite
  client. The header rendered distinct **WoW**, **RoM**, and **CoA** tabs without
  clipping.
- The WoW tab showed only the eight standard WoW classes available to Human;
  no custom CoA class appeared.
- The CoA tab reused the Azeroth race roster and showed the 16 custom CoA
  classes available to Human, defaulting to Barbarian. The character body and
  starter weapon silhouette remained visible while switching categories.
- In the production preview, the saved Human Priest rendered in the roster with
  her full body and starter staff silhouette, then entered Wailing Caverns with
  her body, staff, party, dungeon environment, and HUD visible.
- Browser diagnostics contained no errors.
- Evidence: `playtest-artifacts/character-create-wow-category-2026-08-16.png`
  `playtest-artifacts/character-create-coa-category-2026-08-16.png`,
  `playtest-artifacts/character-category-roster-2026-08-16.png`, and
  `playtest-artifacts/character-category-dungeon-2026-08-16.png`.
- Verification: 71 focused catalog/store/repository/dungeon/ability/party tests
  passed, all 28 avatar tests passed, and `npm run build` completed
  successfully.

# 2026-08-16 — Deadmines party terrain following and vertical animation tolerance

- Entered The Deadmines as the saved Human Priest and selected **Attack**.
  The player crossed 43 m of uneven mine-floor and rail terrain; the tank and
  a damage companion remained visibly beside the player instead of waiting for
  the 12 m stuck-recall threshold. All four companion world labels remained in
  the live scene, and no automatic party recall was observed.
- The character maintained ordinary ground locomotion over small collision
  height changes. Focused timing coverage confirms that terrain gaps shorter
  than 100 ms neither enter the falling state nor trigger a landing animation,
  while intentional jumps and sustained falls still present immediately/after
  the grace period and land once.
- The saved roster preview rendered the complete Human body and starter staff;
  the in-dungeon view rendered the player silhouette, nearby companions, the
  Deadmines environment, and HUD.
- Browser diagnostics contained no errors. The existing third-party
  initialization deprecation warning remained non-blocking.
- Evidence: `playtest-artifacts/party-terrain-roster-2026-08-16.png`,
  `playtest-artifacts/deadmines-party-terrain-2026-08-16.png`, and
  `playtest-artifacts/deadmines-party-follow-view-2026-08-16.png`.
- Verification: 38 focused party/pathing/jump/animation tests passed, all 28
  avatar tests passed, and `npm run build` completed successfully with 954
  modules transformed. The existing large-chunk advisory remains informational.

# 2026-08-20 — shared online dungeon session with AI fill

- Ran two signed-in production clients against the local account server on
  separate browser origins. **SharedLeader / Bulwarka** invited
  **SharedMember / Mendara**; the accepted group roster showed both selected
  characters and their Tank and Healer roles as ready.
- The leader launched Wailing Caverns with **Fill open slots with AI** enabled.
  Both clients automatically entered the same activity, each rendered the
  other player through the normal equipped `CharacterModel` path, and each
  showed a live, targetable remote party frame. Three AI companions filled the
  remaining slots, producing the expected five-member party on both clients.
- The reciprocal scene captures showed Bulwarka and Mendara at the same shared
  dungeon position with matching environment, objective, AI roster, mob state,
  and remote role labels. The leader client continued encounter simulation;
  the member client received the synchronized mob and AI state without running
  a second competing world simulation.
- From the member client, selected Bulwarka's online party frame and completed
  a Lesser Heal cast. The member's mana/resource change propagated to the
  leader client and both clients retained the same live remote health state.
  Enemy combat and AI resource/health changes also continued to propagate while
  the two clients remained connected.
- Evidence reviewed during the browser run: invitation/group roster DOM,
  coordinated-launch DOM from both clients, reciprocal full-window dungeon
  captures, and before/cast-complete remote-healing DOM snapshots. No missing
  character-body or equipment fallback appeared in either dungeon capture.
- Verification: server/API tests passed all 11 cases, the complete Vitest suite
  passed all 755 cases across 132 files (including the 28 avatar cases), and the
  TypeScript production build completed successfully with 983 modules
  transformed. The existing large-chunk advisory remains informational.

# 2026-08-28 — configurable render caps and idle-loop efficiency

- Opened the saved Human Priest roster after a fresh client reload. The full
  body and starter staff rendered correctly in the single character-preview
  canvas, which reported the persisted 30 FPS preview cap and manual frame loop.
- Entered Wailing Caverns as healer and exercised the gameplay setting live.
  The performance probe settled at about 30.5 FPS for the 30 FPS option and 60
  FPS for the 60 FPS option; combat and fixed-step physics continued normally.
- Pausing switched the dungeon canvas to demand rendering. Its performance
  counters remained byte-for-byte unchanged over 1.4 seconds, confirming that
  the local paused scene did not keep rendering in the background.
- Returned gameplay rendering to the default 60 FPS and left the character
  preview at 30 FPS. Evidence reviewed during the browser run included the
  roster-preview and live dungeon screenshots. The final fresh session emitted
  no errors; the existing third-party initialization deprecation warning
  remained non-blocking.
- Verification: the focused efficiency tests passed all 83 cases, all 28 avatar
  tests passed, the complete Vitest suite passed all 809 cases across 148 files,
  and the production build completed with 1,043 modules transformed. KTX2 and
  environment-performance audits remained green; the existing large-chunk
  advisory remains informational.

# 2026-08-30 — authored weapon grips and ready stances

- Reloaded the offline client and switched a Human female preview directly
  between Hunter, Warrior, and Priest. The bow, sword/shield, and staff changed
  immediately to their native HoldBow, Ready1H, and Ready2HL poses; no stale
  class pose or backward procedural axis remained.
- Verified RuneWaker Human Warrior and Scout previews against their explicit
  hand/shield locator frames. The one-hand, shield, and bow silhouettes used the
  RuneWaker socket-axis adapter and remained proportionate to the character.
  Also checked the Ascension Human Ranger fallback with its bow visible in a
  supported ready pose.
- The saved Human Priest roster preview showed the complete body and staff with
  both hands aligned to the shaft. In Wailing Caverns, the player and party
  equipment remained attached through live combat animation without a second
  local weapon swing or load-time placement jump.
- Evidence: `playtest-artifacts/2026-08-30-weapon-grip/weapon-grip-roster.png`,
  `human-hunter-holdbow.png`, `human-warrior-ready1h.png`,
  `human-priest-ready2hl.png`, `runewaker-warrior-sockets.png`,
  `runewaker-scout-bow.png`, `ascension-ranger-bow.png`, and
  `weapon-grip-dungeon.png`.
- Verification: all 41 avatar tests passed, the 37-test focused weapon/stance
  suite passed, and the production build completed with 1,047 modules
  transformed. Asset audits remained green with zero KTX2 fallbacks and zero
  environment-performance regressions; the existing large-chunk advisory
  remains informational.

# 2026-08-30 — bounded combat VFX bursts and silent audio default

- Created a fresh offline Human Priest and confirmed the in-game Options panel
  opened with **Master volume** at 0%. Gameplay and preview frame-rate controls,
  VFX quality, and per-actor VFX/audio controls remained available.
- Entered Wailing Caverns as healer, sent the party into combat, and captured a
  live four-effect spell burst. High quality reported four active effect systems
  and zero combat point lights; additive particles, impacts, melee arcs, the HUD,
  and combat telegraphs remained visible.
- The captured burst reported 60 FPS, a 16.7 ms average frame, a 16.8 ms slowest
  frame, 258 draw calls, and 353,526 triangles at 1.5 pixel ratio. A three-second
  combat sample reached five concurrent effects, stayed between 54.1 and 60 FPS,
  and returned to a 16.7 ms frame average.
- A cold Vite dynamic import failed once while first opening character creation;
  reloading after compilation resolved it. The final dungeon run had no further
  errors; the existing third-party initialization deprecation warning remained.
- Verification: the complete Vitest suite passed all 839 cases across 153 files,
  the production build completed with 1,047 modules transformed, KTX2 and
  environment-performance audits remained green, and the release-candidate
  check passed. The existing large-chunk advisory remains informational.

# 2026-08-31 — Shadowfang Keep navigation freeze regression

- Reproduced the reported freeze with the saved Human Priest at the Shadowfang
  Keep entrance. The visible scene fell to roughly 4–11 FPS with 91–225 ms
  average frames and repeated 500+ ms slow frames, without a runtime exception.
- Traced the stall to synchronous construction of the provisional 545,171-
  triangle party-navigation graph after the dungeon became visible. A direct
  benchmark blocked the main thread for roughly 8 seconds and consumed about
  800 MiB of heap; its subsequent entrance surface/path queries also failed.
- Added a navigation policy that keeps reviewed, budgeted graphs enabled while
  unvalidated or over-50,000-triangle graphs use the existing local and
  breadcrumb party-follow fallback. Shadowfang therefore avoids the unusable
  graph, while Wailing Caverns retains its reviewed navigation.
- Re-entered Shadowfang from a clean reload and kept the live scene open beyond
  the former graph-build window. Eight consecutive samples held 59–60 FPS at
  about 16.7 ms per frame; the slowest final sampled frame was 25.8 ms. The party,
  HUD, minimap, combat state, and full environment remained responsive and
  visible in the reviewed screenshot. No new browser errors were emitted.
- Verification: all 20 focused navigation cases passed, the complete Vitest
  suite passed all 844 cases across 154 files, the production build completed
  with 1,048 modules transformed, KTX2 and environment-performance audits
  remained green, and the release-candidate check passed. The existing
  large-chunk advisory remains informational.

# 2026-08-31 — dungeon-fleet startup freeze audit

- Audited the complete registered fleet: 84 regular-dungeon navigation assets,
  all 75 Manastorm package assets, and all 470 published Manastorm stages. Every
  referenced navigation GLB exists and parses. Under the reviewed 50,000-
  triangle policy, 31 regular and 41 Manastorm package graphs remain enabled;
  the other 53 regular and 34 Manastorm packages use breadcrumb following.
- Added an actual-geometry triangle check before navigation allocations. This
  prevents stale catalog metadata from bypassing the synchronous-build ceiling;
  Wailing Caverns metadata was corrected from 32,077 to 45,732 triangles.
- Found nine zero-length imported patrol routes that could throw during scene
  render: one in Scholomance, two in Botanica, two in Magister's Terrace, and
  four in Escape from Durnholde. Catalog generation now preserves those actors
  as static spawns, and runtime normalization provides the same defense for
  future malformed imports. All remaining 752 regular patrols and every
  effective Manastorm stage patrol have usable routes.
- Live entrance checks loaded Scholomance, Escape from Durnholde, and Forgotten
  Mine without a scene error or startup freeze. Scholomance settled near 28.5
  FPS at 946 draw calls; Durnholde near 29 FPS at 3,663 draw calls; Forgotten
  Mine near 50 FPS at 1,502 draw calls after staged loading. These identify
  remaining environment/population throughput work, but are separate from the
  fixed startup-crash and synchronous-navigation failure modes.
- Verification: the complete Vitest suite passed all 850 cases across 155
  files, the production build completed with 1,048 modules transformed, all
  106 environment assets passed their performance audit with no regressions,
  the KTX2 audit found no fallback textures, and the release-candidate check
  passed. The existing large-chunk advisory remains informational.

# 2026-09-04 — optional content offload and basic Human fallback

- Built and served the production browser output from a clean localhost origin.
  A fresh Undead female Priest preview rendered with the real bundled Human
  female rig and default textures while retaining the Undead/Priest identity in
  the roster. No optional character pack had been installed.
- Entered Wailing Caverns as healer from a separate clean browser origin. The
  Undead player, Undead tank, and Goblin party member all rendered through the
  matching basic Human fallbacks while their saved race, role, class, equipment,
  and party labels remained intact. The dungeon stayed playable and emitted no
  new error; only the existing third-party initialization deprecation warning
  appeared.
- Exercised the browser Downloads lifecycle with a small local two-revision
  content fixture. Missing character and original-media packs showed Download;
  a verified install showed Remove; the changed revision showed Update and
  Remove together; and applying Update returned the pack to offline-ready state.
- Reloaded after service-worker activation. The application and lazy JavaScript
  chunks loaded normally, while the content worker continued to reserve managed
  character, creature, equipment, game, and UI paths for bundled or verified
  cached content. The initial broad `/assets/` interception found during this
  playtest was corrected before the final run.
- In-dungeon Options showed Generated as the active default, kept Original
  disabled until its optional spell-effects-and-sounds pack is downloaded, and
  exposed a Download action for that pack.
- Evidence:
  `playtest-artifacts/2026-09-04-optional-content/roster-undead-priest-basic-human-fallback.png`,
  `dungeon-optional-models-basic-human-fallback.png`,
  `browser-downloads-optional-packs.png`, and
  `browser-downloads-update-and-remove.png`.
- Verification: all 43 avatar tests passed; the complete Vitest suite passed all
  896 cases across 164 files; production browser and Android builds completed;
  both binary inventories contained only the two Human rigs, starter equipment,
  and Wailing Caverns content; and neither output contained optional character
  packs, original combat media, or WAV files. KTX2 and environment-performance
  audits remained green, and the release-candidate check passed. The existing
  large-chunk advisory remains informational.

# 2026-09-05 — Power Word: Shield live character anchoring

- Installed the optional Original Spell Effects & Sounds pack through the
  production Downloads screen, enabled it, and selected Original spell visuals.
- Entered Wailing Caverns as a Human Priest and cast Power Word: Shield on both
  the party tank and the player while exercising party movement and combat. The
  persistent original aura stayed at the rendered character root instead of
  following the throttled gameplay-position snapshot, with no inherited camera
  or character rotation.
- Reviewed live dungeon screenshots before and during the movement test. The
  browser emitted no new errors; only the existing third-party initialization
  deprecation warning remained.
- Verification: all 899 Vitest cases passed across 165 files, including three
  new visual-anchor registry cases. The production build completed with 920
  modules transformed, and asset audits remained green. The existing
  large-chunk advisory remains informational.

# 2026-09-05 — modular RuneWaker boss mechanics

- Built a temporary local production content feed for Forsaken Abbey and the
  first two Ystra Labyrinth boss rooms, then installed each pack through the
  browser Downloads workflow with checksum verification.
- Loaded the source-backed Specter Guardian Creator encounter. Its mechanic
  panel appeared below the minimap and buff/debuff strips with Guard Bomb,
  Transport Magic, and Meteor Shower countdowns. A live Meteor Shower cast
  showed its one-second cast warning over the boss.
- The first pass exposed that four countdown rows could crowd the boss's
  pending 80%, 60%, 40%, and 20% phase triggers out of the panel. The timeline
  now limits only countdown rows and always retains every pending health
  threshold. The corrected production build displayed all eight rows without
  covering the central playfield.
- Loaded Rune Guardian Leader in Ystra Labyrinth Boss Room 2. Its source-backed
  Manastorms Duo Lv5 mechanic displayed a large purple circular floor warning
  while the matching panel row counted down at 0.6 seconds. The area indicator,
  cast timing, party damage, and effect/debuff presentation all advanced in the
  live encounter.
- Browser logs contained no uncaught application errors. The existing
  third-party initialization deprecation warning remained; WebGL context-lost
  messages corresponded to intentional production rebuild reloads and opening
  replacement test tabs.
- Verification: the full suite passed all 913 cases across 169 files, and the
  corrected production build completed with 923 modules transformed. KTX2 and
  environment-performance audits remained green, and the release-candidate
  check passed.

# 2026-09-05 — specialization-aware heirlooms and endgame loot

- Activated the legacy level-10 Priest `Rolecheck`. The mandatory Class Hall
  picker appeared before the main menu, offered Discipline, Holy, and Shadow,
  and presented Cancel, Keep talents, and Reset talents after choosing Holy.
  Keeping talents returned to the main menu with the specialization saved.
- Opened Wailing Caverns from Dungeon Finder. Role selection was restricted to
  Healer, matching the saved Holy specialization; no Tank or DPS override was
  available.
- Reviewed the roster preview and a live Wailing Caverns combat scene. The
  player body, staff, and all four AI party bodies/equipment silhouettes were
  visible and correctly anchored without obscuring the HUD or playfield.
- Opened Inventory & Equipment in the dungeon. The player had zero earned
  items but every applicable slot resolved to a level-10 Priest healer
  heirloom and reported average item level 10. Morvane, the Blood Death Knight
  companion, showed a complete tank heirloom loadout with the intended
  two-handed weapon layout and empty conflicting slots.
- Browser logs contained no uncaught application errors. Only the existing
  third-party initialization deprecation warning appeared.
- Automated coverage validates every classic and CoA specialization, Feral
  Bear/Cat separation, RoM-derived profiles, player and party scaling caps,
  same-level catalog dominance at levels 1/40/79, real item-level-85 upgrade
  availability for every gear archetype, two-hand replacement scoring,
  fallback reappearance, percentage-preserving stat refreshes, specialization
  persistence and talent reset behavior, and the item-level-255 save limit.
- Verification: the complete Vitest suite passed all 928 cases across 171
  files, including all 43 avatar cases. All 15 server persistence/API tests
  passed. The production build completed with 925 modules transformed, KTX2
  and environment-performance audits remained green, and the release-candidate
  check passed. The existing large-chunk advisory remains informational.

# 2026-09-07 — modular encounter completion and boss guide

- Entered Wailing Caverns through the live offline roster and verified the 3D
  dungeon, party HUD, minimap, and buff/debuff strips remained readable. The
  encounter timeline occupies the same top-right HUD column beneath those
  elements, keeping timed casts and pending health thresholds out of the
  central playfield.
- Opened the new Boss Guide from the in-dungeon pause menu. The three-column
  field manual listed both WoW and RuneWaker dungeons, their bosses, source
  provenance, triggers, timings, target/effect descriptions, warning geometry,
  counterplay, and a concrete test procedure for every listed mechanic.
- Selected Archaedas in Uldaman and confirmed all four authored mechanics:
  Awaken Earthen Custodian every 10 seconds, the two-second 18-metre Ground
  Tremor warning circle, six Earthen Guardians at 70% health, and two Vault
  Warders at 40% health. Bosses with no sourced special attack are explicitly
  labelled with zero special mechanics instead of receiving invented behavior.
- The encounter authoring layer now drives serializable timers, health gates,
  guards, actions, summon groups, chained signals, stateful world objects,
  teleports, rewards, checkpoints, and online leader-authoritative interaction
  validation. Harmful area mechanics use the same geometry for both the world
  warning and hit resolution.
- Browser review found no visible layout or interaction regressions. The only
  previously observed browser message remains Rapier's third-party
  initialization deprecation warning.
- Source limitations and the exact remaining RuneWaker evidence, navigation,
  formula, loot, animation, and asset-fidelity gaps are recorded in the linked
  deterministic audit reports and authoring documentation rather than silently
  approximated.
- Full-WDB environment packages for Raksha Temple and The Origin completed and
  passed their GLB gates. Treasure Trove's 908-placement visual/collision stage
  also completed, but its full-resolution 36-tile Recast bake did not terminate
  after nearly two CPU-hours on the watershed path and more than four CPU-hours
  on the monotone path. The fail-closed pipeline preserved the prior validated
  shipping package; this remaining offline bake is documented in
  `ENCOUNTER_AUTHORING.md` instead of lowering navigation resolution.
- Verification passed: the complete Vitest suite passed all 981 cases across
  185 files; online authority passed all 4 server cases; RuneWaker gameplay,
  encounter, environment, navigation, formula/loot, and animation audits passed;
  and the production build completed with 955 modules transformed. The KTX2
  audit covered 12,347 compressed textures with zero fallbacks, and the
  environment-performance audit found no regressions. The existing large-chunk
  advisory and Rapier initialization deprecation warning remain informational.

# 2026-09-08 — WoW dungeon production-functionality audit

- Audited all 54 WoW five-player catalog entries against live runtime contracts,
  pinned AzerothCore population/object drafts, and instance-script source. The
  composed-runtime content audit is green for 1,193 mobs, 273 bosses, 4,026
  attacks, 2,572 specials, and 1,275 native models, but the stricter production-
  fidelity gate remains blocked at 0/54.
- The audit found 3,943 source game-object spawns, including 2,321 actionable
  doors, buttons, chests/resources, transports, traps, and related objects. Raw
  database rows are no longer passed into the typed world-object renderer; all
  54 live definitions now have zero malformed world-object records.
- The follow-up implementation sweep completed source review for all 54
  instances. Fifty-one now have runtime-wired instance orchestration and the three
  genuinely empty stock instance scripts are pinned as source-exact no-ops;
  there are no unreviewed instance scripts left. Wailing Caverns remains the
  only entry with a reviewed off-mesh link, and the three Ascension-specific
  dungeons still lack authoritative server population data.
- Entered Wailing Caverns from the local browser roster as a Priest healer. The
  3D environment, populated combat route, party, objective, minimap, and HUD all
  loaded; the minimap strip showed one Power Word: Fortitude entry. No uncaught
  client runtime errors appeared. Rapier's existing initialization deprecation
  warning remained informational.
- The local content manager exposed all 54 WoW entries but only Wailing Caverns
  was installed; the other 53 were marked download required. Those dungeons were
  therefore checked by source/runtime-contract audit in this pass, not claimed
  as completed visual traversal playtests.
- The follow-up world-object compiler converted all 3,943 source rows—including
  all 2,321 actionable doors, chests/resources, switches, traps, and
  transports—into stable typed definitions with transformed coordinates,
  type-aware states, conservative dimensions, and source/template/spawn/type
  provenance. Wailing Caverns receives its 57 generated records through the
  authored-definition enrichment path as well.
- Client `GameObjectDisplayInfo` metadata resolves a visual display ID for
  3,905 records, a native model path for 3,837, and exact bounds for 3,547. The
  generated export manifest deduplicates those instances into 458 native legacy
  assets, records 38 intentional no-display spell-focus objects, and isolates
  the remaining nine missing-model display IDs and 68 no-bounds display IDs for
  review.
- An initial live pass exposed colored procedural markers for adapted
  chests/resources. Unreviewed source objects are now deliberately hidden,
  non-blocking, and non-interactive so incomplete door scripts cannot soft-lock
  a dungeon and placeholder art cannot leak into normal play. A second live
  Wailing Caverns pass confirmed the clean 3D scene, party combat, HUD, minimap,
  and single Fortitude entry. No uncaught client errors appeared; duplicate
  KTX2-loader warnings occurred only across deliberate hot-reload cycles.
- The final runtime audit reports 54/54 reviewed dungeons. Fifty-one dungeons use
  52 rule-bearing instance-controller definitions (Blackrock Spire has separate
  lower- and upper-spire controllers), while three use verified source-exact
  no-op scripts. Every catalog objective has a completion action, and the audit
  found zero malformed-object, duplicate-id, or encounter-contract issues. All
  51 rule-bearing dungeons retain explicit adaptation/gap notes; native object
  models and full-route live traversal are not claimed complete.
- The source-combat follow-up gives every runtime-spawnable boss a sourced kit
  unless the pinned source really is melee-only. The six deliberate exceptions
  are Kresh, Deviate Faerie Dragon, Spirestone Battle Lord, Spirestone Butcher,
  Crystal Fang, and Grubbis. Crystal Fang retains its source-exact 15% rare-spawn
  behavior and death summon. Trial of the Champion, Utgarde Keep, Hellfire
  Ramparts, Sethekk Halls, Mana-Tombs, and Slave Pens event bosses now use their
  pinned spell IDs and timing windows; unsupported vehicle, proc, queue, shared-
  health, phase, stack, and healing-reduction semantics are explicitly recorded
  as adaptations instead of being silently invented.
- Wipe/reset validation exhaustively covered 673,664 reachable encounter states,
  180 reset paths, every one of the 68 non-initial reset targets, and all 135
  entity-timeout rules. Durable progression now fails closed on a missing entity,
  legacy snapshots with unordered action history fail closed, and Frozen Reach's
  Arktos is attempt-scoped so a wipe respawns him and can still advance the duo
  encounter on the next pull. The two viable post-reset resume states in Scarlet
  Cathedral and Halls of Stone are intentional, not soft locks.
- Same-process checkpoint restore and online follower hydration now accept their
  trusted save/leader snapshots even when a retained local provider has a newer
  reset revision; ordinary stale imports remain rejected. A renderer restart
  preserves the current encounter snapshot. Encounter effects use one delivery
  path at a time, and obsolete headless queues are cleared on reset or dungeon
  switch instead of retaining already-applied envelopes for the process lifetime.
- The generated download topology now contains an installable logical pack for
  every one of the 54 campaign dungeons. The three custom dungeons without their
  own creature payload are represented by dependency-only packs, while every
  non-bundled dungeon pulls its exact environment and shared-creature closure.
  The existing live content feed was not deployed or mutated during this pass;
  browser traversal therefore still reflects the packs already available there.
- Installed packs now persist v2 verification receipts containing their exact
  file set, byte counts, hashes, and dependencies. Browser-cache and native-file
  presence are revalidated before entry, production dungeon entry fails closed
  when a pack or dependency is missing or stale, and `content:publish` validates
  the complete 54-dungeon dependency graph before any release output is written.
- Run completion now leaves the final room active so loot can be collected before
  the player explicitly exits and clears the checkpoint. A wipe returns the party
  to the saved run spawn while preserving already-cleared bosses and durable run
  progress; attempt-scoped actors and effects reset for the next pull.
- The live Priest check used the saved `Rolecheck` character, which advanced from
  level 11 to level 12 during the dungeon passes. The latest Wailing Caverns run
  displayed 677/677 health, one Power Word: Fortitude alongside Arcane Intellect
  (and Holy Concentration while active), and one Frostbolt plus one Icy Touch
  debuff. Repeated Frostbolt applications refreshed that one entry from 7.1 to
  21 seconds rather than duplicating it. The deterministic geared level-8 check
  produced 3,285 mana and increased health from 407 to 437 when Fortitude was
  applied.
- Verification remains deliberately split: source coverage, runtime contracts,
  reset reachability, and automated behavior are green, while the combined
  production-fidelity audit remains 0/54. Its remaining blockers include native
  object export/collision, explicitly listed source-semantic gaps, and end-to-end
  traversal evidence for every production pack—not unfinished instance-script
  review.
- Final release verification passed all 1,109 Vitest cases across 202 files,
  TypeScript project compilation, and the production build with 967 modules
  transformed. The build staged 5,587 browser-bootstrap files, validated all
  eight arenas, audited 12,347 compressed textures with zero fallbacks, and found
  no environment-performance regressions. The release-candidate check accepted
  all 118 changed paths. The existing large-chunk advisory and Rapier
  initialization deprecation warning remain informational.

# 2026-09-09 — fixed per-rank classic mana costs

- Imported WoW mana percentages are now converted into fixed costs at each
  spell rank's source level. Player level, Intellect, equipment, and maximum-mana
  modifiers no longer increase the cost of an already learned rank; legitimate
  talent and aura cost reductions are still applied afterward.
- Shadow Word: Pain rank 4 now costs 133 mana at every level from 26 through 33,
  including level 32. Learning rank 5 at level 34 raises the cost to 188 mana.
- AI companions now use real character-scale mana pools so fixed spell costs do
  not prevent healing or resurrection. The companion display also resolves the
  active spell rank and the same final affordability result as the primary HUD.
- Verification passed all 1,133 Vitest cases across 208 files. The production
  build completed with 974 modules transformed; arena, KTX2, and environment-
  performance audits remained green, and the release-candidate check passed.

# 2026-09-09 — dungeon wall and nameplate performance regressions

- The 0.1.24 material fallback honored BLEND tags on legacy WoW WMO walls,
  floors, and ceilings, disabling their depth writes. WoW environments now use
  their authored material roles again; RuneWaker environments retain their
  source MASK/BLEND semantics. Shipped Wailing Caverns wall materials, vines,
  and water are covered by regression tests.
- The 0.1.25 boss-name occlusion scanned the complete rendered scene with
  Three.js raycasts. A live Kalin's Shrine CPU profile was dominated by triangle
  intersection and vertex decoding. Boss names and nearby NPC prompts now use
  bounded camera-to-label queries against Rapier's existing physical spatial
  index, at most ten times per second, skipping offscreen labels. Walls, floors,
  and ceilings occlude labels; actors and trigger sensors do not.
- In an isolated offline Edge playtest at 1440 x 900 on an AMD RX 9070 XT,
  Kalin's Shrine improved from about 10 FPS (98.5 ms/frame, 167 draw calls,
  DPR 1.125) to 55.5-56.1 FPS (17.8-18.0 ms/frame, 168-170 draw calls,
  DPR 1.5). This is a representative entrance/combat comparison, not an audit
  of all thirty RoM dungeons. The environment alone took about 0.8 ms CPU and
  0.5 ms GPU with 14 draw calls.
- Wailing Caverns screenshots show continuous opaque cavern walls and ceilings.
  A controlled toggle of the old material flags reproduced the missing wall
  sections at the same camera position; restoring depth-writing materials fixed
  them. Its full scene ran at about 57 FPS. No uncaught client errors appeared.
- Evidence and repeatable local benchmark helpers are under
  `playtest-artifacts/2026-09-09-dungeon-rendering/`: `kalin-before.cpuprofile`,
  `kalin-after.cpuprofile`, `kalin-after-metrics.json`, `kalin-before.png`,
  `kalin-after.png`, `wailing-regression.png`, and `wailing-restored.png`.
- All 26 focused material, label occlusion, combat sight, aggro visibility,
  instancing, and NPC-prompt tests passed. Physical-occlusion tests exercise
  actual Rapier worlds rather than just source-code assertions.

# 2026-09-09 — campaign rendering audit and Thor performance preparation

- Audited the actual visual GLB headers for all 54 WoW campaign dungeons:
  9,402 materials, including 3,586 legacy BLEND materials restored to opaque
  rendering by the shared policy. Every WoW package contained affected
  materials. A permanent shipped-asset test covers all 54 definitions and
  retains explicit cutout, blend, and additive roles. This is material-policy
  coverage, not a claim that every room has been visually traversed.
- Live screenshots checked Wailing Caverns, Shadowfang Keep, Utgarde Pinnacle,
  the Nexus, and Tinkertech Showdown. The sampled walls, floors, stairs, and
  ceilings remain solid; authored lighting effects remain transparent. Kalin's
  Shrine and Forsaken Abbey provided RoM rendering checks. Roster and in-dungeon
  screenshots show character bodies and equipped weapon silhouettes after the
  shared character animation change.
- A controlled Utgarde Pinnacle entrance test found about 1,739 draw calls,
  1.95 million triangles, and 30.5 FPS on the desktop GPU. Clipping at the
  existing full-fog distance reduced this to about 209 calls, 567,000 triangles,
  and 60 FPS at the same viewpoint. The camera now follows each dungeon's fog
  distance, including transitions that retain the existing Canvas/camera.
  The background uses the fog color to avoid a clipping seam at the horizon.
  All 84 dungeon definitions have render-distance coverage.
- Static dungeon visual transforms are cached until the root or its parent's
  world transform changes. Tests verify imported matrices, repeated unchanged
  frames, root movement, and parent movement. Characters, moving world objects,
  and physics remain outside this immutable visual cache.
  The final 63-second Utgarde sample held 60 FPS with 213 draw calls:
  main-thread work averaged 9.2 ms/frame, renderer CPU averaged 7.5 ms
  (p95 8.3 ms), and GPU averaged 2.7 ms (p95 4.1 ms). Movement and jump input
  continued to animate the player while the environment remained stationary.
- Character locomotion and jump clip selections are cached by animation
  library and weapon stance. The final Kalin CPU profile contains no sampled
  per-frame base-clip selection/name parsing. Frame-cap deadlines now tolerate
  sub-millisecond timestamp rounding without drifting or emitting catch-up
  frames; tests cover 60 FPS on rounded 60 Hz and 120 Hz callbacks.
- A 31.5-second Kalin sample with a connected browser companion display held
  60 FPS: main-thread work averaged 8.7 ms/frame, renderer CPU averaged 5.4 ms
  (p95 5.9 ms), and GPU averaged 2.4 ms (p95 3.4 ms). The companion used about
  5.1 ms of main-thread time per second. Recent ability-cost snapshot work was
  not a dominant hotspot in this sample. Startup warmup is bounded to loading;
  companion snapshots are already limited to ten per second during gameplay.
- These are isolated offline Edge desktop samples on an AMD RX 9070 XT at
  1920 x 1080 CSS pixels, DPR 1.5 (2880 x 1620 renderbuffer). Party composition
  can change between dungeon entries. They do not certify Android performance
  or sustained busy-combat headroom. ADB listed no connected Android device.
- Thor acceptance should use the actual model, both native displays, target
  resolution, and a sustained 20-minute traversal/combat run after warmup.
  Provisional headroom target: p95 main-thread and GPU work each at or below
  12 ms, with p99 presented frame intervals within the 16.7 ms 60-FPS budget.
  Test large pulls, boss effects, menus, and repeated dungeon transitions;
  record thermal behavior and frame-time spikes, not only average FPS.
  The environment asset audit still has 60 grandfathered budget exceptions;
  its green result means no new asset regression, not a Thor FPS guarantee.
- Evidence is in `playtest-artifacts/2026-09-09-thor-audit/`, including roster,
  dungeon, and companion screenshots, CPU profiles, and CPU/GPU JSON samples.
  The full material report is
  `playtest-artifacts/2026-09-09-dungeon-rendering/wow-material-audit.json`.
- Validation passed 227 focused tests covering avatar rendering and animation,
  camera distances, material policies, frame pacing, static transforms, label
  visibility, and companion snapshots. The final static-cache change also
  passed the 105-test scene subset. The production build and release-candidate
  check passed; the existing large-bundle advisory remains.

# 2026-09-09 — additional CPU headroom from culling and shared skeletons

- Prepared dungeon visuals now group static siblings into conservative spatial
  bounds. One scene hook checks these bounds after the current camera and world
  matrices update, before Three.js traverses individual meshes. Bounds include
  complete wall geometry and all instances; lights, animated meshes, and meshes
  explicitly disabling frustum culling remain outside these groups. The hook
  supports streamed chunks and restores its predecessor after cleanup.
- At the same Utgarde Pinnacle entrance, only 16 of 1,383 spatial groups were
  visible. The renderer examined 108 static meshes instead of 7,002, retaining
  the same 206 draw calls and 566,913 rendered triangles. Controlled culling
  toggles produced byte-identical canvas PNGs across eight camera headings.
- In the same running scene, spatial grouping reduced average main-thread work
  from 9.45 to 6.83 ms/frame, and renderer CPU from 7.81 to 5.31 ms/frame.
  Removing containers emptied by static instancing reduced the final fresh-run
  averages further to 6.49 ms main-thread and 4.85 ms renderer CPU (p95 5.2 ms).
  The final sample lasted 87.8 seconds and held 60 FPS. Authored empty scene
  nodes are preserved; only containers emptied by our batching are removed.
- The shared character renderer still uses SkeletonUtils.clone, then reuses
  equivalent skeleton palettes within that owned clone. Bone identities, order,
  and inverse bind matrices must all match; per-mesh bind transforms remain
  unchanged. Different characters retain independent bones and animations.
  Owned bone textures are now explicitly disposed during character teardown.
- A live character audit found 66 skeleton objects for one player palette and
  256 objects for four party palettes. A Wailing Caverns toggle removed 309
  duplicate skeletons and produced an identical rendered image. Tests verify
  posed vertices, differing bind transforms, incompatible palettes, independent
  cloned characters, and redundant texture disposal.
  In the same Wailing scene, main-thread work fell from 7.40 to 6.70 ms/frame;
  renderer CPU fell from 5.51 to 4.86 ms/frame, and allocated GPU textures fell
  from 292 to 217. This measures palette sharing after spatial culling was
  already enabled. Both samples held 60 FPS.
- The final built runtime showed a complete roster body and staff with one
  skeleton for 66 skinned meshes. Kalin's Shrine showed one player palette and
  four independent party palettes; walking retained the player body and weapon
  silhouette. Wailing Caverns and Utgarde screenshots retained solid dungeon
  surfaces. There were no uncaught client errors during these playtests.
- All 210 targeted avatar, visibility, instancing, material, camera-distance,
  and label tests passed. The production build and release-candidate check
  passed. The existing large-bundle advisory remains.
- Evidence is under `playtest-artifacts/2026-09-09-static-culling/`: paired PNGs,
  eight-heading pixel comparisons, profiles, before/after timing JSON, and
  screenshots. Measurements use the same desktop Edge/RX 9070 XT setup as the
  preceding audit. These CPU improvements do not certify Thor performance.

# 2026-09-09 — bundled arena assets blocked by the browser content worker

- Reproduced the reported collision-matched fallback in the production build
  with the content worker active. Warnorken's `visual.glb?arena-runtime=1`
  returned 404 even though the file was bundled and directly served with 200.
  The worker omitted `/assets/game/arenas/` from its bundled asset prefixes and
  incorrectly rejected these files as uninstalled optional content.
- Added the arena prefix to the worker. Installed updates still take priority;
  missing optional dungeon content still returns 404. Behavioral worker tests
  failed for all eight arenas before the fix and now cover visual, collision,
  and navigation requests with the runtime query parameter.
- The production build passed. Reloading the same browser session activated
  the fixed worker without clearing storage or downloading content. Warnorken
  rendered textured stone geometry and Nagrand rendered its authored arena
  geometry; neither showed the loading/fallback banner after loading, and no
  new console errors were recorded. Checks used offline practice in the Codex
  browser at `http://127.0.0.1:4174/`, backed by a local test server on port 4173.
- All 22 content tests and the eight-arena asset validator passed. Screenshots,
  before/after error logs, asset validation, and build output are saved under
  `playtest-artifacts/2026-09-09-arena-loading/`. The existing large-bundle
  advisory remains. This verifies the local browser build, not a deployment
  or an Android package.

# 2026-09-10 - arena character and gameplay integration

- Replaced the limited arena spell interpreter with a private bundle of the
  existing character combat engine. Offline practice and server matches share
  casting, channels, resources, auras, pets, talents and equipment mechanics.
  Each actor/match keeps its own state; opening panels does not run a second
  client simulation or overwrite dungeon state.
- The controlled character uses CharacterModel with saved appearance, gear,
  class, specialization, talents and bindings. Other participants also use
  character models. The HUD displays actual resources, runes, combo points,
  cast/channel status, cooldowns, buffs, pet commands and PC/controller labels.
  Spellbook, character, talents, bindings, keybinds, map and options work in
  arena. Equipment and talents remain locked to the entered match loadout.
- Fixed Keybinds opening with a shortened custom PC bar: missing slots now
  display as empty instead of crashing the React tree. Live checks opened
  Keybinds, Character, Talents and Options; the four-slot Q/E/R/T bar cast
  Flash Heal and Power Word: Shield with resource/cast/aura feedback.
- Fixed NaN facing, spectator-camera elevation, obstructed starting positions
  and camera collision. Shared right-button drag ends on release, blur, lost
  capture, menus and teardown without pointer lock. Movement, jump, targeting,
  ground placement, feedback, animation and combat effects use arena state.
- Authority uses all collision chunks and navigation triangles. Removed
  invisible static proxy walls. All 3v3 starts are grounded, and bots follow
  connected routes around walls/ramps. Shortcuts stay on level ground to avoid
  falling below waypoints. Detailed scenes use a shaped support surface; the
  loading/error fallback now matches collision instead of using a flat box.
- Black Rook's old navigation/anchors were on terrain above the arena. Rebuilt
  against restored collision, selected its connected interior component (783
  triangles), and placed both starts on the floor near Y=-28.3. Updated the
  asset metadata and generator to stop restoring the old terrain navigation.
- Added PvP defender stats, shields, resource drains, pushback, school locks
  and control diminishing returns. Fixed duplicate control entries counting
  twice, aligned aura/control duration, and made dispels remove movement
  restrictions. Dispel Magic supports friends and enemies. Healers can heal
  allied pets; melee bots start weapon attacks to generate rage.
- Automated checks: 38 server arena tests and 160 distinct avatar, input,
  camera, arena character/rules, combat store and ability catalog tests.
  Coverage includes channels/cancellation, runes, combos, pet orders/damage/
  healing, interrupts, absorbs, dispels, DR, ground spells, mana burns, RoM
  secondary resources, dampening, isolation, matchmaking and match results.
  Both-direction walks between team starts pass across all eight arenas.
- Live checks use an isolated in-memory level-80 priest, ArenaPriest, without
  changing the user's roster. Warnorken showed its interior, the selected
  character and saved spell labels. Movement changed both actor and camera.
  Bots traversed, cast spells, reduced health and settled a defeat when idle.
  No detailed-asset warning appeared. Evidence uses the arena-complete prefix
  under playtest-artifacts; earlier Nagrand/custom-bar evidence retains the
  arena-player prefix. The drag probe uses installed input code with scripted
  right-button events; rotation changes while held, stops after release, and
  pointerLockElement remains null. This is not a physical-device test.
- Final screenshots show the Warnorken interior and character/action bar,
  Black Rook's corrected interior floor, the roster body/staff, and the
  character/party after returning to Wailing Caverns. Files:
  arena-complete-warnorken.png, arena-complete-black-rook.png,
  arena-complete-roster.png, arena-complete-dungeon.png and
  arena-complete-keybinds.png in playtest-artifacts.
- Production build and release checks passed, with the existing bundle-size
  advisory. Run npm run arena:abilities:build after catalog/asset changes;
  prebuild also regenerates the engine and geometry. Verification is local
  offline practice plus authoritative server tests. Hosted online PvP,
  deployment and Android hardware were not exercised.

# 2026-09-10 — profile keybinds and movement-conflict recovery

- Added Pause / Options → Keybinds with 133 actions, PC and Controller tabs,
  search/categories, capture and touch selectors, explicit conflict replacement,
  clear/reset controls, and Restore movement. Desktop defaults to PC; mobile
  and native Android default to Controller. Both tabs remain available.
- Live offline browser check: assigning W to bar 1/button 1 named Move forward
  in the conflict prompt; cancelling preserved W. Rebinding forward to T,
  saving, closing, and reopening retained T. Restore movement returned it to W.
- At 390 × 844, controller selectors assigned L1 + Triangle to Jump, prompted
  before clearing the old layer-2 ability binding, and retained the saved chord
  after reopening. Restored the test bindings afterward. The final phone panel
  starts at (0, 0), has a 443px scrolling action list, and keeps the footer within
  the viewport. Desktop layout checked at 1280 × 900.
- Fixed the inherited 610px panel-height cap, mobile backdrop padding, action
  button accessible names, and companion-toggle overlap found during QA.
- 89 targeted tests passed across input dispatch, mouse-drag lifecycle, analog
  remapping, conflict recovery, menus, account isolation, cloud transfer, UI
  imports, and dual-display synchronization. The full production build and
  release candidate check passed; the existing bundle-size advisory remains.
- Controller chord conflicts also cover a modifier button’s standalone action,
  preventing combinations such as View + A from opening the companion first.
- Hardware controller / Android APK validation and a physical right-button
  drag were unavailable through this browser session. Drag lifecycle and null
  pointer-lock assertions passed in the input tests. Browser viewport screenshots
  are scaled by the in-app compositor; DOM geometry supplied the viewport checks.

## 2026-09-10 - Original spell media defaults and visual fallback

- Original spell visuals and authentic sounds are enabled by default. The original media pack is prepared before gameplay; download failure leaves generated visuals available and missing sounds silent.
- Saves using the previous generated-only default migrate once to original media while retaining quality, group visibility, mute, and volume settings. Explicit later preferences remain saved.
- Removed synthesized combat audio. Original mode now uses generated visuals when no original phase is mapped or when all original graphs/textures for that phase fail to load; usable original layers remain preferred.
- Validation: 64 targeted tests passed, covering defaults, persistence migration, media installation/retry, rendering selection, original audio playback, silence on missing/invalid audio, and existing effect lifecycles. Production TypeScript and Vite build passed after correcting the installer test mock type; prebuild content/asset checks also passed.
- Browser verification used the production CombatEffects and OriginalCombatEffect components with isolated store/media fixtures and the actual packaged Flash Heal graph/textures. The original case rendered one original emitter and zero generated emitters. Unmapped spell, graph HTTP 404, and texture HTTP 404 cases each rendered zero original emitters and one visible generated effect. Combat reset removed the fallback.
- Evidence: playtest-artifacts/spell-media-20260910/original.png, bad-graph.png, bad-texture.png, and missing.png. The first Vite preview timed out; the bundled isolated preview completed successfully. This was an effect-focused browser check, not a full dungeon or device audio audition.
- Final release check: blocked by existing trailing whitespace in server/generated/arena-combat-runtime.mjs. The files changed for spell media pass the whitespace check; the unrelated generated arena file was left intact.

## 2026-09-10 - All regular dungeons bundled with lossless shared storage

- Web and Android packaging include all 84 regular dungeons and their asset dependencies. Shared Manastorm maps are stored once while every Manastorm pack keeps its optional install flag. Manastorm-only maps remain outside the initial package.
- Packaged assets fell from 3,440,440,928 logical bytes to 1,600,690,686 stored bytes, including 179,063,519 bytes saved by sharing exact duplicate files. No source models, texture resolutions, animations, or audio were changed or generated.
- Gzip representations retain both compressed and original SHA-256/size metadata. Publishing keeps original remote objects for older clients. Browser installs verify both representations, retain compressed shared objects with small aliases, and preserve shared objects until their final pack is removed. Native optional downloads remain resumable and verify their expanded originals before activation.
- Android's packager automatically expanded and renamed .gz files; final objects use .gzip with noCompress instead. Added an APK verifier to the normal Android build workflow, checking every object, every original hash, exact object inventory, and the 84-dungeon/optional-Manastorm boundary.
- Final debug APK: 1,622,100,512 bytes. All 9,106 unique objects (9,194 logical asset references) passed verification from inside the actual APK. Verification report and test APK: artifacts/dungeon-texture-audit-20260910/apk-verification.json and healer-man-0.1.28-debug.apk.
- Validation: production TypeScript/Vite build and asset checks passed; 80 targeted content, Downloads, and avatar tests passed across the final runs, plus nine publisher/server tests. Android Java compilation and assembleDebug passed. Existing bundle-size/import advisories remain.
- Live production browser check used an isolated offline Gziphealer character. Roster and Deadmines rendered the player body, party bodies, equipped weapon silhouettes, and textured dungeon geometry using the bundled compressed objects. Downloads showed all regular dungeon packs included and 75 Manastorm rows not installed. Evidence: artifacts/dungeon-texture-audit-20260910/roster-final.png, deadmines-final.png, and downloads-dom.txt. No new console errors appeared in the final-format dungeon run.
- A concurrent build replaced the shared dist directory during the first preview, causing temporary asset 404s. Repeated visual validation against an isolated build completed successfully. Deadmines displayed a route-blocked party status at its entrance; this was a loading/rendering check, not a full dungeon traversal. Physical Android hardware was not exercised.

## 2026-09-11 - Original WoW visuals for every class

- Packaged all 378 directly mapped WoW models plus five model dependencies:
  4,205/4,205 source phase assets across all ten WoW classes, 260 mesh models,
  1,367 particle emitters, 193 ribbon emitters, and 428 original textures.
  The audit found no missing graph, dependency texture, or mapped sound file.
- Added source bone animation, skinning, billboards, material/UV animation,
  layered textures, ribbons, geometry particles, and bounded nested particles
  to the existing combat event rendering path. Graph URLs include payload
  hashes so regenerated models do not reuse stale cached graph files.
- The development preview at scripts/combat-effects/preview.html lists all
  directly mapped models and uses the combat renderer. Browser screenshots
  showed Battle Shout, Devotion Aura, Hunter's Mark, Slice and Dice, Priest
  Shield, Blood Strike, Lightning Bolt, Fireball, Shadow Bolt, and Wrath.
  Cone of Cold also showed the source geometry particles during playback.
- Browser QA exposed and fixed the second-texture UV shader declaration and
  a development media-pack gate: Vite serves original media directly, while
  release builds still require the verified optional pack. The offline Mage
  Visualtest rendered with its staff in the roster; player/party bodies and
  weapon silhouettes also rendered in Wailing Caverns.
- Evidence is in playtest-artifacts/2026-09-11-wow-effects. The initial dungeon
  screenshots preceded the development media-gate fix and demonstrate avatar
  rendering and generated fallback only.
- After the fix, Options reported Original media ready, and live Fireball
  casting showed source fire effects on the Mage (dungeon-original-fireball.png).
  No original-asset load or shader errors appeared. Existing physics-init and
  duplicate KTX2-loader warnings appeared during dungeon reloads. The party
  later reported Route blocked at the entrance; full traversal was not tested.
- Fixed instant-ability phase resolution without adding presentation-bus or
  audio events. Original mode now overlays mapped source-side cast layers on an
  instant release and falls back between source `impact` and `aura-start`
  terminal mappings while preserving the selected source phase. This reaches
  all 1,106 castable instant rank records with packaged source visuals,
  including all 287 latest-rank instant abilities with source visuals. Three
  additional source-backed instant records are passive and emit no combat
  event. The 61 latest-rank instant abilities with no source visual mapping
  continue to use generated fallback.
- Live Frost Nova QA reported original/high with two concurrent render systems,
  showing the source blue ground wave and its root layer instead of the prior
  generated semicircle. Evidence: dungeon-original-frost-nova-fixed.png. No new
  original-asset or shader warning/error appeared after the cast.
- Automated validation passed 20 Node packaging/decoder tests, 47 avatar
  tests, and 121 combat/media/audio/runtime tests (188 distinct
  tests across those suites).
  Source graph tests parse every shipped direct model, sample bone timelines,
  and check malformed geometry, dependencies, and billboard transforms.
- Production TypeScript/Vite build, prebuild content/asset checks, asset audit,
  and the release candidate check passed. Existing bundle-size/import
  advisories remain.
- Coverage is for source visual mappings in the current ability catalog.
  Effects without source mappings retain fallback. This is representative
  desktop browser QA, not exhaustive spell-by-spell fidelity or a physical
  Android/GPU performance test. No hosted release was published.

## 2026-09-12 - Stockades party routing and indoor lighting

- Fixed quantized navigation preparation: normalized integer positions now
  expand to floats before the GLB node transform is baked. Previously the
  Stockades world coordinates clamped into roughly [-1, 1], leaving the live
  party unable to attach to its graph even when an offline route test passed.
- Added three reviewed bidirectional links across the 1.537m Recast seams at
  the entrance/central stairs and the central/north and central/south stairs.
  The shipping 3,538-triangle graph connects the entrance and all five boss
  approaches while discarding the remaining small decorative islands.
- Added navigation-only review state and compiler-preserved runtime overrides.
  Stockades navigation is green; its wider encounter, wipe/reset, checkpoint,
  and completion validation remains unvalidated.
- Leader and follower formation paths now use the graph when a breadcrumb
  rejoin is blocked. Fallback comfort goals and recalls attach to the reviewed
  surface, preventing an off-boundary stair target from causing a false stuck
  loop. Combat approach connectors respect each member's personalized range.
- Replaced the generic dark expansion palette only for Stockades with warmer,
  stronger ambient/hemisphere fill and a warm directional light. No dungeon
  textures or source meshes were changed, and other dungeon palettes remain
  unchanged pending individual visual review.
- Live desktop browser check at 1280 x 720 used offline level-15 Priest
  Rolecheck as healer with Blood DK tank Morvane, Combat Rogue Vale, Unholy DK
  Thorne, and Frost Mage Ilyra. Fresh Defend mode showed PARTY 5/5, Recall on the
  entrance stairs recovered without a route warning, and Attack resumed with
  visible party movement down the stairs. The tank advanced into combat and
  took Defias Basic Attack/Kick/Backhand aggro; Backstab and Icy Touch events
  appeared while the player remained at the entrance.
- Inline task screenshots captured the warm, readable stonework/stairs and
  equipped party silhouettes in Defend and Recall-to-Attack states. The final
  runtime produced no new navigation errors; the existing deprecated physics
  initialization warning remains. Temporary diagnostic logging was removed.
- Automated validation: all 223 Vitest files passed, 1,373 tests passed, and
  six existing tests remained skipped. Regressions cover the actual quantized
  Stockades GLB through the shipping asset-preparation path, normalized integer
  transforms, formation projection, entrance-to-boss routing, and tank melee
  approach. This is an entrance/combat smoke pass, not a complete dungeon clear
  or physical Android test.
- Production TypeScript/Vite build, texture/performance audits, bundle
  verification, and the final release candidate check passed. Existing
  bundle-size/import advisories remain. No hosted release was published.
- The optional all-dungeon functionality audit was stopped after an extended
  run without output; its existing report files were not changed. Stockades
  navigation evidence comes from the actual shipping-asset regressions and
  the live smoke pass above, not a new fleet-wide functionality certification.

## 2026-09-12 - Stockades stair-ramp follow-up

- Follow-up player feedback reported companions still sticking near stairs.
  Actual collision profiles contain approximately 0.44-0.59m risers, above
  the party sweep's 0.32m lower-leg clearance. Collision-backed regressions
  reproduce uphill rejection on the original shipping collision GLB.
- Added three reviewed package-space traversal proxies for the entrance and
  both prison-wing stair flights. Collision tread/riser vertices within the
  reviewed stair relief are reshaped onto smooth ramp planes; upper walls,
  ceilings, other floors, and geometry outside those footprints remain intact.
  Source visual/collision GLBs are unchanged. Only prepared walking collision
  and party navigation use the overrides; the rendered stairs remain steps.
- Navigation receives the same ramp planes with 0.05m surface clearance and
  six new triangles stitching the exact source boundary portals. This replaces
  the three vertex-only connections with filled triangle corridors, allowing
  mid-stair replans and recalls to project onto the interior walking surface.
  The reviewed graph now has 3,544 source triangles and 31 components, retaining
  entrance-to-all-five-boss routes and discarding minor decorative islands.
- Live QA caught a second issue: terrain queries included the player's
  kinematic capsule. A failing collision-backed regression confirmed false
  route blockage through a nearby player. Party direct, route, and grounding
  queries now exclude dynamic/kinematic actor bodies and sensors; crowd spacing
  remains a separate party-separation concern. Static walls still block paths.
- Automated tests exercise shipping normalized/quantized GLBs through runtime
  preparation, downhill/uphill party capsule checks on all three flights at
  center and lateral offsets, repeated mid-stair replans, all boss routes with
  live wall checks, tank melee approach with real collision validation, and
  player capsule traversal down/back up all three ramps without jumping.
- Final desktop browser smoke pass at 1280 x 720 used offline level-15 Priest
  Rolecheck and the same Blood DK/Combat Rogue/Unholy DK/Frost Mage party.
  Attack began without ROUTE BLOCKED, companions changed position down the
  entrance staircase, the tank moved out of the entrance view and took aggro,
  and Backstab/Icy Touch/Plague Strike combat events appeared. Experience rose
  from 6,576 to 11,098 as the party defeated trash with the player stationary.
  Recall brought companions back to the entrance stairs. The unattended tank
  eventually died during continued combat; this was not a healed full clear.
- Inline task screenshots captured the unchanged visible steps, equipped
  character bodies, warm lighting, and party movement. Both wing flights have
  collision-backed automated coverage; live end-to-end wing traversal, a full
  dungeon clear, wipe/checkpoint behavior, and physical Android QA remain open.
- Validation: all 224 Vitest files passed, 1,380 tests passed, and six existing
  tests remained skipped. Production TypeScript/Vite compilation passed using
  the project's 8GB heap setting; texture/performance audits, static bundle
  compression, and release-bundle verification passed. Existing build-size
  advisories and deprecated physics initialization warnings remain. No hosted
  release was published.

## 2026-09-12 - Non-blocking players and combat actors

- Removed party soft separation and crowd-triggered combat repositioning.
  Companions may overlap or cross one another without push-apart corrections
  changing their routes. Normal follow formations and role-specific attack
  ranges remain; a tank already in melee range does not move just because an
  ally occupies the same position.
- Added a shared actor collision layer that interacts only with physical
  terrain. The player capsule uses this layer. Player controller movement,
  ground probes, camera sweeps, party direct/route validation, party grounding,
  and combat ground-position queries use terrain-only filters. Actor body type
  cannot turn an actor into a floor or movement obstruction. Party members,
  remote players, mobs, and bosses currently render without physical colliders;
  any future actor capsule must use the shared actor layer.
- Walls, floors, ceilings, dungeon objects, and the Stockades walking ramps
  retain their terrain collision. Target-selection meshes, combat visibility,
  damage, and attack-range checks were not removed.
- Ten new Rapier regressions cover overlapping actors with no contacts or
  displacement, movement through fixed/kinematic/dynamic actors (including an
  overlapped starting position), continued wall/floor collision, and ignoring
  actors in ground, party-route, and camera queries. A terrain-layer control
  confirms real actor/terrain contacts are still detected. Stockades shipping-
  asset regressions now use the live player collision policy and still pass
  down/up all three ramps without jumping.
- Browser smoke pass at 1280 x 720 used offline level-15 Priest Rolecheck with
  Blood DK, Combat Rogue, Unholy DK, and Frost Mage companions. Inline roster
  and dungeon screenshots show character bodies, weapons, unchanged visible
  stairs, and the existing warm lighting. Attack moved companions down from
  the entrance and out of its view without ROUTE BLOCKED; the tank took aggro
  and melee/caster combat events continued. Experience rose from 11,098 to
  13,359. Stop followed by Recall returned the party to the nearby player.
  Selecting Morvane and casting Power Word: Shield applied the buff successfully.
- Live QA was an entrance/trash/recall smoke pass, not a full dungeon clear or
  a manual traversal of both wings. Actor overlap and passage guarantees are
  backed by the physics regressions; physical Android and multiplayer QA remain
  open.
- Validation: all 225 Vitest files passed, 1,390 tests passed, and six existing
  tests remained skipped. The final focused checks passed again after cleanup.
  Production TypeScript/Vite build, texture/performance audits, static bundle
  compression, and release-bundle verification passed. Existing chunk-size
  advisories and deprecated physics initialization warnings remain. No hosted
  release was published.

## 2026-09-12 - Individual companion navigation recovery

- Replaced the old distant, whole-party stuck Recall with an individual
  navigation watchdog in a separately tested game module. Two seconds without
  meaningful waypoint progress invalidates that companion's route; five seconds
  requests a safe recovery landing. Failed replans do not restart the timer,
  back-and-forth movement cancels forward progress, and dense waypoint chains
  still count genuine movement around corners.
- Recovery prefers a recently visited party/player trail point near the group
  on the same floor, then falls back to a validated player position. Actual
  terrain checks require walkable ground, clear companion body space, and a
  route back to the group. Blocked or invalid landings are not applied. Each
  companion has an independent ten-second teleport cooldown.
- Intentional attack-range holds, casting/channeling, crowd control, Stop,
  death, inactive simulation, and UI pauses do not accumulate stuck time.
  Automatic recovery changes only the affected companion's position and routing
  state; it does not heal, resurrect, reset threat, or teleport other companions.
  Manual Recall keeps its existing whole-party behavior and party order.
- Added regressions for recovery timing, oscillation, replanning, independent
  cooldowns and member isolation, exclusions, safe-trail/player selection, wall
  and ceiling rejection, missing floors, and actor-independent landing checks.
  Shipping Stockades navigation/collision assets validate recovery landings on
  all three ramps and along approaches to all five boss objectives.
- Browser smoke pass at 1280 x 720 used offline Priest Rolecheck and Blood DK,
  Combat Rogue, Unholy DK, and Frost Mage companions. Inline roster and dungeon
  screenshots show unchanged character bodies, equipment, stairs, and warm
  lighting. The first catalog load needed a retry. Attack moved companions down
  from the entrance without ROUTE BLOCKED, and combat continued with the player
  stationary. The unattended party eventually died; Stop remained selected
  after Recall returned the fallen companions without resurrecting them.
- The intermittent side-room trap and its automatic recovery were not reproduced
  live. Recovery triggers and terrain safety are backed by automated tests;
  extended side-room playtesting, a healed full clear, and physical Android QA
  remain open. This pass did not modify existing ramp or actor-collision policy.
- Validation: all 227 Vitest files passed, 1,416 tests passed, and six existing
  tests remained skipped. The final 37 focused checks passed after correcting a
  test-helper type annotation. Production TypeScript/Vite compilation, texture
  and performance audits, static compression, and bundle verification passed.
  Existing chunk-size advisories and deprecated physics initialization warnings
  remain. No hosted release was published.

## 2026-09-12 - Wall-aware combat pursuit and individual mob recovery

- Hostile mob actions now require both attack range and terrain line of sight.
  Losing either cancels a pending hostile cast so pursuit can resume; companion
  hostile casts also cancel when sight is lost. Self buffs and friendly heals
  cannot make an otherwise obstructed mob hold an attack position. Proximity
  aggro also checks sight instead of pulling enemies through nearby walls.
- Static enemies, roaming packs, and summoned enemies share separately tested
  imperative pursuit rules over the already-loaded dungeon navigation surface.
  Blocked direct approaches use approved routes through doorways. Routes may
  finish at a reachable attack position only when that position is both in
  range and visible; off-surface connectors still require body clearance.
- Two seconds without meaningful waypoint progress requests a replan. After
  five seconds, a stuck mob within ten meters of its target may recover to a
  validated nearby landing on the same floor and within its home leash. Ground,
  body clearance, route connectivity, attack range, and sight must all pass.
  This teleport preserves health and threat and has a ten-second cooldown.
- Distant, wrong-floor, out-of-leash, or unsafe near-target recovery requests
  instead reset only that mob's combat state and restore its health. A validated
  home landing is preferred; otherwise it evades and navigates home without a
  blind teleport. Ten seconds of evade prevents immediate proximity, damage,
  or taunt re-engagement. Dead enemies are never resurrected. Casting, control,
  pauses, inactive simulation, and non-authoritative online clients do not
  accumulate navigation failures or trigger automatic recovery.
- Regression coverage includes range-plus-sight attack/cast gating, immediate
  cast cancellation, doorway detours, recovery timing and safety, reset/member
  isolation, cooldown bypass protection, and near-teleport health/threat
  preservation. Shipping Stockades meshes independently verify real wall
  occlusion in both directions and clear sight across all three stair ramps.
  Pursuing mobs traverse each actual ramp uphill and downhill without false
  teleport/reset recovery. Existing party and player ramp checks still pass.
- Browser smoke pass at 1280 x 720 used offline level-16 Priest Rolecheck and
  Blood DK, Combat Rogue, Unholy DK, and Frost Mage companions. Power Word:
  Shield applied to Morvane; Attack moved companions down from the entrance
  and out of its view without ROUTE BLOCKED, and Defias melee/cast events and
  tank damage followed. Stop plus Recall returned all four living companions
  to the nearby player while retaining Stop. Inline screenshots show character
  bodies, equipment, unchanged visible steps, and the existing warm lighting.
  The browser console had no errors and the existing physics initialization
  deprecation warning only.
- Live QA was a short entrance/combat/recall smoke test, not a healed full clear
  or a manually reproduced side-room wall trap. Wall-blocked pursuit and mob
  recovery triggers are backed by automated tests; extended side-room combat,
  physical Android, and live multiplayer QA remain open. Existing lighting,
  ramp assets, and non-blocking actor collision policy were not changed.
- Validation: all 229 Vitest files passed, 1,444 tests passed, and six existing
  tests remained skipped. Final production TypeScript/Vite compilation, texture
  and performance audits, static compression, and bundle verification passed.
  Existing chunk-size advisories remain. No hosted release was published.

## 2026-09-12 - HoTs and regeneration cannot heal corpses

- Reproduced the reported death-with-HoT state through lethal damage followed
  by overdue Rejuvenation ticks; the original implementation restored health
  while retaining the death animation. Separate failures confirmed passive
  spirit/equipment regeneration, ground healing, triggered healing procs, and
  a lingering damage tick's level-up refill could also restore corpse health.
- Kept the fix in combat simulation rather than scene movement. Ordinary
  player healing in mutable combat work now uses a shared living-target guard;
  passive regeneration preserves zero health. Timed effects validate their
  target before ticks and procs, discard HoTs on dead targets even before the
  next tick, and retain HoTs on living allies when their caster dies. Newly
  applied actor HoTs also reject dead targets. Ground healing skips corpses
  while continuing to heal living teammates. Experience may still be awarded
  after death, but its level-up refill no longer implicitly resurrects.
- Explicit player and actor-based party resurrection remain available and
  synchronize restored actor health/alive state. Party resurrection also
  clears the death animation using the existing ability-cancel event. Normal
  checkpoint respawn still restores health and clears lingering effects.
  The existing player movement gate remains unchanged and blocks zero health.
- Fifteen new regressions cover the original self-HoT failure, actual Priest
  Renew, overdue/future ticks, explicit companion-to-player HoTs and their
  healing logs, dead allies, regeneration, ground healing, proc healing,
  post-death level rewards, living-target healing, resurrection, and respawn.
  The final focused combat/death/support checks passed 101 tests; the full
  suite passed all 230 files and 1,459 tests with six existing skips.
- No live corpse-movement replay or physical Android check was performed in
  this pass. Evidence is combat-state regression coverage, not a new visual
  claim. Character rendering, camera/input, dungeon assets, lighting, and
  navigation policy were not modified. No hosted release was published.

## 2026-09-12 - Selectable resurrection checkpoints and fallen NPC recovery

- Normal dungeons now offer the entrance plus every defeated boss as named
  resurrection destinations. Choices use authored boss locations, not corpse
  coordinates, and unlock from actual defeated actors or durable encounter
  objective completion. Alive/script-despawned bosses and invalid destination
  IDs cannot grant a checkpoint. Boss-wipe resets retain completed checkpoints;
  a fresh dungeon run clears its progress.
- The player defeat prompt includes a destination chooser. Respawning at the
  entrance or a cleared boss retains defeated enemies and run progress while
  resetting the current pull and restoring the party through the existing
  respawn lifecycle. Legacy callers without a choice retain saved-spawn behavior.
- A contextual `Rez dead members` button and matching destination chooser appear
  below the party frames whenever an NPC teammate is dead. Only fallen NPCs are
  revived, at full health, into their formation slots at the selected checkpoint.
  Living NPCs, the player's health/location, current enemies, and party orders
  are not reset. Lingering timed effects on revived NPCs are removed without
  removing living allies' HoTs. Manastorm charges and remote human resurrection
  behavior are unchanged; non-authority online clients cannot mutate NPC revival.
- Living, engaged bosses disable the NPC chooser/button with an explanation;
  the same restriction is checked again when the action executes. Trash combat
  does not block it. Individual return tasks use approved navigation toward the
  player's current location, including while the party is on Stop, then resume
  the current order. Returning members do not cast/taunt from the checkpoint or
  skew the engaged squad's target-selection center. Existing validated stuck
  recovery remains available. Recall, death, actor unmount, and run reset cancel
  pending return tasks. Scene relocation clears the corpse animation explicitly.
- Live localhost Stockades screenshots at 1280x720 showed the panel with a dead
  Morvane tank, an enabled entrance chooser/button during trash combat, and the
  tank restored to 1707 health at the entrance after clicking while Stop remained
  selected. The button disappeared when no NPC was dead; living Thorne continued
  taking damage rather than receiving a party-wide refill. At 384x844 both native
  chooser and button remained within the viewport, about 176px wide and 44px high.
  The narrow screenshot capture was backend-scaled, so DOM bounds supplemented
  the visual check. Normal viewport sizing was restored and the temporary tab
  and development server were closed.
- Twenty-nine new regressions cover checkpoint unlocking/retention, invalid and
  locked choices, player checkpoint respawn, boss/trash restrictions, selective
  NPC revival/effect cleanup, authority/Manastorm isolation, Stop/rejoin behavior,
  wall-aware return navigation, registry lifecycle, and UI visibility/placement.
  The shipping Stockades navigation/collision test also simulates all four NPC
  formation slots walking from the entrance and every boss checkpoint back to
  the entrance through approved corridors, without requiring stuck recovery.
  Final Vitest validation passed all 234 files and 1,488 tests, with six existing
  skips. No live boss-kill destination-selection, distant-player return replay,
  multiplayer session, or physical Android check is claimed in this pass.
- Browser logs contained no errors; Rapier initialization and KTX2-loader warnings
  occurred during development hot reloads. Character assets, camera/input,
  dungeon geometry, and lighting were not edited. Final production TypeScript/Vite
  compilation, static compression, and release-bundle verification passed; texture
  and environment-performance audits also passed during prebuild. Existing build
  chunk advisories remain. No hosted release was published.

## Optional DPS meter — September 12, 2026

- Added a compact party DPS meter with current/last-fight damage, class-colored
  rankings, and a shared fight timer. It consumes the lossless combat event
  stream and counts effective damage, including pets/summons under their owner.
  Completed fights remain visible until the next pull; encounter resets clear
  the meter. Hiding it does not stop recording. Local menu/background pauses
  do not dilute DPS.
- Automatic placement uses the PC lower-left corner and the Thor companion
  display. The UI editor supports visibility, dragging/keyboard nudges, width,
  and screen selection. These settings participate in saved layouts, all six
  template slots, and export/import, including migration of older UI files.
  Dedicated companion snapshots carry meter data and draft placement; lower
  display movement commands update the authoritative editor draft.
- Live localhost Stockades checks at 1280x720 showed increasing party damage
  and readable bars in the PC lower-left corner. Resizing from 230 to 240 pixels
  worked; dragging moved the handle from approximately (9, 509) to (269, 400).
  Saving visibility off hid the meter, and reopening the editor retained that
  setting while damage continued accumulating. Default position/width and
  visibility were restored after this check.
- In Thor preview, the meter appeared only inside the 620x540 companion surface.
  Populated bars remained clear of the party strip, vital controls, and action
  bar. With the main editor open, dragging on the bottom screen moved its handle
  from approximately (526, 322) to (661, 369) within companion bounds. Those
  preview movement edits were canceled. Screenshots are saved in
  `artifacts/dps-meter/pc-default.png`, `thor-default.png`, and
  `thor-editor-moved.png`. Full-page capture scaled incorrectly in the browser
  backend, so the Thor screenshots use the scrolled normal viewport.
- All 47 targeted tests passed across seven files, including damage accounting,
  ownership, reset/pause behavior, UI migration/persistence, and dedicated
  companion publication/unsubscription. The production build and release
  checks passed. Browser error logs were empty. Physical Thor hardware and a
  multiplayer session were not exercised; the new per-party totals consume
  this client's combat stream. No hosted release was published.

## Player-only revival during ongoing party combat — September 12, 2026

- In normal offline dungeons, a defeated player with living companions and an
  engaged, living enemy can revive at the entrance or an unlocked boss checkpoint
  without resetting the pull or relocating/restoring companions. The player-only
  path advances the relocation revision, not the party or encounter reset signals.
  Recall remains an explicit, optional command. Full-wipe recovery is unchanged;
  Manastorm's shared-charge resurrection remains separate.
- Player recovery restores health/resources, clears corpse animation, casts,
  control, shields, harmful player auras, and player-targeted periodic effects.
  It preserves NPC/enemy actor snapshots, positions, health, targets, casts,
  periodic effects on other actors, and ongoing cooldowns. The existing checkpoint
  chooser is retained; its defeat-prompt copy explains that companions keep fighting.
- Nine dedicated regressions cover trash/boss combat, selected checkpoints,
  unchanged companion/enemy snapshots and Recall revisions, continued simulation,
  periodic-effect isolation, full-wipe fallback, canonical actor health, and invalid
  requests. A prompt-rendering regression checks the optional-Recall explanation.
  Final Vitest validation passed 239 files and 1,511 tests, with six existing skips.
  Test fixtures use a fixed combat clock and respect surviving crowd-control timers.
- Localhost Stockades smoke testing at 1280x720 showed the rendered dungeon,
  party frames, unchanged Recall control, and an active party fight with increasing
  damage and declining tank health. Session screenshots captured the entrance and
  combat HUD. Startup navigation initially timed out and recovered after reload;
  browser error logs were empty after recovery. A live player-death/mid-fight
  checkpoint-revival replay was not achieved, so spatial proof of that exact
  sequence is not claimed. Multiplayer and physical Thor hardware were not tested.
- Production TypeScript/Vite compilation, static compression, release-bundle
  verification, and prebuild texture/performance audits passed. Existing chunk
  advisories remain. No hosted release was published. Character rendering,
  camera/input, dungeon geometry, lighting, and unrelated in-progress party damage
  changes were not edited for this request.

## RoM dungeon entry and level-up crash — September 12, 2026

- Reproduced in the local PC browser with an offline RoM Human Priest. Stockades
  loaded at level 1, then the first level-up unmounted the entire app. The console
  recorded `Unsupported party specialization: rom-champion:Primary` from
  `partyStore.syncLevel` through `PartyBridge`. Initial party creation already
  skipped WoW talents for RoM classes; level synchronization did not. The same
  call also runs during entry when the saved character level replaces the
  previous combat level.
- Level synchronization now retains empty, frozen talent ranks for RoM
  companions, matching party creation. Other classes still receive their
  existing talent plans. Two regressions failed with the original exception
  before the fix, then passed. They cover all ten RoM companion classes, all
  three player roles, levels 1/2/10/70 and downward synchronization, stable
  identities, fallen companions, and living health/resource ratios.
- After reload, the same saved level 2 Priest entered Stockades successfully,
  fought with the RoM party, and reached level 3 without losing the scene. The
  same character then entered Forsaken Abbey and progressed from level 3 to 6.
  At the 74-second check after Attack, the scene and HUD remained present,
  combat continued, and the renderer reported about 60 FPS. No new browser
  errors or context-loss messages appeared. Rapier initialization and multiple
  active KTX2-loader warnings occurred during these development sessions.
- Screenshots and timestamped browser evidence are in
  `artifacts/rom-dungeon-level-sync/`. Screenshot capture retained the browser
  backend's existing scaling issue; DOM progression and renderer counters
  supplemented the rendered dungeon/body/equipment checks. This was an offline
  PC smoke test, not full dungeon completion, multiplayer, or physical Android
  validation. The temporary desktop viewport was restored and test tab closed.
- All 54 focused party, talent-plan, and RoM combat tests passed. The production
  TypeScript/Vite build, prebuild texture/performance audits, compression, and
  release-bundle verification passed with existing chunk advisories. The final
  release-candidate check includes untracked changes. No hosted release was
  published. Existing combat, death/respawn, AI, and HUD work was preserved.

## NPC party damage — September 12, 2026

- Reproduced Garrick, Arms, and Fury dealing zero damage in level 5/20
  sustained fights: Rage starts at zero, companion Rage regeneration is
  disabled, and the ability loop had no regular weapon swings. Priests,
  hunters, and warlocks also dealt zero because they repeatedly refreshed
  their DoTs before the first tick could fire.
- Added independent weapon swings for melee companions and hunters, with
  Rage earned from successful hits. Swings respect range, visibility, line
  of sight, facing, control, casting, rejoining, and party orders. Maintained
  DoTs and debuffs now allow other abilities to run between applications;
  utility taunts/interrupts/control require a useful combat context.
- Self-centered area attacks now validate their radius and hit around the
  caster. Companion direct and periodic amounts use the existing spell
  scaling calculation, including spell base damage and equipment bonuses.
- The new audit runs 30-second fights for all 26 WoW party templates at
  levels 1, 3, 5, 20, and 80. Every template produces positive damage from
  actual abilities, beyond weapon swings, with resources within bounds.
  Separate regressions cover zero-Rage short pulls, swings during the GCD,
  Stop, Thunder Clap targeting, and zero-coefficient spell base damage.
- In an offline PC Stockades run, the saved level 3 Priest commanded Attack
  without player attacks. Garrick dealt 6 damage in the first three seconds
  and 80 damage by approximately 20 seconds; Kael dealt 147, Korra 220, and
  Ash 146 at that check. The party defeated enemies and the player reached
  level 4. The saved party's Garrick is a Tauren; the class/role fix also
  applies to Dwarf Garrick. This was a PC combat smoke test, not physical
  Thor, multiplayer, or a full dungeon completion test.
- `artifacts/npc-damage/stockades-warrior-damage.png` shows the rendered
  dungeon, player body/equipment, and active DPS rows for both warriors.
  A paused accessibility snapshot records totals. No browser errors were
  reported. The temporary browser tab was closed after verification.
- All 310 focused combat, party, death/respawn, and DPS tests passed. The
  production build, texture/performance audits, compression, bundle
  verification, and release candidate check passed. Existing concurrent
  party-level synchronization, death/respawn, catalog, and server changes
  were preserved.

## Dungeon release stages and issue reports — September 12, 2026

- The complete 84-dungeon browser now sorts Full Release, Beta, then Alpha,
  alphabetically within each stage. Stormwind Stockade is the only Beta;
  the other 83 entries are Alpha. WoW/RoM category tabs remain available.
- Dungeon pause menus can submit a nonempty issue comment of up to 500
  characters to the shared SQLite report database. Connected guest players
  can submit reports too. New reports are stamped Submitted by the server;
  comments remain in the form after a failed network request.
- View Reports on the main menu shows all reports, including Closed, by
  default. Status and Dungeon filters combine, and Clear filters restores
  the complete list. GM accounts can select Submitted, In-Progress,
  Testing Fixes, or Closed, use Close report, and reopen closed reports.
  The server rejects non-GM updates and stale revisions.
- Dungeon rows with Submitted, In-Progress, or Testing Fixes reports have
  an orange border and a top-right open-report count. Closing the final
  open report removes both treatments. Counts refresh when opening the
  browser, returning to the app, and every 30 seconds while visible.
- A headless Edge browser exercised an actual Stockades entry, its pause
  report form, a 500-character submission, whitespace rejection, failed
  submission/retry, the shared player report list, all GM status changes,
  combined filters, and badge removal. Keyboard Back and controller filter
  navigation/GM status changes passed. No browser page errors occurred.
- Visually inspected the form, orange dungeon badge, player report list,
  and GM report screens at 1280×800 and 390×844. Long comments wrap and the
  mobile report screen has no horizontal overflow. Evidence and the test
  result are in `playtest-artifacts/dungeon-reports/`. This validates the
  reporting flow; it is not an exhaustive dungeon completion or physical
  Android/Thor playtest.
- All 43 focused report, catalog, store, shell, and display tests passed
  (five existing shelved-mode tests skipped), along with all 61 server tests.
  The 32 party AI/store/sustained-damage checks also passed after preserving
  immutable cast metadata across companion channel updates to resolve three
  concurrent combat type-check errors. A separate, in-progress expanded
  class-damage audit failed its level-1 CoA player case for Witch Doctor
  and Sun Cleric; those combat behaviors are outside the reporting change.
- The final production TypeScript/Vite build, prebuild texture/performance
  audits, compression, and release-bundle verification passed with existing
  chunk advisories. `npm run release:check` passed after a transient Windows
  Git crash, including untracked files. No hosted release was published.

## CoA and RoM class damage audit — September 12, 2026

- Ran 342 sustained, 30-second combat scenarios: all 21 CoA classes as player
  characters and synthetic NPC slots at levels 1, 2, 4, 9, 20, and 80; all 10
  RoM classes in both runtimes at levels 1, 20, and 70; and the actual shipped
  RoM companion templates with their assigned roles and starting resources at
  those three levels. All scenarios passed. CoA expeditions currently use WoW
  companion templates, so the synthetic CoA slots exercise class execution
  without implying that a CoA companion roster has been added.
- Found and fixed companion spacing based only on WoW class names. RoM and
  CoA casters now use their class metadata and learned offensive spell ranges;
  short-range attacks remain reachable. A thrown summon such as Caltrops does
  not force an early Witch Hunter away from its melee attacks. Hunter aliases
  also use ranged weapon stats and projectile presentations for ranged swings.
- Fixed Dawn Blade's melee classification: a five-yard hunter-archetype attack
  previously counted as ranged and could not initiate the player's weapon
  swings needed to generate Rage. The companion AI now rejects unlearned CoA
  tree entries and resolves the rank actually learned.
- Fixed channel execution in both runtimes. Normal amounts are distributed
  across five pulses, while guaranteed periodic trigger payloads fire per
  pulse. Radiant Flame's imported periodic damage now happens during the
  channel. Focused tests verify five damaging pulses and no duplicate amount
  on completion. Conditional proc payloads are not treated as guaranteed hits.
- Corrected the RoM vocation import to use the player's `FixArgTableID = 1`
  record, as selected in the source client's level-up calculation. The previous
  adjacent table made Champion and Warlock stats grow exponentially; a level-20
  Champion weapon hit reached a raw 704 billion damage in the initial audit.
  The corrected Champion instead sustained attacks and damaging abilities for
  the full test. Source record sentinels and all ten classes' growth bounds are
  covered by regression checks; this is a source-table correction, not a balance
  multiplier adjustment.
- Corrected tooltip fallback damage on retaliation auras. Drakkari Barrier
  stays friendly and deals its authored damage through its damage-taken proc.
  The imported first offensive unlocks are Witch Doctor 9, Sun Cleric 4, and
  Primalist 2. The player audit requires offensive ability damage once unlocked;
  these classes have no learned offensive cast at level 1. Their NPC weapon
  attacks deal damage at level 1. This supersedes the preliminary player
  failures mentioned in the reporting notes above.
- Live PC check: created the isolated offline test character Romaudit, entered
  Forsaken Abbey as a healer, and issued Attack. At 26 seconds the meter showed
  Ilyen (Mage) 492 damage, Morgha (Warlock) 370, Veya (Rogue) 217, and Bronn
  (Champion tank) 148. All four NPCs dealt damage; visible enemy effects also
  confirmed Heavy Bash, Electrocution, Weakening Weave Curse, and Shadowstab.
  No player healing was applied during this bounded damage check, and the tank
  eventually died. This was not a dungeon-completion test.
- Visually verified the loaded roster body/staff and in-dungeon player
  body/staff. The DPS meter remained readable at the PC bottom-left default.
  The first attempt used the downloadable shelved preview without its content
  server; the final captures use the full bundled development asset set and
  recorded no browser errors. No physical Ayn Thor device was tested here.
- Evidence: `artifacts/class-damage-audit/report.json`,
  `level-20-damage.csv`, `rom-roster.png`, `rom-dungeon-start.png`,
  `rom-party-damage.png`, and `rom-browser-errors.json`. Training-target totals
  demonstrate execution and sustained damage; they are not class-balance
  rankings or exhaustive coverage of every talent build and dual-class pair.
- All 434 focused combat, party, CoA, RoM, DPS-meter, stat, and avatar tests
  passed, along with the native vocation-import regression. The production
  TypeScript/Vite build and release-bundle verification passed. The final
  release candidate check includes the new tests and generated catalogs.

## RoM dungeon enemy durability — September 13, 2026

- Reproduced Barren Caves' low health with a RoM Priest and the four shipped
  companions. Tested 108 combinations: 12 party seeds at levels 1, 2, 4, 8, 9,
  20, 40, 70, and 80. The old fallback had 94 health at level 1 and 1,832 at
  level 80; kills took 0–1.9 seconds from the first damaging NPC hit.
- Normal RoM dungeons now use a party-health baseline of
  `1000 + 100 * enemyLevel + 8 * enemyLevel * enemyLevel`, with authored
  creature/boss multipliers applied separately. This is explicit Healer Man
  balancing because native RoM creature health tables are not packaged.
  Barren Caves' Anglo retains his 7.5 multiplier and one-level advantage.
- Loaded RoM creatures receive the new health when the character levels up.
  Re-registration preserves wounds as a health percentage and keeps corpses
  dead. Enemy attack damage also rebases on a level change instead of retaining
  its original low-level value. Regression checks cover these transitions.
- With the new health, the 108 isolated party-damage pulls lasted 3–17.5
  seconds. A separate pull exercised actual enemy attacks and Priest healing
  before the enemy died, with all companions surviving. These are opening
  pulls with baseline equipment, rather than an audit of every geared build.
- Live PC check at 1280×720: created the offline Human Priest Barrenaudit and
  entered Barren Caves as Healer. A level-1 Cave Sand Bug had 1,108 health.
  At 10 seconds it still had 374 health, with 734 party damage on the meter.
  The completed first fight showed 14 seconds and 1,108 total damage:
  Morgha 370, Ilyen 369, Maerwen 289, and Cael 80. Urgent Heal then restored
  Maerwen to 407/407 while the next pull was active. No browser errors were
  captured during that bounded combat check.
- Concurrent navigation edits triggered development hot reloads around the
  playtest; the captures above were taken during uninterrupted combat. This
  check did not complete the dungeon or use physical Ayn Thor hardware.
  Restored the original Hudtest roster selection, closed the test tab, and
  stopped the temporary development servers.
- Evidence: `artifacts/rom-dungeon-health/report.json`,
  `barren-caves-health.png`, `barren-caves-combat.png`, `barren-caves-healed.png`,
  `live-combat-dom.txt`, and `browser-errors.json`.
- Validation: all 407 focused combat, party, CoA, RoM, DPS-meter, stat, and
  dungeon-scaling tests passed across 52 files. The final production
  TypeScript/Vite build and release-bundle verification passed.

## 2026-09-13 - WoW dungeon lighting and stair fleet pass

- Applied environment-specific readable lighting to all 54 WoW dungeon entries,
  using the reviewed Stockades ambient/hemisphere fill as the baseline. Stone,
  cavern, fire, arcane, crypt, and outdoor profiles preserve distinct coloration.
  Stockades' existing presentation/three ramps and RuneWaker lighting are unchanged.
- Scanned every shipped WoW collision package, including authored Wailing
  Caverns' actual source assets. Added 347 verified walking proxies across 22
  other dungeons and eight filled navigation portal bridges in Uldaman. Rendered
  steps and source visual/collision/navigation GLBs remain unchanged.
- Added package-identity gates and a compiler-independent shared runtime adapter.
  Collision preparation spatially indexes flights; rejected candidates cannot
  supply geometry needed by accepted proxies. Existing navigation validation,
  content, encounter, and availability gates were not promoted to green.
- Shipping-asset regressions independently pass all 347 new proxies through
  prepared collision and the live NPC segment validator uphill/downhill at the
  center and two interior lateral lanes. Exact portal matching and all existing
  Stockades player/NPC/wall tests pass.
- The game-playtest visual viewer uses shipped meshes and runtime lighting and
  material policies without touching saved character/party state. At 1280x720,
  Uldaman's old/new-fill comparison showed clearer stonework; Wailing Caverns
  retained green, readable rock; Mechanar retained cool arcane interiors;
  Scholomance's visible entrance steps/masonry remained intact and readable.
  Blood Furnace's dark source textures remained visible at its wall-facing
  entrance viewpoint. Durnholde's visual load completed without console errors.
- Zul'Farrak entrance/boss views and Durnholde's exterior boss view showed
  geometry/terrain/viewpoint gaps, recorded for separate investigation. This is
  not exhaustive staircase certification: curved/short/heavily-decimated flights,
  rejected candidates, outer stair edges/landing lips, complete dungeon routes,
  full clears, and physical Android QA still require individual review.
- Evidence/coverage: `dungeon-pipeline/WOW_ENVIRONMENT_REVIEW.md`,
  `src/game/generated/wowDungeonStairs.generated.json`, and screenshots in
  `artifacts/wow-dungeon-environment/`. No hosted release was published.
- Validation: 247 Vitest files passed, 1,599 tests passed, six existing tests
  skipped. Production TypeScript/Vite build, texture/performance audits, bundle
  compression/verification, and release candidate check passed. Existing physics
  initialization and bundle-size/import advisories remain.

## 2026-09-13 - Dungeon completion results and run leaderboards

- Fixed missing RuneWaker completion bookkeeping for the 21 dungeons with
  configured bosses. Barren Caves now awards Anglo's objective on his actual
  death and announces the completed run. Bossless placeholder entries do not
  fabricate a completion.
- Run timing starts when the dungeon is playable, survives wipes, pauses, and
  saved resumes, and freezes at the final-boss objective. Older checkpoints
  without a recorded start time still announce completion but remain unranked.
- The results dialog shows the final-boss time, final-clear and full-clear
  placement, party classes/levels, and defeated bosses. Stay and loot, reopening
  results, and Return to main menu were verified through the live UI. Returning
  clears the completed checkpoint.
- Replaced the dungeon picker's minimap with persistent completion-time
  leaderboards. Final Boss clears includes all qualifying runs; Full dungeon
  clears requires every available boss for the selected difficulty. Expanded
  entries show each party member and boss status. Normal/heroic rankings are
  separate; disconnected results use an explicitly labeled device leaderboard.
- An actual combat integration check fought Barren Caves' level-2 Anglo with
  9,240 health using NPC ability selection, enemy attacks, and Priest healing.
  NPCs dealt 9,239 damage after a one-damage player tag; healing totaled 441.
  The final death awarded completion at 84.1 seconds with all companions alive.
- Manual browser combat confirmed NPC damage and healing on Barren Caves
  trash. The live results check restored the integration fight's completed
  checkpoint through an isolated local QA account: 01:24.1 ranked #1 of three
  seeded local-server entries in both filters. This was not a manual browser
  full clear and no production leaderboard entries were published.
- PC results and expanded final/full leaderboard layouts were visually
  reviewed at 1280x720. The built-in Thor preview displayed the leaderboard,
  filters, and role-selection control within its 960x540 main surface. Compact
  480x360 DOM bounds kept the dialog footer inside the viewport; the in-app
  browser's viewport-override screenshot scaled incorrectly, so that capture
  does not establish compact visual quality. Physical Thor hardware was not used.
- Evidence: `artifacts/dungeon-runs/combat-completion.json`, `results-world.png`,
  `leaderboard-final-expanded.png`, `leaderboard-full-expanded.png`, and
  `leaderboard-thor.png`. The bounded final browser check captured no errors.
- Validation: 70 completion, checkpoint, encounter, catalog, client, and UI
  tests plus the real-combat integration test passed. The 432-test focused
  combat/party/class/health/input regression suite and 18 server tests passed.
  Production build, final TypeScript check, and release candidate check passed.

## 2026-09-13 - RoM character nameplate class labels

- RoM overhead nameplates and AI party frames now show the active class alone,
  or primary/secondary class names joined with `/`, without race or `Primary`.
- An isolated local fixture using the real scene, party frames, and shared
  character preview verified `Warrior`, `Priest`, and `Priest/Warrior` labels.
  The two-class dungeon screenshot showed the label above the rendered body
  and equipped staff silhouettes. The roster preview also retained its body
  and equipment. The fixture used in-memory identities without saving a roster.
- Evidence: `artifacts/rom-nameplates/dungeon-dual-class.png` and
  `artifacts/rom-nameplates/roster-preview.png`.
- Validation: 70 avatar, party-store, and mob-nameplate tests passed. Production
  build, asset audits, compression/bundle verification, and release check passed.

## 2026-09-13 - Dungeon boss durability and difficulty leaderboards

- Campaign bosses now use a level-scaled health floor above their original
  health when needed. Final bosses receive an additional 1.8 multiplier on
  that floor. Larger authored health values remain effective.
- The dungeon browser offers a 1-10 boss difficulty slider, native keyboard
  adjustment, and controller decrement/increment buttons. Difficulty 1 is
  Easy; each step adds 25% health and 15% damage to bosses. The selected source
  Normal/Heroic variant also applies to entry. Trash retains its existing stats.
- Boss difficulty is fixed at entry, preserved through saves and renderer
  restarts, and passed from the online group leader to followers. Changing
  difficulty starts a new run; completed checkpoints also start fresh.
- Device and world leaderboards isolate each difficulty's final/full-clear
  records, ranks, and pages. Existing records migrate to a separate legacy
  bracket. Migration and repeat submissions were verified against SQLite;
  HTTP tests verify difficulty selection and reject invalid values.
- Deterministic Ragefire simulations tested 12 RoM party compositions at levels
  1, 8, 20, 40, and 80 against the shared runtime scaling function. All ordinary
  boss samples exceeded 25 seconds and final-boss samples exceeded 45 seconds.
  Additional damage-role Mage samples passed those bounds; at level 8 the
  ordinary boss took 29.8 seconds and the final boss took 55 seconds. Timings
  remain dependent on party composition, player actions, and equipment.
- A final-boss fight with enemy attacks and Priest healing verified repeated
  attacks, incoming damage, healing, and party survival before the boss died.
  Balance evidence: `artifacts/boss-difficulty/ragefire-balance.json`.
- Desktop screenshots at 1280x720 verified the Easy and difficulty-10 layouts,
  stat preview, and matching leaderboard labels. Pointer and keyboard changes
  were exercised. At 390x844 the slider remained operable and its DOM width
  stayed within the viewport, but the in-app browser's viewport-override
  screenshots scaled incorrectly, as in the earlier compact playtest. Those
  captures do not establish mobile visual quality. The viewport was restored.
- Validation: 169 focused game/client/UI tests and 20 server tests passed.
  The production build, asset audits, bundle verification, and release check
  passed. Existing RoM nameplate changes were retained.

## 2026-09-13 - Original WoW dungeon atlas

- Added 107 unique original map images across 51 standard WoW dungeons using
  the supplied patch-enUS-M archive and existing client archives. A duplicated
  Trial of the Champion floor record is deduplicated before packaging. The
  source MPQs and Mapster addon were read without modification; no Lua addon
  code is installed into HealerMan.
- Coordinate inverses were checked against 10,342 matching source spawns and
  the event-driven Trial of the Champion source entrance. Maximum discrepancy
  is 0.189 yards in authored Wailing Caverns rounding; other imported maps
  have substantially smaller discrepancies. All 107 image checksums match
  their local immutable content objects.
- The production preview at localhost:5174 loaded Wailing Caverns and showed
  original artwork with the player/camera direction, entrance, numbered
  encounters, and a 45.7%, 60.5% entrance-position readout. The map was visually
  reviewed at 1280x720, with the rendered dungeon visible behind it. Pressing
  M opened it; toggling patrol routes then pressing M closed it despite the
  checkbox retaining focus. Reopening restores the clean default overlay.
- Evidence: `artifacts/wow-maps/wailing-caverns-atlas.png`. No production-preview
  errors were captured in the bounded console check. Earlier development HMR
  errors from concurrent equipment edits were not treated as production errors.
  Testing used the existing offline Hudtest QA character, not a live account.
- Multi-floor rendering, floor controls, rebound shortcuts, repeat/capture
  guards, PNG dimensions/checksums, axis projection, shifted/rotated scene
  transforms, and RuneWaker survey fallback are covered by automated tests.
  Every dungeon was not individually traversed in the browser. Auto floor
  selection is horizontal best fit, not WMO-group/height detection; overlapping
  stacked rooms require manual floor selection. Mobile hardware was not tested.
- Tinkertech Showdown, Frozen Reach, and Forgotten Mine have no matching
  artwork records and retain their survey maps. RuneWaker maps are unchanged.
  Failed artwork loads also retain a correctly reprojected survey fallback.
- Validation: 50 focused map/UI/input tests and three BLP/tile-layout tests
  passed. The production build, texture/performance audits, and compressed
  release-bundle verification passed. Local downloadable packs were published;
  no online deployment was performed.

## 2026-09-13 - Boss card collection and leveling loot

- Boss deaths roll a guaranteed Normal card plus independent Rare (1/10),
  Epic (1/50), and Legendary (1/100) bonuses. The collection catalogue covers
  330 distinct WoW/RoM bosses and 1,320 rarity variants using existing models.
  Repeated appearances share counts and list every source dungeon.
- Copies and the latest claim per boss spawn persist per character through the
  roster/cloud-save path. Shared-world hydration grants the authority's rolls
  to followers and ignores duplicate/reconnected and older reward snapshots.
- Equipment drops and bonus equipment are suppressed below player level 80;
  existing inventory and authored non-equipment drops are preserved. Level-80
  equipment loot remains enabled. Bosses without equipment tables award cards.
- The production browser at 1280x720 verified the main-menu entry, all four
  Lady Anacondra model portraits, dungeon location, odds, search, All/Earned,
  dungeon/rarity grouping, and in-dungeon pause-menu access. Roster and dungeon
  screenshots show player/party bodies and starter weapon/shield silhouettes.
- At 390x844 the collection switched to two columns, with no horizontal page
  overflow and dialog bounds within the viewport. The in-app browser again
  scaled the viewport-override screenshot incorrectly; that capture does not
  establish mobile visual quality. The viewport was restored.
- Deterministic tests cover exact odds boundaries, counts, persistence,
  no-loot-table bosses, loot restrictions, and online replay. Live browser QA
  did not perform a boss clear; earned-copy display was checked with seeded
  render tests. Evidence: `artifacts/boss-cards/collection-desktop.png`,
  `collection-mobile.png`, `roster.png`, and `dungeon.png`.

## WoW and CoA default boss difficulty review — September 13, 2026

- Reviewed all 10 WoW and 21 CoA classes: 61 canonical class-role builds at
  levels 8, 20, 40, and 80 with three deterministic companion compositions.
  The 1,464 durability simulations used live boss scaling, starter heirlooms,
  legal level-appropriate talents, live companion AI, and the existing baseline
  player rotation. CoA characters currently use WoW companions.
- Fixed Death Pact and other instant-kill effects bypassing boss health in the
  player, triggered-spell, and companion paths. Fixed repeated proc activations
  across branches, duplicate execution of expanded CoA payloads, invented
  damage on proc auras and Waterfall, and critical-damage talent bonuses
  affecting ordinary hits. Regenerated the CoA runtime catalog and shards.
- Raised the shared health floor by 40% to cover the remaining CoA burst.
  Every class uses the same boss stats at the same level and difficulty.
  Fastest sampled kills were 26.2 seconds for an ordinary boss and 48.6 for
  the final boss. All duration assertions passed. The refreshed RoM regression
  also passed; its level-eight damage-role Mage sample took 41.8/74.9 seconds.
- Durability samples disabled enemy attacks and stopped at 60 seconds to
  isolate damage. Most WoW and CoA samples still had a living boss at the cap;
  these are lower bounds, not proof of typical 25-30/45-second kill times.
  Default class throughput is still uneven, especially for WoW parties.
- An additional 26 final-boss attack samples exercised healer builds at levels
  8 and 80. Paladin, Priest, Shaman, and Tinker kept their parties alive in both.
  Druid and several CoA samples lost companions; Bloodmage also lost the player
  at 80. This basic policy can refresh HoTs too early or choose self-healing
  hybrids, so these outcomes need a dedicated healer rotation/playtest pass
  before treating them as class throughput limits. No live boss clear was
  performed in this numerical review.
- Validation: 122 combat/catalog checks, 91 class-damage/avatar checks, eight
  RoM boss regressions, TypeScript, and the production build passed. All six
  large-matrix assertions passed, but its progress reporter timed out during
  CPU-heavy loops. Added worker yields and reran the affected level-40 test;
  it exited cleanly without the reporter error.
- Evidence: `artifacts/boss-difficulty/wow-coa-balance.json` and
  `artifacts/boss-difficulty/wow-coa-review.md`. Coverage is a canonical build
  per role with starter gear, rather than every specialization or later gear
  combination. `npm run release:check` is required for handoff.

## CoA healer rotation pass — September 13, 2026

- Corrected base-mana percentage costs, Tinker's secondary mana maximum,
  imported heal power/AP/spirit scaling, self-healing riders, and party heal
  recipients. Removed invented instant heals from conditional procs and
  invented hostile damage from healing/mitigation tooltip text. Regenerated
  the CoA runtime catalog and all class shards.
- Companion support AI maintains existing HoTs, covers other injured allies,
  and avoids redundant shields. The benchmark player policy now prefers
  effective emergency/area healing, avoids wasteful self-healing, positions
  ground heals on the injured ally, and funds rage heals through weapon attacks
  and actual resource generators. Normal gameplay still uses player input.
- Tested nine canonical CoA healer builds at levels 8, 20, 40, and 80 against
  final-boss attacks for 60 seconds with three deterministic WoW companion
  compositions each: 108 samples. Every player survived and every sample
  delivered effective ally healing. Entire-party survival was 24/27 at eight,
  25/27 at twenty, and 27/27 at both forty and eighty.
- The remaining low-level losses are Primalist companions. Its imported Geode
  Barrage lacks the server-scripted stone/rage payload; that mechanic still
  needs implementation. Survival at the time cap does not establish a boss
  clear, indefinite mana sustainability, or complete CoA mechanic parity.
- Reran the full WoW/CoA durability and healer-attack suite: all six assertions
  passed without reporter errors. Fastest sampled ordinary/final kills remain
  26.2/48.6 seconds. WoW Druid's comparable eight/eighty attack samples now keep
  the party alive with the revised HoT policy.
- Validation: 167 combat/catalog/runtime regressions, four healer matrix
  assertions, 47 avatar checks, TypeScript, the production build, and release
  bundle validation passed. The final release hygiene check covers untracked
  regression tests as well as tracked edits.
- Live production-preview smoke check at 1280x720 used an offline Human Witch
  Doctor (Brewing). Roster and Wailing Caverns entrance captures show the
  player and companion bodies and starter weapon/shield silhouettes. This
  visual check did not perform a live boss clear or validate manual rotations.
  Evidence: `artifacts/boss-difficulty/coa-healer-smoke/roster.png` and
  `artifacts/boss-difficulty/coa-healer-smoke/dungeon.png`.
- Numerical evidence and remaining limitations:
  `artifacts/boss-difficulty/coa-healer-balance.json`,
  `artifacts/boss-difficulty/wow-coa-balance.json`, and
  `artifacts/boss-difficulty/coa-healer-review.md`.

## Primalist rage generation follow-up — September 13, 2026

- Restored Geode Barrage's three physical stone pulses at cast completion,
  +500ms, and +1000ms, using the ranked parent amount, 23% nature/spell power,
  and 9% attack power. Each pulse dealing effective damage grants its caster
  5 rage, for 15 from a full Barrage. Removed the unscheduled hidden placeholder
  proc so another periodic event cannot substitute for or duplicate the stones.
- The 5-rage stone amount is HealerMan's normalization using the related
  imported Geode spell 804002 payload. Barrage's imported DBC does not expose
  an exact rage amount; this does not establish exact server-script parity.
  Dead targets and fully absorbed stones grant no rage. Player and companion
  grants cap at the source pool maximum and synchronize the actor's pools.
- Seismic Crash grants 20 rage once per cast and executes its initial damage
  pulse. Seismic Spike grants 4 per target actually damaged. Primalist energize
  effects retain their power type and convert internal rage units to displayed
  points: Primal Rush and passive Protective Roar grant 20 rather than 200.
  Protective Roar retains beneficial party recipients when its payload fires.
- Scoped imported power-unit conversion to Primalist. A global conversion
  caused a Bloodmage survival regression; other classes retain their existing
  script-payload interpretation pending their own power-record review.
- The benchmark policy considers all usable rage generators, including Crash
  and Barrage, before rage heals. Normal gameplay still uses player input.
  All 108 attacking-final-boss CoA healer samples now keep the entire party
  alive, including the five previously failing low-level Primalist samples.
  The regression requires zero companion deaths at every tested level.
- Validation: 207 combat/catalog/runtime/avatar checks passed, including eight
  focused Primalist checks for timing, coefficients, capped gains, absorbed and
  dead targets, companion attribution, passive payloads, and Seismic gains.
  All four strict healer matrix assertions passed. All six WoW/CoA boss suite
  assertions passed across 1,464 durability runs and 26 attacking-boss samples;
  fastest sampled ordinary/final kills remain 26.2/48.6 seconds. TypeScript,
  the production build, release bundle validation, and release hygiene passed.
- Live production-preview smoke check at 1280x720 used an offline female
  Tauren Primalist (Life). Roster and Wailing Caverns entrance captures show
  a character body with a staff and companion bodies with weapon silhouettes.
  The Tauren identity rendered a human-looking fallback; the optional content
  service was unavailable in this static preview. These captures establish
  body/equipment visibility, not correct Tauren appearance. Entrance trash
  reached level two with the full party alive. Barrage remained locked until
  level six, so pulse/rage behavior was checked by deterministic runtime tests,
  rather than a live manual Barrage cast. No live boss clear was performed.
- Evidence: `artifacts/boss-difficulty/primalist-smoke/roster.png`,
  `artifacts/boss-difficulty/primalist-smoke/dungeon.png`,
  `artifacts/boss-difficulty/coa-healer-balance.json`,
  `artifacts/boss-difficulty/wow-coa-balance.json`, and
  `artifacts/boss-difficulty/coa-healer-review.md`. Survival at the 60-second
  cap does not establish a boss clear or indefinite resource sustainability.

## WoW action bars in Ability controls — September 13, 2026

- Added Controller bars / WoW bars selection to Ability controls. On PC, the
  dialog opens on WoW bars when that UI layout is enabled. The existing three
  controller layers remain available through the same screen.
- The WoW editor exposes all five bars and twelve buttons per bar, including
  inactive bars. Slots show their saved keyboard shortcuts and assigned
  abilities. Assigning or leaving a slot empty saves immediately through the
  existing profile UI persistence path and updates the gameplay action bar.
- Restore Bar resets that bar's abilities to the current character's class
  defaults; Restore all WoW bars resets all sixty ability slots. Both use a
  confirmation with cancellation. Keyboard shortcuts, bar geometry, PC/mobile
  UI templates, and controller bindings remain independent of these resets.
- Long WoW ability lists scroll inside the picker. Controller navigation keeps
  the selected control visible. Future abilities retain their unlock-level
  labels and can be assigned ahead of time; passive abilities are excluded.
- Live 1280x720 production-preview QA used an isolated offline Human Priest
  profile at port 5192. Verified first-slot assignment, inactive fifth-bar
  button-twelve assignment and clearing, canceling a reset, restoring one bar
  while retaining the fifth-bar assignment, and restoring all WoW bars.
  Controller Cross retained Smite through the WoW edits and resets.
- Enabled the WoW layout through Edit UI & templates, confirmed Ability controls
  opened directly on WoW bars, and changed the first button to Lesser Heal.
  The gameplay bar updated immediately. After a full page reload and dungeon
  reentry, both the gameplay bar and editor retained Lesser Heal and key 1.
  The final picker capture shows access to the late-level abilities and Leave
  empty while all bar controls stay visible. No browser UI errors were logged.
- Validation: all 55 UI-layout, action-binding, keybind, profile-sync, input,
  HUD-layering, and panel-integration tests passed. The production build,
  including TypeScript with the project's configured 8 GB heap, prebuild asset
  audits, and postbuild release-bundle validation, passed. A separate manual
  TypeScript attempt using Node's default heap exhausted memory; the standard
  build completed successfully. Final release hygiene covers all changed paths.
- Evidence: `artifacts/ability-controls/wow-bars.png`,
  `artifacts/ability-controls/controller-bars.png`,
  `artifacts/ability-controls/wow-picker.png`, and
  `artifacts/ability-controls/wow-picker-end.png`. Cloud upload and physical
  gamepad input were not exercised by this offline browser pass.

## Account-wide boss card collections — 2026-09-14

- Moved boss card counts and reward receipts out of character profiles into one
  collection per account owner. Gameplay starts with that account collection;
  menu and in-game collection screens show the same account totals.
- Migrates existing character logs by summing their rarity counts, retaining the
  newest reward receipts, and removing the old character fields. Reloading or
  downloading an old cloud roster again does not repeat the migration.
- Cloud uploads now include the account collection, including accounts with an
  empty roster. Migrated cloud logs upload without waiting for another boss kill.
  Higher cached counts and newer receipts survive an older cloud download.
- Automated integration checks verified earning on a second character,
  switching back, account sign-out isolation, character deletion, and restoration
  into a separate device cache. Collection markup identifies the account scope.
- Validation: 164 tests passed across 19 files, with six shelved-mode tests
  skipped. This persistence change did not alter character models or equipment.
- TypeScript, prebuild asset audits, and release hygiene passed. The first
  shared-output bundle check encountered a missing staged asset; an isolated
  production build and complete bundle check passed with all 5,529 objects.
  Validation output is under `artifacts/boss-cards/account-build`.

## Main-menu talents and controls — 2026-09-14

- Added Talents, Keybinds, and Button assignments destinations using the existing
  editors. RoM characters show their automatically ranked Skills instead of
  a talent-point editor. Closing a panel or using Back returns to the main menu.
- A menu-only character subscription initializes the selected character and
  saves talent ranks and controller ability assignments without mounting the
  dungeon simulation. Level, experience, inventory, and equipment are preserved.
  Keyboard bindings and WoW bar assignments retain the existing profile saving
  behavior; keybind drafts require Save changes.
- Live offline browser QA at 1280x720 opened all three editors for the level-17
  Priest Rolecheck. The talent tree showed eight available points, controls
  showed the character's skill names, and the assignment picker showed current
  Smite and future unlock levels. Canceling a keybind edit and using Done returned
  correctly; Escape first canceled the ability picker and then closed the editor.
  This browser pass did not change the character's build or saved controls.
- At a 390x844 viewport, the talent dialog's measured bounds fit the screen,
  with no horizontal content overflow and a visible main-menu Back button.
  Browser screenshot scaling was unreliable at this override; visual evidence
  below covers the default desktop viewport. Physical gamepads and cloud saves
  were not exercised.
- Persistence tests cover allocation, reset, assignment, reopening, storage
  failure, and owner/selection/phase/close isolation. UI and profile regressions
  passed. TypeScript, production bundling, and complete release-bundle validation
  passed after retrying a shared-output staging race; release hygiene passed.
- Evidence: `playtest-artifacts/main-menu-editors/main-menu.png`,
  `playtest-artifacts/main-menu-editors/talents.png`,
  `playtest-artifacts/main-menu-editors/keybinds.png`, and
  `playtest-artifacts/main-menu-editors/button-assignments.png`.

## AYN Thor physical-size UI review — 2026-09-14

- Browser profiles use the Thor's 6-inch 1920x1080 upper panel and 3.92-inch
  1240x1080 lower panel. Panel sizes come from
  [AYN's specifications](https://www.ayntec.com/products/ayn-thor); resolutions
  are also documented in the
  [Thor manual](https://ae-pic-a1.aliexpress-media.com/kf/S21d979b6c8d840ddaebfeb976094e0cfP.pdf).
  Derived physical rectangles are 132.8x74.7 mm and 75.1x65.4 mm. At a common
  160 CSS pixels per inch, browser viewports are 837x471 and 473x412, with
  independent device scale factors. A 44-pixel control represents about 7 mm
  on either panel. The combined preview uses these relative physical sizes.
- Dungeon defaults put party/self health, targets, minimap, and encounter
  information upstairs. Controller actions, cast bar, buffs/debuffs, experience,
  and DPS meter appear downstairs. Shared HUD components render once on their
  assigned panel; the old duplicate tactical party-health panel is replaced.
- Every HUD element has Top / Bottom / Hidden in the mobile UI editor. Touch
  dragging from the lower page updates the upper page's authoritative draft.
  Selection raises the widget above overlapping controls. Save, cancel,
  reset, rapid reopening, and a full reload/dungeon reentry were verified.
  Legacy templates without a screen field retain their defaults. Single-screen
  play keeps visible elements on its primary display.
- Enlarged text, icons, and touch controls; corrected roster, creator, class
  hall, downloads, and dungeon/role layouts for the actual panel dimensions.
  Character creation uses the lower model preview. Editor controls scroll with
  Save and Cancel kept visible. Buff tooltips and the completion banner use
  readable text; a lower-screen completion banner wraps its buttons below it.
- Edge browser QA uses two Android/touch-emulated pages in one isolated offline
  context. BroadcastChannel carries actual snapshots and touch commands between
  pages. Covered login, empty/populated roster, both creation steps, main menu,
  class hall, menu talents/keybinds/button assignments, reports, downloads,
  dungeon/role selection, gameplay, pause, map, party, boss guide, spellbook,
  talents, ability controls, inventory, boss cards, options, keybinds, issue
  reporting, UI editor, defeat/respawn, completion/results, and combined preview.
- Verified lower action casting changes the upper combat cooldown; party frames
  can move downstairs, hide, save, reload, and select the player remotely.
  Buff inspection opens by touch. Results open automatically and can reopen
  from the banner; finishing through the lower banner returns both pages to
  the main menu. Completion uses an isolated presentation fixture, rather than
  a live boss clear. Dungeon captures include equipped fallback bodies; this
  review does not establish authentic race-model loading.
- All 42 layout, keybind, HUD-layering, display, snapshot, and dual-screen
  regressions passed. A complete production build with asset audits and release
  bundle validation passed; final UI edits also passed production compilation.
  Browser captures have no unscrollable controls clipped outside their panel
  and no uncaught UI errors. Physical hardware, Android WebView installation,
  live online grouping, and shelved game modes were not exercised.
- Reproduce with `node scripts/playtest-thor-ui.cjs` while Vite runs at port
  5176, or set `THOR_TEST_URL`. Evidence is in
  `playtest-artifacts/thor-ui/results.json` and paired PNGs, including
  `14-final-defaults`, `14d-results`, `14e-complete-bottom`, and
  `15-combined-preview.png`.

## Boss Testing (2026-09-14)

- Main menu now opens a searchable boss catalog with game and readiness
  filters, selected-boss fights, and random fights. Selecting a boss installs
  its missing dungeon content pack and dependencies automatically.
- Testing uses an empty procedural room with one initial boss. Existing
  combat, encounter phases, summoned adds, and relevant mechanic objects are
  retained. Tests grant no experience, loot, or boss cards, and do not persist
  changes to the character's progression or inventory.
- Pause and victory review offer Ready, Needs testing, and Not ready labels,
  boss-specific reports, retry, and return to the main menu. Labels persist
  through the server, with local pending saves retained while offline.
- Regression checks passed: 98 focused client tests, 19 server tests, and a
  complete production build with asset audits and release bundle validation.
- Production Edge playtest passed selection, active Lady Anacondra mechanics,
  all three readiness labels, report submission, retry, victory review,
  persistence, random selection, and a 390-pixel mobile layout without
  horizontal overflow. Victory uses deterministic damage through the combat
  store; this is not a natural full-fight clear. Experience and saved character
  data were unchanged, and no uncaught browser errors occurred.
- Selecting Taragaman the Hungerer from an uninstalled Ragefire Chasm pack
  downloaded the pack and dependencies, then loaded the boss model and active
  mechanics in the test room. A second browser page received the testing room
  snapshot and opened boss review through the lower-display button. Physical
  dual-screen hardware was not exercised.
- Reproduce with `node scripts/playtest-boss-testing.cjs`; the script starts
  an isolated production server and test database. Evidence is in
  `playtest-artifacts/boss-testing/results.json` and screenshots
  `01-main-menu.png` through `12-bottom-display-review.png`.

### Boss Testing health correction

- Testing rooms now use the boss's authored health at the scaled enemy level,
  without the campaign durability floor or final-boss bonus. A lone testing
  objective previously made every selected boss qualify as a campaign finale.
  At character level 1 this produced a 10,987 HP minimum instead of Celebras
  the Cursed's 575 HP from his level-2 native stats and authored multiplier.
- Regression checks passed: 24 tests covering testing, campaign boss balance,
  level scaling, asset preparation, and display snapshots. A full production
  build passed its asset audits and release bundle validation.
- Production Edge playtest confirmed a level-1 WoW Priest fighting a level-2
  Celebras with 575 maximum health. The boss model and active mechanics were
  visible, and readiness labels, reports, retries, victory review, random
  selection, mobile layout, automatic downloads, and lower-display review
  passed without uncaught errors. Victory still uses deterministic damage;
  natural time to defeat was not measured.
- Reproduce in PowerShell with `$env:BOSS_TEST_BOSS = 'Celebras the Cursed'`
  followed by `node scripts/playtest-boss-testing.cjs`. Evidence is in
  `playtest-artifacts/boss-testing-scaling/results.json` and paired screenshots.

### Regular dungeon health correction

- Regular dungeon scaling now selects the durability minimum using the
  character's game. WoW uses an interpolated level curve, RoM uses 1.15 times
  the expanded-class baseline, and CoA uses 1.75 times that baseline. These
  independently calibrated floors keep every canonical class-role build above
  the same 25-second ordinary and 45-second final-boss mechanic windows. Native
  authored health can still exceed the floor. Final-boss and difficulty bonuses
  remain part of campaign scaling. Boss Testing continues using scaled authored
  health without a campaign minimum.
- On Easy, a level-3 WoW Priest now encounters a level-4 Targorr the Dread with
  1,300 HP in regular Stockades, compared with the prior shared 10,766 HP
  minimum. The WoW final boss has a 2,340 HP minimum at that level; equivalent
  RoM and CoA parties receive 7,074.8 and 10,766 HP ordinary-boss floors.
- Level-3 WoW Priest simulations with three companion seeds cleared the opening
  boss in 25.7-43.6 seconds and the final boss in 48.2-79.6 seconds. These timing
  samples assume companions are in range and omit boss attacks; a separate
  level-3 test verifies real boss attacks execute during the shortened fight.
- Validation passed for all 271 application test files: 1,759 tests passed and
  six were skipped. The shared balance cohort covers every WoW, RoM, and CoA
  class-role build at levels 8, 20, 40, and 80 with three deterministic seeds.
- The earlier production Edge capture recorded the intermediate 1,200 HP WoW
  floor and is superseded by the final 1,300 HP calibration. The browser script
  now checks the final value; a new live screenshot has not been recorded.
- Reproduce with `node scripts/playtest-dungeon-health.cjs`. Browser evidence is
  in `playtest-artifacts/dungeon-health-scaling/results.json` and
  `stockades-level-three.png`. Run the scaling tests with `BOSS_BALANCE_REPORT=1`
  to record companion timing samples in `artifacts/boss-difficulty/ragefire-balance.json`;
  the all-class report is `artifacts/boss-difficulty/wow-rom-coa-balance.json`.

## Boss spell-effect repair (2026-09-15)

- The dungeon compiler now reads source DBC effect/aura payloads, durations,
  radii, summon creature entries, and targets instead of treating unknown boss
  casts as ordinary damage. The repaired set covers control, damage-plus-control,
  summons, summon-plus-damage, drains, defensive absorbs/reflection/immunity,
  armor debuffs, persistent vampirism, and teleport-to-caster movement.
- The source review removed three controller/helper casts from boss loadouts:
  Noxxion helper `21708`, Gortok stasis-orb awaken `47669`, and Drakkari
  Elemental reset visual `54888`. The new audit reports 427 explicit boss-effect
  casts, 424 cast warnings, 61 ground telegraphs, 427 release/impact cues, 332
  aura/control HUD indicators, and no high-confidence semantic issues.
- The media package supplies original source visuals for 375 effect casts and
  source audio for 377. The remaining 52 visual and 50 audio cases retain the
  generated school VFX and new short spatial school cue. Media packaging
  completed with 5,066 of 5,066 referenced phase assets and no failures.
- Focused validation passed 60 gameplay/audio/catalog tests and all 20 effect
  packaging tests. The production build, texture/performance audits, and
  release-bundle verification passed. The complete application suite passed
  after refreshing the original-model coverage floor for the expanded package.
- Production Edge Boss Testing passed with The Black Stalker. The screenshot
  shows the live boss effect, the Mechanics panel including Summon Player, the
  party/buff/debuff indicators, and the isolated testing room. Report/retry,
  readiness, victory, automatic content download, mobile selection, and the
  lower display also completed with no browser runtime errors; XP and the saved
  profile remained unchanged.
- Reproduce the data/media audit with `npm run dungeon:boss-effects:audit` and
  the browser pass with `BOSS_TEST_BOSS="The Black Stalker"` plus
  `BOSS_TEST_ARTIFACT_DIR="playtest-artifacts/boss-spell-effects-black-stalker"`
  before `node scripts/playtest-boss-testing.cjs`. Evidence is in
  `playtest-artifacts/boss-spell-effects-black-stalker/results.json` and
  `03-empty-room.png`. Headless automation verified audio node selection and
  fallback behavior but did not perform a subjective loudspeaker/listening test.

## Boss Testing random follow-up (2026-09-16)

- The Boss Testing review shown after victory now includes **Fight another
  random boss**. It excludes the boss just completed, preserves the active party
  role, downloads missing content through the existing verified pack flow, and
  enters the new testing room directly. Review controls remain disabled while
  the next boss is being prepared.
- All 15 focused Boss Testing, preparation, review, and display-snapshot tests
  passed. The production build, asset audits, and release-bundle validation
  passed.
- Production Edge defeated Lady Anacondra, displayed the new victory action,
  selected Varos Cloudstrider, and entered his active testing room without
  returning to the menu. No browser runtime errors occurred. Evidence is in
  `playtest-artifacts/boss-testing/06-victory-review.png`,
  `06b-another-random-boss.png`, and `results.json`.

## Eck the Ferocious visibility and healer pressure (2026-09-16)

- Eck's Oracle GLB used a bind-pose-derived `groundOffset` of `11.859`, which
  lifted the animated body out of the gameplay camera while the encounter ring
  remained visible. The Gundrak adapter now keeps the original model but uses a
  reviewed zero ground offset and a seven-unit nameplate height.
- Eck no longer repeats Spring Init and Berserk as 135% damage attacks. The
  runtime now follows `boss_eck.cpp`: Bite opens at 5 seconds and repeats every
  8–12 seconds, Spit opens in 10–37 seconds and repeats every 8–12 seconds,
  Spring targets one random party member every 10–24 seconds, and Berserk is a
  one-time non-damaging aura at 90 seconds. The original spell IDs remain bound
  so release/impact visual and sound presentation plus mechanic indicators are
  preserved; Spring also retains a target-circle warning.
- The reusable `npm run dungeon:boss-pressure:audit` command compares all WoW
  bosses at level 80 over 60 seconds and writes the complete ranking plus
  per-attack contribution to `dungeon-pipeline/WOW_BOSS_PRESSURE_REPORT.json`
  and a top-30 summary to the adjacent Markdown report.
- Production Edge rendered Eck, his nameplate, health bar, mechanic timers, and
  target marker with no browser errors. Evidence is in
  `playtest-artifacts/eck-final-visual/03-empty-room.png` and `results.json`.

## Balance Lab run history and analytics (2026-09-16)

- Server Balance Lab jobs now write private `job.json` metadata, restore the
  newest 100 runs after a restart, import completed pre-history reports, and
  mark interrupted work failed instead of losing it from the GM screen.
- The selected completed run now has five in-app graphs for boss-to-party
  pressure, party-to-boss damage, kill time, class/spec boss damage, and
  class/spec damage taken. Boss, class/spec, and party tabs retain every row and
  add source, role, search, metric, and direction filtering plus wipe, defeat,
  death, healing, absorb, and tuning-outlier context.
- Seven focused server job tests passed, including restart recovery, legacy
  import, and interrupted-run behavior; the combined Balance Lab/server set
  passed nine tests. The 8 GB TypeScript check and production Vite build passed.
- Production Edge automation submitted a real Eck simulation through the GM
  UI, opened the generated JSON in the dashboard, exercised all tables and
  filters, and verified the 480×900 layout with no browser errors. Evidence is
  under `playtest-artifacts/balance-lab/`, especially
  `04-run-analytics.png` and `05-mobile-analytics.png`.

## Balance Lab class/spec ability breakdown (2026-09-16)

- Each simulated member now exports the full learned, level-appropriate,
  executable ability list in addition to observed combat events. Suite reports
  aggregate every class/spec ability with active-run coverage, usage rate,
  casts, hits, casts/minute, raw/effective damage, DPS, damage share, damage per
  hit, healing, overheal, absorbs, AI categories, and cooldowns. Learned
  abilities that produce no events remain visible as `available · unused`.
- The GM dashboard adds an **Abilities** tab with class/spec, source, role,
  usage-state, free-text, metric, and direction filters; an unused-damage view;
  a damage-contribution graph; a complete horizontally scrollable table; and a
  downloadable `class_abilities_across_parties.csv` artifact.
- Six Python aggregation tests, nine server tests, the 8 GB TypeScript check,
  and the production Vite build passed. Production Edge automation completed a
  real Eck run, verified 11 downloadable suite artifacts, exercised the unused
  ability filter, and passed desktop plus 480×900 overflow checks with no
  browser errors. Evidence is in
  `playtest-artifacts/balance-lab/04b-ability-breakdown.png` and
  `playtest-artifacts/balance-lab/06-mobile-abilities.png`.

## Balance Lab server Python preflight (2026-09-16)

- The production failure was a deployment mismatch: the TrueNAS app used the
  canonical Node-only startup command even though server Balance Lab jobs need
  the Python controller. The canonical YAML in both `DEPLOYMENT.md` and the
  release-runbook appendix now installs `python3`, prints its version during
  startup, sets `BALANCE_LAB_PYTHON=python3`, and includes the queue limits.
- A GM-authenticated runtime endpoint now probes the configured executable with
  `--version`. The Balance Lab setup shows the resolved Python 3 version, and a
  missing runtime disables **Run on server**. POST requests independently
  repeat the preflight and return a dedicated 503 without creating a failed job.
- Nine focused Balance Lab service tests plus two server integration tests, the
  8 GB TypeScript check, and the production Vite build passed. Production Edge
  displayed `Server runtime ready · Python 3.14.0`, submitted a real Eck run,
  generated all reports, and completed desktop/mobile QA without browser errors.
  Evidence is in `playtest-artifacts/balance-lab/02-server-job-config.png`.

## Party melee animation cadence (2026-09-16)

- Party character animation now follows authoritative combat-presentation
  phases instead of treating every `nextActionAt` retry timestamp as an attack.
  Auto-attacks produce one ordinary attack event, while real abilities retain
  their cast-start, release, and cancel identities.
- Warrior and Paladin auto-attacks now prefer the normal one-handed/two-handed
  attack clips before special clips. Special/spinning clips remain available to
  explicitly matched abilities rather than every weapon swing.
- The focused combat-animation suite passed 10 tests, all 47 avatar tests
  passed, and the production build plus texture/performance and release-bundle
  audits passed. The live level-three Stockades pass completed without browser
  errors; all five bodies and their equipped weapon/staff silhouettes rendered
  intact. Evidence is in
  `playtest-artifacts/dungeon-health-scaling/stockades-level-three.png`.

## Balance Lab healing done and received analytics (2026-09-17)

- Raw Balance Lab member metrics now attribute effective healing to both the
  caster and the party target. Party and suite reports expose healing done and
  received per active second plus absolute totals; self-heals count on both
  sides of the attribution.
- The GM overview adds separate **Healing done by class/spec** and **Healing
  received by class/spec** graphs. The class/spec table can display and sort
  both rates and totals. Older saved reports remain readable but show no target
  healing until they are rerun.
- All six Python controller tests and the production build passed. A live
  one-seed, 15-second Eck the Ferocious run recorded the Holy Paladin's 867
  effective healing as 57.8 HPS done and the Blood Death Knight's same 867 as
  57.8 HPS received. A matching automatic-suite run verified the new JSON and
  CSV fields under `playtest-artifacts/balance-lab-healing-suite/`.

## Balance Lab deterministic class/spec coverage (2026-09-17)

- Automatic sweeps now offer **Random parties**, **All specs**, and **Specific
  specs**. All-spec mode derives the minimum valid 1-tank/1-healer/3-damage
  party set that covers every available build. Specific mode guarantees every
  selected build appears and uses distinct compatible fillers for missing
  roles; selected builds are not reused as fillers when alternatives exist.
- A real level-20 WoW all-spec sweep against Eck the Ferocious generated eight
  distinct parties and placed all 31 catalog builds in `suite-report.json`
  with zero missing. The combined catalog audit found 70 CoA builds requiring
  16 coverage parties and 16 RoM builds requiring four; every source exposes
  enough tank, healer, and damage builds for valid coverage parties.
- Eight Python controller tests, nine Balance Lab service tests, two server
  integration tests, and the production build passed. Production Edge
  automation verified the All Specs estimate, selected Paladin Holy through the
  Specific Specs picker, completed the server job, rendered all seven overview
  graphs, and passed desktop plus 480×900 overflow checks without browser
  errors. Evidence is in `playtest-artifacts/balance-lab/02-server-job-config.png`
  and `playtest-artifacts/balance-lab/03-server-job-complete.png`.

## Balance Lab class/spec ability damage drill-down (2026-09-17)

- Every class/spec row now exposes **View ability damage** when ability telemetry
  exists. It opens the Abilities tab on that exact source/build, clears
  conflicting search and role filters, and defaults to abilities with positive
  effective damage sorted highest first.
- The drill-down graph and table retain ability name/ID, casts, casts per minute,
  hits, effective/raw damage, damage share, DPS, and damage per hit. Healing,
  overheal, absorbs, usage, and unused-ability filters remain available.
- The production build passed. A real five-second level-20 WoW run against Eck
  the Ferocious opened a Death Knight Unholy row through the new action and
  verified positive effective ability damage plus the expected cast, damage
  share, and DPS columns. Desktop and 480×900 checks passed with no browser
  errors. Evidence is in
  `playtest-artifacts/balance-lab/04b-ability-breakdown.png` and
  `playtest-artifacts/balance-lab/06-mobile-abilities.png`.

## WoW full-spec mechanics and rotation pass (2026-09-18)

- `playerbots-inspired-v3` now gives every one of the 31 Wrath specs a legal,
  level-scaled talent plan and executable profile. The pass repaired stale
  talent prerequisites, companion Rage/Runic Power persistence, short melee
  reach, warrior rage banking, and explicit profile ordering.
- Scripted payloads now make Bloodthirst, Bloodrage, Execute, Mutilate, Killing
  Spree, Penance, Feral Spirit, Conflagrate, and Earth Shield mechanically
  meaningful. Frost/Unholy disease setup, Arms Rend/Mortal Strike order, Arcane
  Barrage, and healer maintenance behavior have direct regression coverage.
- The exhaustive level-80 headless audit completed 2,672 fights with no
  simulation failures under `artifacts/balance-fixes-20260918-level80/`.
  The final 40-run Stockades confirmation under
  `artifacts/balance-fixes-20260918-stockade-level80-final/` exercised all 31
  specs: Feral Spirit, Bloodrage, and Earth Shield were active in every eligible
  run, while the repaired damage signatures produced damage in their applicable
  runs. Low-level boss floors were recalibrated so ordinary/final Stockades
  encounters still retain their 25/45-second mechanic windows. These are
  structured combat checks, so no scene screenshot is applicable. Remaining
  cross-spec DPS/HPS outliers require a separate coefficient-tuning pass.

## Balance Lab level batches and parallel workers (2026-09-19)

- The GM setup now switches between one level and all level brackets. The three
  schedules expose 9 jobs (level 1 plus every 10), 17 jobs (level 1 plus every
  5), or 80 jobs (every level). Every-boss, source-specific boss, and explicit
  boss selection remain available, and the estimate totals all scheduled jobs.
- A browser-driven nine-level exploratory submission created nine persisted
  jobs in one request, started four independent Python/TypeScript processes,
  and completed the first four level reports before the automation time limit;
  the test server then stopped the remaining isolated test jobs. Server tests
  separately prove parallel worker utilization, queued overflow, and all-or-
  nothing rejection when a complete batch will not fit.
- The final production-build smoke pass verified the 9/17/80 schedule controls,
  worker/queue status, specific Paladin Holy and Eck selection, one complete
  real level-20 report, all report/ability analytics, and 480x900 layout with no
  browser errors. Evidence is in
  `playtest-artifacts/balance-lab/02-level-batch-config.png`,
  `playtest-artifacts/balance-lab/04-server-job-complete.png`, and
  `playtest-artifacts/balance-lab/06-mobile-analytics.png`.

## Balance Lab parallel watcher isolation (2026-09-19)

- A production all-WoW, every-10-level Stockades batch completed only the level
  1 and level 80 jobs. Levels 10 through 70 failed during startup before a
  party ran. Every failed job recorded `ENOSPC: System limit for number of file
  watchers reached` beneath Vite's watcher stack, while the Balance Lab runtime
  reported zero active jobs and normal queue capacity after the failures.
- Each parallel Python worker had launched its own `vite-node` process. The
  inherited Vite development configuration watched the large shipped public
  asset tree, so ten simultaneous jobs exhausted the container's Linux watcher
  allowance. Which two levels completed was a startup resource race, not a
  class, level, boss, or combat-model failure.
- The headless bridge now launches with `--mode balance-lab`. That Vite mode
  sets `publicDir: false` and ignores every watch path, eliminating hot-reload
  watchers that one-shot simulations do not use. Server job failures also
  translate the known watcher-limit signature into a specific GM-facing error.
- The Python controller tests, all 14 Balance Lab service tests, the complete
  84-test server suite, and the 62-test benchmark/party suite passed. A resolved
  Vite configuration check returned an empty public directory and
  `watch.ignored: ["**/*"]`; a real level-10 WoW catalog invocation completed
  through the new mode and wrote its 72,723-byte catalog. The production build
  passed, and the authenticated GM smoke flow completed a real server job,
  loaded its analytics and ability breakdown, and passed desktop/mobile checks.

## Default Git and TrueNAS release path (2026-09-19)

- The visual release manager now starts with **Git + TrueNAS only** enabled.
  This mode retains the candidate check, content/server/application tests,
  content publication, production build, staged-file review, commit,
  remote-main verification, and pinned TrueNAS deployment commands.
- The default path does not bump package or Android versions, request signing
  credentials, run Android sync, build an APK, or publish a Gitea release.
  Clearing the toggle restores the existing full APK workflow; **Publish
  Current APK Only** remains available for retrying publication from an already
  versioned and pushed commit.
- **Run test suites** is independently enabled by default. Clearing it skips
  only the content, server, normal-application, and shelved-application test
  commands. The candidate check, content publication, production build,
  staged review, Git verification, and selected deployment/APK stages remain.
  Both the log and final staged review identify the skipped suites.
- All 12 release-manager unit tests and Python compilation passed. The real
  Windows Tk smoke launch also initialized the default-on mode and closed
  cleanly.

## Thor character shell redesign (2026-09-20)

- The `960 x 540` top-display roster now keeps its character stage on the left,
  a 300-pixel saved-character rail on the right, and the primary Continue action
  inside a 62-pixel bottom bar. Both creator steps retain the shared live model
  preview and fit without document overflow.
- All 47 avatar/rendering tests and the full production build passed. Browser
  automation created a Human Priest, exercised both creator steps, and verified
  the roster geometry without page errors. Evidence is in
  `playtest-artifacts/character-shell/01-creator-path.png`,
  `playtest-artifacts/character-shell/02-creator-appearance.png`, and
  `playtest-artifacts/character-shell/03-character-select.png`.
- A production Wailing Caverns smoke run rendered the player and all four party
  members with intact bodies and starter weapon silhouettes. Evidence is in
  `playtest-artifacts/character-shell/04-in-dungeon.png`.

## Thor main-menu redesign (2026-09-20)

- The `960 x 540` top display now gives the destination grid the wide left pane
  and uses a 288-pixel right rail for the selected character's level, name,
  class, and specialization. The character-model preview no longer mounts on
  this screen; authenticated sessions open the online-group invite panel below
  the identity card.
- A live browser pass measured a full `960 x 540` surface with no horizontal or
  document overflow. The destination pane scrolls independently, the identity
  card remains fully visible, no character-preview node mounts, and all removed
  introductory copy is absent. Nineteen focused tests and the production build
  passed.
- The destinations are now split into a dedicated Adventures card for Dungeons
  and Randomized Bite-Sized plus a secondary Character & Collection card. Boss
  testing, companion readiness, reports, and the authorized GM/server tools now
  live behind one controller-navigable GM Menu dialog; unauthorized accounts
  see GM Tools and Balance Lab as locked rather than losing the menu structure.

## Discipline Priest Holy Nova (2026-09-22)

- A level-80 Priest cast Holy Nova in a built Wailing Caverns scene. The live
  cast restored 770 health to the injured player. At 400 ms, one Nova burst was
  visible; by 850 ms, it had ended. Screenshots are in
  `playtest-artifacts/holy-nova/holy-nova-100ms.png`,
  `playtest-artifacts/holy-nova/holy-nova-400ms.png`, and
  `playtest-artifacts/holy-nova/holy-nova-850ms.png`.
- The nearby-enemy damage and nearby-ally healing paths passed the focused
  combat-store test. The four focused suites, Ready-five certification,
  TypeScript build, production build, and release candidate check passed.

## RoM primary five validation (2026-09-22)

- A fresh local account created a Human RoM Rogue. Character creation showed
  level-based skill unlocks. The roster preview rendered the complete character
  body with a starter weapon silhouette after the 3D canvas loaded.
- The same Rogue entered a live Wailing Caverns scene as DPS. The player and
  companions rendered bodies and equipped weapon or staff silhouettes; the
  browser reported no page errors. The left-hand Rogue dagger is not clear in
  these camera angles, so this is a body/equipment-presence check rather than
  visual proof of dual daggers. The resolved virtual loadout test checks both
  dagger slots.
- Evidence: `artifacts/playtest/rom-five/roster-preview.png` and
  `artifacts/playtest/rom-five/dungeon-rogue.png`. The in-dungeon Skills and
  Spellbook panels also showed level unlocks without numbered skill ranks;
  evidence is `skills-no-ranks.png` and `spellbook-no-ranks.png` in that same
  directory. The repeatable flow is `node scripts/playtest-rom-five.cjs`.
- At 1440 x 900 the dual-display "Roster details" toggle overlays the roster
  Continue button. The playtest used a programmatic click to continue; this
  existing layout collision should be addressed in the shell UI.

## Named RoM class pairs (2026-09-23)

- A fresh local account created a male RoM Dwarf Mage. Its live creation and
  roster previews rendered the complete body and starter staff silhouette.
  Evidence: `artifacts/playtest/rom-five/dwarf-mage-creator.png` and
  `artifacts/playtest/rom-five/dwarf-mage-roster.png`.
- The same session entered Wailing Caverns with a RoM Rogue. The party frames
  showed paired-cohort NPC names Serenna, Caelan, and Veyra for that seed; the
  player and companions rendered bodies and weapon or staff silhouettes in the
  scene. The browser reported no page errors. Evidence:
  `artifacts/playtest/rom-five/dungeon-rogue.png`. The companion pair and race
  assignments, including Durnik as a dwarf, are covered by
  `src/game/romFiveValidation.test.ts`; this level-1 scene does not exercise
  the level-10 secondary-class unlock or elite-skill mechanics.

## RoM leveling loot and package catalog (2026-09-23)

- Rebuilt the client and reran `node scripts/playtest-rom-five.cjs` after the
  leveling boss-loot rule and source package catalog changes. The live Rogue
  roster preview showed the complete body and starter blade silhouette;
  the in-dungeon scene showed the player and companion bodies with weapon,
  shield, and staff silhouettes. No browser page errors occurred. Evidence:
  `artifacts/playtest/rom-five/roster-preview.png` and
  `artifacts/playtest/rom-five/dungeon-rogue.png`.
- This level-1 scene checks rendering. Focused boss-card, SQL loot, RuneWaker
  loot, and RoM class tests check that leveling boss item drops are empty,
  cards remain granted, and secondary stat growth uses the primary level.
  Source package contents are cataloged as unverified and were not granted.

## RoM heirloom reference atlas (2026-09-23)

- Rebuilt the client and reran `node scripts/playtest-rom-five.cjs` after adding
  the source-backed 1–79 native-stat references. The live Rogue roster showed
  its body and starter blade; the Dwarf Mage roster showed its body and staff.
  Wailing Caverns showed the Rogue player and companion bodies with equipped
  blade, shield, and staff silhouettes. No browser page errors occurred.
  Evidence: `artifacts/playtest/rom-five/roster-preview.png`,
  `dwarf-mage-roster.png`, and `dungeon-rogue.png` in the same directory.
- These references are not equipped by the runtime, so this level-1 visual check
  confirms the existing character path still renders. The reference suite
  verifies the five class curves and source IDs at every level 1–79.

## Five-class RoM hybrid equipment (2026-09-23)

- Rebuilt and reran `node scripts/playtest-rom-five.cjs` after changing the five
  active heirloom layouts to legal RoM weapon combinations with WoW combat
  stats. The Rogue and Dwarf Mage roster previews showed complete bodies and
  starter blade/staff silhouettes. Wailing Caverns showed the player and
  companion bodies with visible blade, mace, shield, and staff silhouettes;
  there were no browser page errors. Evidence: `artifacts/playtest/rom-five/`
  `roster-preview.png`, `dwarf-mage-roster.png`, and `dungeon-rogue.png`.
- At level 1, the screenshot does not confirm each member's full inventory or
  the level-10 secondary unlock. `src/game/romFiveValidation.test.ts` checks
  every hybrid weapon slot for legality and stat growth at levels 1, 20, 40,
  60, and 79. Targeted Shadoj reports at levels 20, 60, and 79 are under
  `artifacts/balance-lab/rom-hybrid-l*-shadoj/`; they exposed low Mage damage
  and a level-79 Priest death in each seed, so balance remains Needs Testing.

## RoM independent weapon slots (2026-09-23)

- Rebuilt and ran `node scripts/playtest-rom-five.cjs` after adding the trained
  off-hand and ranged heirlooms. The Rogue roster preview and Wailing Caverns
  scene showed a complete character body and starter blade; companion bodies,
  shields, and weapons remained visible. The Dwarf Mage roster still showed its
  body and two-hand staff. The playtest reported no browser page errors.
  Evidence: `artifacts/playtest/rom-five/roster-preview.png`,
  `dungeon-rogue.png`, and `dwarf-mage-roster.png` in the same directory.
- These level-1 screenshots do not show the later training unlocks. Focused
  equipment tests verify the Knight/Scout bow at level 10, Rogue projectile at
  level 12, and Scout/Rogue and Rogue off-hand weapons at level 16.

## RuneWaker sprite particle pilot (2026-09-24)

- Opened `scripts/combat-effects/rom-preview.html` in the in-app browser against
  Vite and inspected frozen 250 ms frames after regenerating the archive. Knight
  Charge showed a blue upward spray with orange sparks (3 layers), Whirlwind
  Shield showed a small gold circular impact (3 layers), and Mage Electric Bolt
  showed a blue-white projectile with streaks (3 layers). Browser console had
  no errors or warnings. The preview tab is retained for inspection.
- These screenshots verify the sprite renderer and packaged textures at the
  preview camera scale. A level-18 Knight and level-10 Mage were then run in
  Wailing Caverns with Original visuals on using disposable local QA saves.
  Whirlwind Shield's impact graph loaded at the caster position, but its
  33 ms emission burst could disappear when a frame skipped past the last
  rate key. Integrating the source rate over each frame restored the burst.
  A 2.5 presentation scale made the gold shield ring visible around the
  Knight at the dungeon camera distance during an isolated cast. Electric
  Bolt cast and resolved in combat with no missing-original warning, though
  the fast projectile was not visually captured in a still frame. Charge
  remains to be checked in live movement. The next visual decoder target is
  the Priest's holy `act_casting_c_hand.ros`: its only unsupported controller
  is `CRuController_EventTrigger`, which the source uses for action callbacks.

## RuneWaker Priest holy cast (2026-09-24)

- The native exporter now records the `EventTrigger` payload. Source event 43
  in holy `act_casting_c_hand.ros` requests looping 3D playback of
  `sound/fx/skill/holy/act_casting_c_hand.wav`; it adds no visual layer. The
  particle archive validates this exact event and marks the partial graph
  renderer-ready. The shared combat manifest attaches its three sprite layers
  and looped cast sound to Priest Urgent Heal, Heal, and Group Heal only.
- The in-app browser preview rendered the holy model as a gold star and swirling
  particles with three loaded textures. In a disposable level-14 Priest save in
  Wailing Caverns, Urgent Heal (2), Heal (6), and Group Heal (Shift+6) each
  showed the gold cast effect on the isolated player during the cast. Group
  Heal's longer cast kept the effect visible across successive frames. The
  browser reported no original-effect load errors; its only warning was an
  unrelated dependency deprecation warning. Audio loop and release/cancel
  behavior passed focused unit tests, but the browser pass did not assess sound
  by listening. The generated map is `docs/ROM_SPELL_MEDIA_MAPPING.md`.

## RoM Priest target and Rising Tide media (2026-09-24)

- Traced Urgent Heal's MagicObject component 501543 to ImageObject 573048 and
  its target `act_benedictory_heal_t_down.ros`. Its `explode` motion requests a
  one-shot 3D `act_benedictory_heal_t_down.mp3` at the healed target. The
  browser graph renders two original sprite layers, the emitted 72-triangle
  rune cylinder, and the animated gold ground decal with source color and
  alpha tracks. It remains marked partial because the decal is placed on a
  level ground plane rather than projected onto nearby terrain triangles.
- Traced Rising Tide's component 620408 to ImageObject 572062 and target
  `act_riptide_t_freeze_hit.ros`. Its five-sprite target graph and the
  `act_icecone01_c_hand.ros` cast and `act_icecone01_f.ros` projectile graphs
  are attached to their respective phases. ROS direct-audio events identify
  `act_icecone01_f.ros.mp3` on projectile release and
  `act_riptide_t_freeze_hit.mp3` on target impact. Impact audio is positioned
  at the target; release audio remains at the caster.
- The focused combat audio suite passed 12 tests, including distinct release
  and target panning. The manifest test checked all three new MP3 paths against
  the source ROS EventTrigger ID 43 payloads and packaged files.
- In the in-app RuneWaker preview, Urgent Heal's impact showed gold stars, a
  green rune cylinder, and a gold target-ground ring at the frozen 250 ms frame.
  Rising Tide's cast showed a blue ring, its release a bright blue flash, and
  its target impact a blue and magenta burst. All four graphs loaded without
  browser errors or warnings. Audio mapping was source and test verified; this
  browser pass did not assess the sound by listening.
- Live Priest combat in Wailing Caverns confirmed Urgent Heal's original gold
  rune circle and sparkles around the selected party target Veyra. In Boss
  Testing against Verdan the Everliving, Rising Tide showed blue source
  particles around the Priest during the cast, a blue projectile on release,
  and a bright blue and magenta burst across Verdan's visible lower body at
  arrival. The impact presentation now follows the projectile's 180–550 ms
  travel time; large-target attachment uses visual bounds so the burst remains
  in view. The browser reported no original-media or WebGL errors. Audio
  timing and target panning are source and test verified, not listening tested.

## RoM primary-five media batch preview (2026-09-24)

- The batch now follows direct MagicCollect component ImageObjects to their
  target `ExtendACT` models, packages 170 unique ROS graphs and source textures,
  and attaches 74 renderer-ready graphs plus Urgent Heal's partial target graph.
  The generated class audit counts 84 castable skills with at least one playable
  original phase. Source audio is packaged for 125 of 129 cohort spell entries.
- In the in-app RuneWaker preview, frozen at 250 ms, Priest Ice Fog showed a
  bright ice spray, Knight Holy Seal a blue sigil, Scout Snipe an orange target
  ring, Rogue Sneak Attack a blue arc, and Mage Electric Explosion a layered
  blue burst. All five graphs loaded and rendered; each status showed two or
  three source sprite layers with matching textures.
- The first exporter pass falsely marked seven graphs ready despite emitter
  geometry or sprite billboard modes the WebGL renderer rejects. The packager
  now blocks those modes, and the parser test loads every ready graph. This
  preview pass did not assess sound by listening or cast the five spells in
  live combat; those are still needed to confirm placement and phase timing.
- A shared source alpha track rule then enabled four additional graphs whose
  alpha anchor covers every sprite layer in the motion. The Mage Fireball
  cast graph rendered a bright red and gold burst in the preview at 250 ms.
- RuneWaker audio phases now play all distinct packaged source cues together;
  the previous weighted selector could drop one of several cast or impact
  layers. A focused audio test confirms two impact files create two voices.
  Listening QA is still needed for the balance of these layers.

## RoM Priest channel and stacked status icons (2026-09-25)

- In a disposable level-30 Priest/Knight Wailing Caverns run,
  `scripts/playtest-rom-status-effects.cjs` cast Chain of Light at a nearby
  Deviate Ravager. The cast bar showed Channeling and the rendered scene showed
  a sustained white-gold beam between the Priest and target. The browser
  reported no page errors. Evidence: `artifacts/playtest/rom-status-effects/chain-of-light-and-stacked-dot.png`.
- Three same-spell damage-over-time applications on that target appeared as one
  icon with a small `3`; the browser assertion also checked the icon count.
  This exercise checked HUD grouping, while focused combat tests checked four
  separate channel damage pulses. Source audio was not assessed by listening.

## Raksha Temple collision performance (2026-09-26)

- A production-build offline playtest in headless Edge measured about 0.6 FPS
  after Raksha Temple loaded. A CPU profile placed nearly all frame time in
  Rapier character movement and physics. Its collision GLBs use interleaved
  position buffers; passing the packed float array as XYZ vertices created
  malformed trimeshes.
- After unpacking positions, the same 15-second sample recorded 600 frames at
  16.67 ms average frame time (60 FPS). The entrance's WDB collision lacks a
  walkable floor, so Raksha now uses its navigation mesh as physical ground.
  The final browser pass retained 60 FPS, showed the party at 5/5 instead of
  `ROUTE BLOCKED`, and kept the player at the entrance. An eight-second forward
  movement sample explored 2 m and still held 60 FPS. Screenshots:
  `playtest-artifacts/raksha-final-ground.png` and
  `playtest-artifacts/raksha-moved.png`. The scene remains very dark;
  this pass verified frame rate and floor support, not lighting quality.

## Scholomance rendering and population performance (2026-09-26)

- In a production-build offline run in headless Edge, Scholomance's entrance
  averaged 31.78 ms per frame (about 31 FPS) over 15 seconds. A CPU profile
  was dominated by Three.js render traversal and matrix updates, while the
  visual asset audit counted 5,140 raw draw calls and 3,175 mesh nodes.
  WebGL instrumentation saw about 867 draw calls per rendered frame before
  the visual batching change.
- Scholomance had 191 static mob spawns within its former 95 m activation
  radius at the entrance, across stacked indoor floors. Its 48 m radius and
  global opaque instancing reduced the 15-second average to about 20.1 ms
  per frame (about 50 FPS). Holding forward for 12 seconds explored 30 m;
  the subsequent sample remained near 20.1 ms per frame with the party at
  5/5. Entrance and movement screenshots showed intact geometry and no page
  errors: `playtest-artifacts/scholomance-after.png` and
  `playtest-artifacts/scholomance-moved.png`.

## Bite-Sized Dungeon Race (2026-09-26)

- `scripts/playtest-dungeon-race.cjs` used headless Edge with a fresh offline
  character. Practice entered a Bite-Sized dungeon with five player-party seats,
  displayed the top clock and NPC progress, and kept the game scene active
  after a forced NPC win showed the “You lose” toast. The NPC roster contained
  five members. Desktop and 960×540 Thor-size captures are in
  `playtest-artifacts/dungeon-race/`.
- `scripts/playtest-dungeon-race-online.cjs` created two online accounts in
  separate browser contexts. Both entered the same human match, dungeon, and
  seed with each other as opponents. A third account queued alone and received
  a five-member NPC team after the five-second fallback. Both scripts reported
  no browser page errors.
- The practice screenshot check used a time jump to trigger the loss toast,
  then injected a defeated boss-card fixture into the active combat store. The
  normal automatic loot action added one Bazil Thredd card after the loss and
  showed its reward window (`05-card-after-loss-fixture.png`). It did not
  traverse and defeat the final boss through live combat. Server tests verify
  that a losing clear is accepted for the dungeon time board.

## Dungeon Race exit regression (2026-09-26)

- `src/app/dungeonRaceStore.test.ts` covers immediate cleanup of active and
  practice races, cancellation during an in-flight queue request, ordered
  server leave before requeue, and an old preparation response after start.
- `scripts/playtest-dungeon-race-exit.cjs` used headless Edge with a fresh
  online account. It left an NPC race through the pause menu, entered an
  ordinary Bite-Sized dungeon without a race HUD, and requeued into a new
  online match without an “already in a race” error. Screenshots are in
  `playtest-artifacts/dungeon-race-exit/`. The offline practice playtest also
  left after a modeled loss, confirmed the local race state cleared, and
  started practice again. Browser page errors were absent in both passes.

## Bite-Sized spawn floor regression (2026-09-26)

- Fresh offline Dungeon Race practice entries in Wailing Caverns and Utgarde
  Pinnacle previously settled about 0.5 m below the visible floor. During the
  race start gate, Rapier physics was paused while PlayerRig accumulated
  downward movement. Holding the player controller for that gate removed the
  stored fall. Wailing keeps its original entrance; Utgarde now starts on the
  connected Skadi wave corridor with three source first-wave guards.
- `scripts/playtest-bite-sized-spawns.cjs` entered both dungeons through the
  menu, waited for content, intro, and race gate completion, then sampled five
  seconds of play. Wailing was grounded at y=-31.016 from a -30.92 spawn;
  Utgarde was grounded at y=104.812 from a 104.8 spawn. Both kept their entry
  X/Z position, and there were no browser page errors. Before/after screenshots
  and `result.json` are in `playtest-artifacts/bite-sized-spawns/`. The player
  stands visibly on the floor in both after screenshots. Utgarde corridor samples
  at z=370, 390, 410, 430, 445, and 460 also stood on collision in the live
  client; the source guard points and boss position have collision and
  navigation surfaces.

## Paused browser memory (2026-09-26)

- A fresh offline Wailing Caverns Bite-Sized run in headless Edge entered the
  dungeon, opened the pause menu, and held it for 60 seconds. The paused scene
  used R3F's `demand` frame loop and reported no browser page errors. Before
  the classic talent lookup change, forced-GC JavaScript heap remained at
  266 MiB across 0, 15, 30, and 60 seconds; DOM node count also stayed flat.
- `src/game/wow335AbilityCatalog.ts` now validates classic talent IDs through
  the already loaded compact runtime talent catalog. The built classic ability
  chunk fell from 41.4 MiB to 2.1 MiB, because it no longer includes the
  36.6 MiB CoA source catalog. The same paused browser run measured 155, 155,
  155, and 156 MiB after forced GC at 0, 15, 30, and 60 seconds. The paired
  screenshots and samples are in `playtest-artifacts/pause-memory/`.
- The one-minute sample did not show a growing JavaScript heap. It does not
  measure GPU allocations or rule out growth in a different dungeon or over
  a longer idle period.

## Character-specific catalog loading (2026-09-26)

- The production build now imports `src/game/romBaseStatGrowth.generated.json`
  (4.5 KB) for shared stat calculations. The full 1.7 MB
  `romPlayerCatalog.generated.json` remains in the RoM ability chunk and is
  requested only when a RoM class needs its abilities.
- `src/game/talentRuntimeCatalog.ts` loads its generated WoW/CoA talent data
  after a non-RoM class requests combat catalogs. A fresh RoM Rogue main-menu
  session fetched `romAbilityCatalog-*.js` without fetching either the
  `wow335AbilityCatalog-*.js` or the talent-data chunk. A fresh WoW Priest
  fetched the classic ability and talent-data chunks without fetching the
  RoM ability chunk. Both sessions had no browser page errors.
- The RoM Rogue roster and Wailing Caverns dungeon smoke test passed with the
  WoW/CoA talent data absent from the initial RoM page load. Browser request
  results and menu screenshots are in `playtest-artifacts/catalog-loading/`;
  dungeon screenshots are in `artifacts/playtest/rom-five/`.

## Browser memory reset at shell boundaries (2026-09-26)

- Browser ES modules cannot be unloaded individually from an active page.
  Changing to a different character now reloads the page with a one-use
  `sessionStorage` handoff to restore the selected character and destination.
  Leaving a gameplay mode for the roster or main menu also reloads after the
  run's normal save path. A live online group activity suppresses this reset
  to avoid re-entering the same activity on reconnect.
- The main-menu catalog gate no longer reloads the full gameplay runtime or
  retains talent art up front; it preloads those resources when gameplay starts.
- In a production Edge playtest, switching from a WoW Priest to a RoM Rogue
  produced one reload and returned to the RoM main menu. The WoW talent editor
  opened on demand before that switch. The new RoM page loaded
  `romAbilityCatalog-*.js` without a WoW ability or `GameRuntime` chunk.
  Leaving a RoM Wailing Caverns run produced a second reload and returned to
  the same character's main menu without a `GameRuntime` chunk. Switching to
  another RoM character also reloaded. The one-use handoffs were consumed,
  and the page reported no errors. Results and screenshots are in
  `playtest-artifacts/memory-reset/`.