Files
aetherbound-guild/docs/06_DEVELOPMENT_AND_REVIEW.md
T

50 KiB

Aetherbound Guild - Development Plan And Review

Complete-game Goal, hand-drawn repair boundary, ordered implementation plan, roadmap and decision ledger.

Owner Review / Owner 审核

当前唯一执行顺序是:启动与首次设置、标题首页、首次四选二建立持久远征团、公会与市场、自动排阵与风险、战斗与结算、第二轮经营、八个区域、结局与通关后玩法。每一步都从启动页真实可达。

本文件是当前六份审核文档之一。历史拆分文件仅保留在 Git 历史中;文内规则、表格和 ID 仍是完整开发依据。

Aetherbound Guild Complete Development Goal

Program ID: ABG-COMPLETE-GAME Status: ACTIVE Owner approval: 2026-08-16 Activation baseline: ee33f775527c20a79703a4fcfe073596c00ca179 Execution topology: one development session; no multi-Agent implementation Master roadmap: 06_DEVELOPMENT_AND_REVIEW.md

Owner Outcome

Continue plan-driven development from the accepted P10.3 Step-2 source until Aetherbound Guild is a complete local release candidate, rather than treating each small runtime slice as an unrelated endpoint. Ordinary implementation may advance without repeated Owner authorization. Every milestone remains bounded, verified, committed, pushed and recorded before the next milestone starts.

Complete Player-Visible Stopping Point

The local candidate must let a new player complete this authored arc:

open the finished landscape game
-> learn Market valuation, Company building and front-to-rear order
-> clear sixteen Contracts across eight mechanically distinct Regions
-> recruit, equip, promote and continuously grow generated Recruits
-> recover knocked-out members at the next stage without losing their identity
-> discover meaningful Profession, Trait, equipment and Artifact combinations
-> defeat sixteen disclosed-rule Bosses
-> choose the finite Chronicle ending or enter Oath Season
-> start a strategically different postgame Contract with the persistent Company

The candidate must contain the complete approved envelope: 12 base, 24 regular advanced and 6 hidden hybrid Professions; 36 Traits; 320 equipment items; 60 Artifacts; 100 ordinary/elite enemies; 16 Bosses; 8 Regions and 16 Contracts.

Experience Contract

  • Fantasy: build a changing guild Company from generated Recruits and keep it alive while restoring a chain of dangerous fantasy routes.
  • Primary action: compare visible capability, cost and threat, then commit a Recruit, item, order, route or run-law choice.
  • Automation: the prepared Party line resolves battle without attack spam; automation exposes causes instead of hiding them.
  • Bottleneck: Coin, one item per Recruit, Line capacity, order, current enemy grammar and current-battle knockouts prevent one solved build.
  • Recurring decision: spend, recruit, reserve, equip, reorder, promote, recover or accept route risk according to the current Company state.
  • Mastery: predict the first decisive interaction and deliberately build a counter without requiring one exact Profession or item.
  • Persistent transformation: Contract clears add Records, Renown, catalog breadth, Region access and optional laws without permanent flat power.
  • Completion: a 15-25 hour first clear and a 60+ hour mastery structure are fully implemented, locally verified and ready for human/device/release gates.

Program Acceptance Gates

Complete Runtime

  • All approved content identities resolve from scalable runtime data and alter a real decision; design-only catalog rows do not count as implemented.
  • The complete Title, Initial Company, Guild Charter, Contract creation, Market, battle, result, recovery, Boss, ending and Oath Season state graph is reachable.
  • One canonical simulation and save authority replaces milestone-only behavior before bulk Region production depends on it.
  • Runtime presentation uses final-direction landscape composition, authored character motion, event-bound damage/healing/status feedback, VFX, music and SFX rather than app-like cards or static debug substitutes.

Verified Candidate

  • Each milestone passes focused negative tests and its actual rendered flow.
  • Campaign-wide normal, active, efficient, adversarial and returning profiles prove pacing, decision density, recovery, anti-dominance and ending reach.
  • Save migration, corruption, rollback, duplicate input, interruption, clock change and zero-offline-value behavior remain authoritative.
  • English, Simplified Chinese, small/reference/large landscape, 130% text, reduced motion/flashes, mute and low power remain usable.
  • The final local candidate has a clean pushed source, source-bound evidence and a fresh H5 review delivery only when the Owner checkpoint authorizes it.

Owner Checkpoints

Ordinary work does not pause between milestones. Owner review is requested at only three planned boundaries:

  1. production-quality visual/audio vertical slice;
  2. the completed RG-04 midpoint build;
  3. the complete first-clear beta/final local candidate.

Owner feedback may reopen the earliest failed player-experience boundary. Automated evidence never closes understanding, fun, visual or listening gates.

Frozen Scope

  • Third-party commercial code, art, music, audio or copied expressive content.
  • Ads, gacha, paid currency, energy, daily-streak or online-required systems.
  • A second battle lane, manual skill spam, same-battle resurrection or fixed authored protagonist relationships.
  • Multi-Agent development unless the Owner explicitly reauthorizes it.
  • Native device packaging, signing, TestFlight, Play Console, store submission, publishing and release acceptance without separate Owner authorization.

Image and TTS generation may use the workspace standing authorization only inside its per-task call and cost ceilings. Video, ElevenLabs music/SFX and any expanded paid batch still require their own authority.

Current Milestone

P10.3 and FG-2 are functionally complete. FG-3 is locally accepted at pushed source 370f2a5: both RG-02 Contracts, 17 ordinary Rounds, both Bosses, permanent casualty/recovery, two Company Records, first-clear Renown and reload-stable RG-03 unlock pass five-profile simulation, Save-v10 persistence, complete runtime, parsed-touch, audio and MovieWriter gates. FG-4 gameplay, simulation, Save-v11 and Web transport reached a combined midpoint candidate, but Owner review on 2026-08-17 rejected its pixel-art presentation as a contradiction of the accepted Japanese 2D hand-drawn direction. The pixel H5 is not a visual baseline. FG-5 and bulk content remain frozen while the active hand-drawn presentation repair preserves the working systems and replaces only the rejected visual layer. Human comprehension, fun, visual and listening judgments remain open rather than inferred from automation.

Aetherbound Guild complete local candidate ready

or:

Aetherbound Guild complete development redesign required:
<earliest failed product boundary>

FG-4 Japanese 2D Hand-Drawn Presentation Repair Goal

Task ID: ABG-FG4-HANDDRAWN-PRESENTATION-REPAIR Status: OWNER_APPROVED_ENTRY_FIRST Owner direction: 2026-08-17 Planning baseline: d59555c5a3cf95fc44de79ba06c02e650eb3df3b Contract revision: fg4-handdrawn-repair-r1

Earliest Failed Boundary

Gallery item 5vl2EqsYWkey renders the playable midpoint as 16-bit pixel art. The Owner rejected that direction because the accepted product identity is a landscape Japanese 2D hand-drawn fantasy game. Automated layout, asset and runtime checks did not have authority to overrule that human visual decision.

Owner Outcome

Keep the implemented game and restore its approved presentation. Do not restart the battle, progression, content, Save or input systems. Replace the rejected pixel visual layer with thin-line/cel-color actors, softly painted environments and compact field-attached game HUD.

Planning Outcome

The Owner reviewed the redevelopment order and approved player-entry-first implementation. The six numbered documents under docs/ retain:

  1. the canonical product and game-system documents;
  2. the complete 50-page inventory and development order;
  3. the Japanese 2D hand-drawn visual contract;
  4. the runtime art, animation, UI, VFX and audio production plan; and
  5. the full-game roadmap and explicit frozen scope.

The authorization starts only DEV-1: Boot, first-launch setup, Title and essential Settings. Initial Company, Guild, Market, formation, battle, later Regions, bulk generation and delivery remain frozen until their ordered step.

Intended Player-Visible Stopping Point After Approval

open one replacement landscape H5
-> recognize a Japanese hand-drawn fantasy guild battle within three seconds
-> read the imminent enemy action first
-> read the ordered Company line second
-> read the objective and compact controls third
-> watch contact, guard, ranged attack, healing, damage and defeat animation
-> reach result or knockout recovery without leaving the painted world
-> continue into the unchanged RG-03/RG-04 midpoint state

Preservation Boundary

  • Preserve accepted gameplay, economy, content, Save-v11, localization, accessibility, touch, audio-event and simulation behavior.
  • Preserve the rejected pixel assets and H5 as Git history; do not use them as visual authority or expand their asset families.
  • Reuse accepted Phase-6 Japanese hand-drawn concepts and motion sources before requesting new images.
  • Keep engine text and numerical values crisp. Non-text runtime art is smooth hand-drawn/cel-painted work, not nearest-scaled pixels.

Frozen Scope

  • Runtime or gameplay implementation before Owner plan approval.
  • New gameplay, economy, Save version, content rows or balance changes.
  • FG-5 and later Region implementation.
  • New image generation during planning review.
  • Native packages, devices, signing, TestFlight, stores and release.
  • Multi-Agent development.

Planning Terminal

Aetherbound Guild hand-drawn redevelopment plan ready for Owner review

FG-4 Japanese 2D Hand-Drawn Presentation Repair Contract

Revision: fg4-handdrawn-repair-r1 Fixed functional baseline: d59555c5a3cf95fc44de79ba06c02e650eb3df3b Implementation state: DEV_1_ENTRY_FLOW_READY

Visual Identity

An ordinary but capable Japanese fantasy guild forms one continuous horizontal Company line inside a bright, softly painted unstable world. Characters use thin controlled linework, restrained cel color and reference-led proportions. The world, imminent action and line relationship dominate; UI behaves as attached field instruments rather than a separate application dashboard.

Authority

The controlling visual references are:

The generic portfolio pixel-art rule is inapplicable to this repository. The Owner's 2026-08-17 rejection supersedes every machine visual-pass claim made for the pixel H5.

Fixed Battle Comparison

  • Viewport: 844x390, with 760x360 and 1280x720 responsive checks.
  • State: one unchanged midpoint battle fixture with a visible incoming threat, ordered Company, melee contact, ranged action, healing, damage and defeat.
  • First focal read: imminent enemy action and affected target.
  • Second focal read: front-to-rear Company order and active profession action.
  • Third focal read: objective, pause, speed and inspect controls.
  • Ordinary actor scale: no fixed percentage; preserve the complete Company line, enemy relation and movement corridor, then confirm the real runtime frame with the Owner. Bosses may be larger without erasing crowd readability.

Art And Import Rules

  • Background, midground and foreground remain separate editable painted layers.
  • Actor identity is locked before action frames. Body animation and detached VFX remain separate assets with stable pivots, baseline, facing and light.
  • Smooth non-pixel filtering is required for painted sources. Palette reduction, hard square clusters, nearest-neighbor magnification and pixelation are prohibited.
  • Existing accepted sources are audited for reuse before generation. Any future image request uses the installed image-gen route, begins with a dry run and remains within four paid calls and USD 5 for its approved task.
  • Raw generated sheets are never runtime-final. Preserve prompts, task IDs, costs, accepted/rejected cells, processing, hashes and import settings.

UI Rules

  • World and actors occupy the primary surface.
  • Top and bottom HUD rails remain compact and attached to screen edges.
  • One primary action is visually dominant per decision state.
  • Cards are allowed only for repeated offers or inventory entries; nested cards, generic dark panels and explanatory walls are rejected.
  • Engine-rendered Simplified Chinese and English remain readable at 130% text.
  • Touch targets remain at least 48 logical pixels and retain non-color states.

Motion And Feedback

Every representative action has anticipation, committed impact, readable value, recovery and legal terminal state. Required states are idle, advance/contact, melee impact, ranged release, heal, guard, hurt, defeat and victory recovery. Damage/healing values, hit pause, shake and detached VFX follow authoritative events and obey reduced motion/flashes and low-power settings.

Rejection Conditions

  • The result still reads as pixel art, a card application or a text debugger.
  • Actor identity, anatomy, facing, scale, pivot or baseline drifts between frames.
  • The player must read the title or explanatory copy to recognize the fantasy.
  • UI or effects cover the imminent threat, active action, HP or primary control.
  • A concept image is presented as proof of real runtime integration.

Evidence Boundary

Agent inspection may reach agent_visual_reviewed. Only the Owner can restore human_reviewed. Device, packaging, signing, stores and release remain open.

Aetherbound Guild Full-Game Implementation Roadmap

Status: ACTIVE Owner-approved: 2026-08-16 Active phase: FG-4 Japanese 2D Hand-Drawn Entry Flow Program Goal: 06_DEVELOPMENT_AND_REVIEW.md Active milestone plan: 06_DEVELOPMENT_AND_REVIEW.md

Current Truth

  • The complete design catalogs exist, but a design row is not implemented content and a production brief is not a finished asset.
  • RG-01's two Contracts have a verified playable runtime baseline.
  • RG-02 is runtime-verified through Round 4 and stable Market 5.
  • P10.3 is complete through stable Market 8 at simulation_verified + runtime_verified + agent_visual_reviewed. Five profiles and the complete Save-v9/layout candidate matrix pass from fixed source d00c7ad.
  • The executable and production presentation authority now covers 4 of 16 Contracts through the complete RG-02 candidate.
  • FG-2 functional runtime, Web and touch evidence passed at source 3302134, but its pixel presentation is superseded by the 2026-08-17 Owner visual rejection. It is not current art authority.
  • FG-3 source 370f2a5 passes five-profile simulation, 15-state Save-v10 persistence, complete RG-02 normal rendering, parsed touch, media, audio and MovieWriter gates. Read-only PMO acceptance is complete.
  • FG-4 Step 5 source 791187c closes the complete RG-03 candidate. Five profiles cover 265 decisions, 16 winning secondary responses, legacy permanent-loss recovery, nine reloads and 15 Save-v11 states; the retained production runtime passes 12-state layout/accessibility, parsed touch, event-bound audio and a 24.83-second recording. Six of sixteen Contracts now reach simulation_verified + runtime_verified production presentation.
  • The combined RG-03/RG-04 midpoint systems and Web transport pass, but Owner review on 2026-08-17 rejects the pixel-art runtime as a direct contradiction of the accepted Japanese 2D hand-drawn product contract. Systems evidence is retained; the Owner has approved rebuilding presentation in actual player order, beginning with Boot, first-launch setup and Title.

Program Sequence

Only one phase and one child-plan step may be IN_PROGRESS.

Phase Status Player outcome Owned production Exit evidence Owner gate
FG-0 Program activation COMPLETE Development has one visible endpoint and ordered route. Program Goal, roadmap, checkpoint and ledger. Clean pushed planning commit. Approved 2026-08-16.
FG-1 Finish P10.3 COMPLETE Clear Siltwake Rounds 5-7, choose an Artifact, optionally promote and reload at changed Market 8. Existing P10.3 Steps 3-5 only. Battle counterfactuals, five profiles, save/layout matrix, parsed-touch runtime and recording. No pause.
FG-2 Production vertical slice and canonical runtime FUNCTIONAL_COMPLETE_VISUAL_SUPERSEDED One Market -> preparation -> battle -> result/recovery chain executes through the canonical runtime. Runtime/state integration, event-bound feedback and audio routing; pixel art is rejected. Functional regressions, recording, touch and H5 transport remain supporting evidence only. Visual direction reopened 2026-08-17.
FG-3 Complete RG-02 COMPLETE Finish both Siltwake Contracts, both Bosses and unlock RG-03 with a changed persistent Guild. Remaining EQ-065..080, AR-009..015, both RG-02 Bosses, Round 8-12 and complete second Contract content, art and audio. Source 370f2a5; five profiles, 15-state persistence, complete runtime/media/audio candidate. PMO accepted 2026-08-16.
FG-4 Complete RG-03 and RG-04 ENTRY_FLOW_READY Preserve the verified midpoint systems and restore the approved Japanese 2D hand-drawn experience in player order. Boot, first launch, Title, Initial Company, Guild, Market, Party, battle, result/recovery and regional presentation, converting legacy lifecycle semantics only in their named step. End-to-end launch reachability, normal-render motion, responsive states and replacement H5. Checkpoint 2: midgame depth, clarity and continuation.
FG-5 Complete RG-05 and RG-06 PENDING Adapt to echoes, rear intrusion, moving rail order and airborne timing across four Contracts. EQ-161..240, AR-031..045, EN-R5/R6, four Bosses and complete production media. Same content, balance, save, runtime, accessibility and media gates as FG-4. No pause.
FG-6 Complete RG-07 and RG-08 PENDING Resolve hidden-information and synthesis Regions, defeat the final Boss and choose Chronicle or Oath Season. EQ-241..320, AR-046..060, EN-R7/R8, four Bosses, ending, Final Ledger and full campaign presentation. End-to-end first-clear simulations, ending atomicity, full campaign runtime checkpoints and complete content/media coverage. Checkpoint 3: first-clear beta.
FG-7 Mastery and postgame PENDING Discover six hidden hybrids and play strategically different Oath runs without flat-power grinding. Hidden Profession unlocks, optional run laws, alternate Boss breadth, Records, collection knowledge and mastery challenges. 60+ hour modeled breadth, anti-dominance profiles, save migration and representative postgame runtime. No pause unless product depth fails.
FG-8 Alpha, beta and local release candidate PENDING A new player can understand, finish and replay the complete game in a stable local build. Whole-game UX polish, balance, localization, accessibility, performance, lifecycle, final media coverage and defect closure. Full suite, campaign simulations, complete state matrix, normal-render recordings, fresh review H5 and clean pushed source. Final Owner human acceptance remains open.

Milestone Rules

Every phase receives its own child Goal, frozen contract and implementation plan before product edits. Each plan must:

  1. define one player-visible stopping point;
  2. keep exactly one step IN_PROGRESS;
  3. list owned and forbidden paths;
  4. test static/schema, focused behavior, regressions, simulations and runtime in increasing cost order;
  5. create an explicit focused commit and push it before advancing;
  6. update this roadmap, ../CURRENT_CHECKPOINT.md and the append-only ledger only after the actual state changes.

Ordinary completion advances directly to the next approved phase. Work pauses only at the three Owner checkpoints, a real product redesign boundary, an external service/account blocker, or a separately authorized delivery boundary.

Content Completion Definition

A content identity is complete only when its schema validates, mechanic is implemented, references resolve, localization exists, decision counterfactual passes, production asset/audio state is recorded and the content appears in a real rendered encounter. Count-only JSON or Markdown does not close content.

Bulk Region production must reuse the FG-2 canonical runtime and production pipeline. Do not create another phase-specific duplicate simulator for each Region. Compatibility adapters may preserve immutable regressions, but new gameplay authority must have one source of truth.

Planned Human Review Deliveries

  • Checkpoint 1: short H5 focused on final-direction Market, battle motion, result and recovery. This decides presentation before bulk production.
  • Checkpoint 2: continuous build through RG-04 for strategy-depth and comprehension review.
  • Checkpoint 3: complete first-clear beta and later final local candidate.

H5 export/deployment is activated at each checkpoint; native packaging, signing, device installation, stores and publishing remain outside this plan.

Program Issue Register

ID Status Boundary
FG-I01 RESOLVED_FG1 P10.3 Step 3 proves three Artifact and two promotion valuation reversals.
FG-I02 RESOLVED_FG2_STEP1 The campaign adapter now preserves exact P10.0 snapshots/events and forwards all transactions to canonical services.
FG-I03 REOPENED_VISUAL Event-bound motion/audio logic passed local gates, but pixel production art is rejected and must be replaced.
FG-I04 OPEN_OWNER Human comprehension, fun, visual and listening acceptance remain unproven.
FG-I05 FROZEN_DELIVERY Device, packaging, signing, stores and release require separate Owner authorization.
FG-I06 OPEN_VISUAL_REPAIR Owner rejected pixel Gallery item 5vl2EqsYWkey; accepted hand-drawn direction must be restored before FG-5.

Immediate Next Action

Complete and push only DEV-2A: align all six current authorities to the persistent Expedition Company and battle-local knockout model. Activate DEV-2B in a later focused planning commit; do not implement the draft service while the authority-correction commit remains unpushed.

FG-4 Japanese 2D Hand-Drawn Redevelopment Plan

Status: OWNER_APPROVED Active implementation boundary: DEV-2A_AUTHORITY_CORRECTION_CLOSURE Goal: 06_DEVELOPMENT_AND_REVIEW.md Contract: 06_DEVELOPMENT_AND_REVIEW.md Page plan: 03_PAGES_AND_UX.md

Development Sequence

Step Status Player-visible outcome Approval boundary
DEV-0 Planning and cleanup COMPLETE Six current documents define the game, pages, art direction and order. Owner approved entry-first development.
DEV-1 Boot, first launch and Title READY Launch, verification, language/accessibility setup, Title and essential Settings form one working game entry in the accepted Japanese 2D fantasy UI language. Owner can reopen, reset, change settings and reach New Game without dead controls, and accepts the play-size visual result.
DEV-2 Initial Expedition Company IN_PROGRESS Four equal stable generated offers can be inspected; exactly two are selected and committed as the persistent opening roster of one continuous incremental game. Owner understands the four offers, creates the intended two-Recruit Expedition Company and can restart without rerolling or losing the draft.
DEV-3 Guild and first Market PENDING The Guild identifies the next action; Market comparison and recruit review build the opening Company. Every purchase, cancel, affordability and return path is truthful.
DEV-4 Automatic formation and risk PENDING Player chooses members; the system orders them and exposes protection, reach, threat and reward before commitment. Owner can explain the formation and selected risk without coaching.
DEV-5 Battle, result and recovery PENDING Real melee, ranged, magic, healing, guard, damage, knockout, reward and next-stage recovery complete the first 15-20 minute loop. Owner completes one full first cycle in a playable H5 and sees the same knocked-out Recruit return next stage.
DEV-6 Second cycle and Company management PENDING Persistent roster, equipment, Artifacts, levels, promotion, Save/resume, dismissal and Contract return support continued play. Owner understands retained growth, battle-local knockout and the next decision.
DEV-7 RG-01 through RG-04 production conversion PENDING Eight Contracts and eight Bosses use the approved hand-drawn presentation over preserved systems. Replacement midpoint H5 passes depth, clarity and continuation review.
DEV-8 RG-05 through RG-08 and ending PENDING Remaining Regions, Contracts, Bosses, catalogs and the First Chronicle ending are complete. End-to-end first-clear beta reaches the finite ending or Oath Season choice.
DEV-9 Postgame and local release candidate PENDING Hidden Professions, Oath laws, Records, all 50 page states, localization, accessibility, performance and defect closure are complete. Final local candidate is ready for separate human/device/package gates.

Why Player Entry Comes First

The Owner must be able to experience every increment in the same order as a new player. A standalone battle frame can validate art but cannot prove launch, navigation, settings, run creation or tutorial continuity. Development therefore keeps one continuous path from Boot to the current stopping point. Battle visual and motion work begins only after New Run, Market, formation and risk can reach it through real navigation.

Execution Boundary

DEV-1 is locally complete and its human gates remain open. The Owner authorized DEV-2 on 2026-08-17 with one corrected product boundary: the choose-two Recruits become a persistent initial Expedition Company, Contract runs are bounded expeditions inside continuous incremental growth, and a zero-HP Recruit returns for the next stage rather than dying permanently.

DEV-2 may correct current design authority, implement the stable initial draft, the choose-two page, its review/commit persistence and a bounded Guild handoff. It may not implement the Guild/Market, retrofit legacy battle/casualty code, change existing FG-4 results, begin DEV-3, generate bulk assets, package or release. The legacy FG-4 implementation remains preserved evidence until its named player-order conversion step.

DEV-1 Durable Implementation Plan

Player-visible stopping point:

launch the real Godot game
-> watch named local verification progress or recover through Retry
-> complete first-launch language, safe-area and accessibility setup once
-> reach the painted Title with truthful New Game, Continue, Settings and Credits states
-> change and persist every essential DEV-1 setting
-> restart or reset the entry state and repeat the complete path

Only one row may be IN_PROGRESS. A failed check keeps that row active.

Substep Status Owned product boundary Narrow acceptance check Commit boundary
DEV-1A SYS-001 Boot COMPLETE New entry controller, resource/settings/save verification, named progress, recoverable error, Retry and safe diagnostics ABG_DEV1_BOOT_OK; ABG_DEV1_BOOT_RUNTIME_OK; strict attempt runner PASS Focused Boot implementation, test, plan/checkpoint update and push
DEV-1B SYS-002 First launch COMPLETE Persistent language, safe area, 100/115/130% text, reduced motion/flashes, captions and usable first-launch completion ABG_DEV1_FIRST_LAUNCH_OK; Boot runtime handoff regression PASS Focused setup implementation, persistence/layout test, plan/checkpoint update and push
DEV-1C SYS-003 Title COMPLETE Painted guild-threshold Title, truthful New Game/Continue/Settings/Credits state and exact return/back behavior ABG_DEV1_TITLE_OK; first-launch regression PASS Focused Title implementation, state/navigation test, plan/checkpoint update and push
DEV-1D SET-001..006 + SYS-008 COMPLETE Settings index, display/text, motion/flash/power, audio/captions, controls, language/locale, credits/licenses/data notice ABG_DEV1_SETTINGS_OK; Title and first-launch regressions PASS Focused Settings/Credits implementation, behavior/localization test, plan/checkpoint update and push
DEV-1E Entry-flow candidate COMPLETE End-to-end player order, responsive refinement, restart/reset fixtures and candidate evidence Source 43e5bac; ABG_DEV1_CANDIDATE_SUITE_OK; ABG_FG4_MIDPOINT_SUITE_OK; strict diagnostics 0 Candidate docs/evidence closure commit and push
DEV-1F Title chrome visual repair COMPLETE Fixed Title frame translates the accepted reference's enamel, ivory, gold-trim, cut-corner, heraldic and icon-forward grammar without copying its battle composition Six DEV-1 headless checks and parsed touch PASS; 20 normal-render development frames cover 760x360/844x390/1280x720, zh_CN/en and 100/130%; focused, primary and disabled states are legible Focused Title chrome implementation, capture, plan/checkpoint update and push
DEV-1G Entry-page visual repair COMPLETE The approved Title grammar extends coherently to Boot, first launch, Settings and Credits without a card wall Six DEV-1 headless checks and parsed touch PASS; final 20-frame normal-render development matrix covers all required sizes/languages/scales with strict diagnostics 0 Focused remaining-page styling, test, plan/checkpoint update and push
DEV-1H Revised entry candidate COMPLETE Complete player-order entry path is source-bound and ready for renewed Owner review Source d3b9326; revised DEV-1 candidate suite and complete FG-4 regression PASS with strict diagnostics 0 Candidate evidence closure commit and push

DEV-1 owned paths are runtime/entry/, the minimum integration changes to runtime/main.gd, runtime/main.tscn and runtime/project.godot, DEV-1 tests under runtime/tests/dev1/, and existing plan/checkpoint/ledger sections. Existing hand-drawn entry art may be reused with its provenance unchanged. Later gameplay controllers, domain services, content catalogs, Save-v11 rules, battle assets and later-page implementation are preserved and outside this step.

DEV-1 verification increases in cost: static parse/resource checks, the active substep's focused headless behavior test, applicable repository regression, then normal-render runtime and real parsed input. Formal fresh import/capture attempts use tools/run_test_attempt.py with exact required markers. Candidate evidence is bound to a clean committed source; dirty-tree captures are guidance only.

DEV-1 final handoff is exactly ABG DEV-1 candidate ready. It requires focused pushed commits, HEAD == upstream, a clean worktree, no owned test processes, and aligned plan/checkpoint/ledger facts. Human visual/comprehension, device, package, store and release gates remain OPEN.

DEV-1 Issue Register

ID Status Boundary
DEV1-I01 RESOLVED_DEV-1A A bounded entry controller now owns Boot while the historical gameplay facade remains preserved and hidden during normal launch.
DEV1-I02 RESOLVED_REUSE Existing p9_3_guild_hall.png is a usable hand-drawn entry background; no DEV-1 paid image request is currently required.
DEV1-I03 OPEN_HUMAN_REVIEW_REVISED Owner rejected the earlier generic DEV-1 chrome. The revised source-bound runtime repairs that boundary locally; only renewed Owner review can accept it.
DEV1-I04 RESOLVED_DEV-1F The Title now carries deep-navy enamel, ivory controls, warm-gold layered trim, cut corners, heraldry, compact icon framing and authored focused/primary/disabled states at actual play size. Remaining entry pages are bounded by DEV-1G.
DEV1-I05 RESOLVED_DEV-1G Boot, first launch, Settings and Credits now reuse the Title's enamel, ivory, gold-trimmed, heraldic and icon-forward controls without changing real behavior.

DEV-2 Durable Implementation Plan

Player-visible stopping point:

launch and reach the real Title
-> choose New Game
-> receive four stable generated Recruit offers in fixed positions
-> inspect each Profession, attributes and one Trait benefit/cost
-> compare any two without changing selection
-> select exactly two and review the opening Expedition Company
-> revise or atomically create the persistent opening roster
-> restart at any pre-commit point without rerolling offers or losing selection
-> stop at a truthful Guild handoff; Market and Contract play remain DEV-3+

Only one row may be IN_PROGRESS. A failed check keeps that row active.

Substep Status Owned product boundary Narrow acceptance check Commit boundary
DEV-2A Continuous-growth authority correction COMPLETE Six current documents consistently define one persistent Expedition Company, Contract expeditions that retain roster growth, non-permanent zero-HP recovery at the next stage, and the exact DEV-2 stopping point Cross-document terminology/semantic scan has no authoritative permanent-death or Contract-roster-discard promise; legacy runtime divergence is explicitly frozen; consolidated design and Recruit-loadout validators pass Focused authority/checkpoint/ledger correction and push
DEV-2B Stable initial draft PENDING Seed reservation, exactly four valid generated Recruit records, stable IDs/order, selected-ID persistence, retry-safe load and invalid/corrupt recovery Focused service/store tests cover create/load/restart, no reroll, duplicate/missing offer rejection and no pre-commit roster Focused draft service, persistence tests, plan/checkpoint update and push
DEV-2C Recruit choice PENDING Title reaches one real Japanese-fantasy selection page with four equal offers, Inspect, Compare A/B, independent Select toggles and exact 0/2..2/2 gating Headless state/navigation tests plus real parsed touch; 760x360/844x390/1280x720, zh_CN/en and 100/130% normal-render fixtures Focused selection UI, tests, plan/checkpoint update and push
DEV-2D Review, commit and handoff PENDING Review names selected two and unchosen two; Revise preserves the draft; Create atomically writes one persistent roster and reaches a truthful Guild boundary without implementing DEV-3 Duplicate commit is idempotent; interruption/restart resumes draft or committed Guild handoff exactly; no unchosen Recruit or Contract reset is stored Focused commit/handoff implementation, tests, plan/checkpoint update and push
DEV-2E Initial Company candidate PENDING Complete Boot -> setup if needed -> Title -> choose-two -> review -> restart/revise/create -> Guild handoff path is source-bound DEV-2 suite, applicable DEV-1 and preserved FG-4 regressions, strict diagnostics 0 and actual-size agent visual review Candidate evidence closure commit and push

DEV-2 owns runtime/entry/, new DEV-2 tests under runtime/tests/dev2/, the minimum main-entry integration and existing authority/checkpoint/ledger sections. It reuses the accepted entry background and procedural chrome; no new image generation is currently justified. Guild/Market, formation, battle, legacy casualty conversion, economy retuning, later Regions and delivery remain outside DEV-2.

DEV-2 final handoff is exactly ABG DEV-2 candidate ready. It requires focused pushed commits, HEAD == upstream, a clean worktree, no owned test processes and aligned plan/checkpoint/ledger facts. Human comprehension/choice/visual, device, package, store and release gates remain OPEN.

DEV-2 Issue Register

ID Status Boundary
DEV2-I01 RESOLVED_DEV-2A All six current authorities now place the persistent Company/knockout rule first; contradictory FG-4 lifecycle details are explicitly frozen legacy evidence and current New Game/Company pages use the corrected model.
DEV2-I02 FROZEN_LATER_CONVERSION Existing FG-4 gameplay and Save-v11 implement the superseded casualty/run model; DEV-2 preserves them and does not claim they already satisfy the new authority.
DEV2-I03 RESOLVED_REUSE Accepted painted entry art plus procedural enamel/ivory/gold chrome cover the initial-company page; no paid generation is required.

Aetherbound Guild Development Ledger

This append-only ledger records current milestone decisions. Superseded test attempt details remain available in Git history and external evidence roots; they are not duplicated here.

2026-08-16 - Complete Game Program Approved

  • The Owner approved plan-driven development toward eight Regions, sixteen Contracts, the First Chronicle ending and Oath Season postgame.
  • Single-session implementation is required. Native delivery remains separate.

2026-08-17 - Functional Midpoint Reached

  • RG-03 and RG-04 gameplay, four Contracts, four Bosses, Save-v11 and five profiles reached a combined automated/simulation/runtime midpoint candidate.
  • Runtime-only Web and Gallery transport passed.
  • These checks did not grant human visual, comprehension, fun or listening acceptance.

2026-08-17 - Pixel Presentation Rejected

  • Owner reviewed Gallery item 5vl2EqsYWkey and rejected its 16-bit pixel-art presentation.
  • Accepted authority remains Japanese 2D thin-line/cel-color characters over softly painted environments. Unit scale follows the accepted reference and Owner review of a real runtime frame, not a fixed percentage.
  • Root cause: FG-2 incorrectly treated a generic portfolio pixel-art rule as stronger than this game's approved product and visual contracts.
  • Gameplay, content, Save, input and event-bound audio remain preserved.
  • FG-5, bulk asset production and delivery are frozen.

2026-08-17 - Planning And Documentation Reset

  • The repository-specific Agent contract now excludes Aetherbound from the portfolio pixel-art rule.
  • Current product documentation is exactly the six numbered Markdown files under docs/; README.md is the clickable review index.
  • Historical milestone Goals, plans, H5 notes, attempt reports and rejected generation Markdown are removed from the current tree; Git history retains them.
  • All 50 canonical pages and the implementation order are defined in 03_PAGES_AND_UX.md.
  • The planning freeze ended when the Owner approved entry-first development. The first product action is the real Boot, first-launch and Title path.
  • Consolidation validation proves both page authorities contain the same 50 IDs and retains 42 Professions, 36 Traits, 320 equipment items, 60 Artifacts, 100 enemies, 16 Bosses and 8 Regions.

2026-08-17 - Entry-First Development Order Approved

  • The Owner rejected a battle-first implementation order and required the game to be developed in the same sequence a new player experiences it.
  • DEV-1 begins with Boot, first-launch setup, Title and essential Settings.
  • Initial Company, Guild, Market, automatic formation, risk, battle, result and later content remain ordered follow-up steps rather than parallel work.
  • One HAPI development session works directly in the current repository and branch; no multi-Agent or new worktree is authorized.

2026-08-17 - DEV-1A Boot Implemented

  • Normal launch now enters a bounded Godot entry shell over the existing hand-drawn guild threshold; the historical gameplay facade remains preserved.
  • Boot verifies required local resources, user storage, settings data and the current save envelope with named progress. Unsafe data is not reset.
  • Recoverable failure retains the branded screen, a stable non-path diagnostic and Retry. Focused service and real-main-scene tests pass with strict markers ABG_DEV1_BOOT_OK and ABG_DEV1_BOOT_RUNTIME_OK.

2026-08-17 - DEV-1B First Launch Implemented

  • A clean profile must apply language, safe-area, 100/115/130% text, reduced motion/flashes and caption settings before reaching Title.
  • Settings use schema validation and verified candidate/backup promotion. A completed setup survives restart and is not forced again; the dedicated reset fixture restores first-launch behavior without changing run data.
  • The focused persistence and screen-state marker ABG_DEV1_FIRST_LAUNCH_OK and the Boot handoff regression pass.

2026-08-17 - DEV-1C Title Implemented

  • The Title uses the approved hand-drawn guild threshold and keeps New Game, Continue, Settings and Credits in a compact operational hierarchy.
  • Continue reflects the verified local-save envelope and shows a visible reason when disabled. New Game and Continue preserve frozen DEV-2/save semantics and state the unavailable downstream review instead of silently doing nothing.
  • Settings/Credits return exactly to Title, while Back opens a non-destructive exit review. ABG_DEV1_TITLE_OK passes with no-save and resumable fixtures.

2026-08-17 - DEV-1D Settings And Credits Implemented

  • SET-001..006 provide stable category navigation and persisted display/text, motion/flash/power, audio/caption, control and locale behavior with explicit Apply/Discard handling and exact Title return.
  • Master mute plus Music/SFX levels apply to real Godot buses. Absent separate Ambience/Voice assets and conflict-safe remapping remain visible with precise unavailable reasons rather than inert controls.
  • SYS-008 names authorship, bundled SIL font license, offline behavior, local data/privacy and the no-external-link policy. ABG_DEV1_SETTINGS_OK and the preceding Title/first-launch regressions pass.

2026-08-17 - DEV-1 Candidate Verified

  • Candidate product source: 43e5bacbf88953ae99676400532189f02143be20.
  • Source-bound DEV-1 attempt: /private/tmp/aetherbound-guild-test-evidence/dev1-candidate/20260817T072754Z-25813; PASS with archive SHA-256 46fedaed64f73519ad0e55aa5094cecbebe95e8c94a1e2fb6ab58a3b5356e947. It covers fresh import, Boot/setup/Title/Settings/Credits behavior, real parsed ScreenTouch, restart/reset, 20 normal-render frames, 760x360/844x390/ 1280x720, English/Chinese and 100/130% text with strict diagnostics 0.
  • Source-bound preserved-system regression: /private/tmp/aetherbound-guild-test-evidence/dev1-fg4-regression-3/20260817T081131Z-73011; PASS with archive SHA-256 6b1dd652e8351d1f4f993dc26cd4a6650513e1905bf626d4a9ff20411864738f. All 30 Godot checks, five Python/media checks, 24 legacy normal-render frames, Save-v11, five-profile midpoint simulation and existing parsed-touch paths remain unchanged and strict.
  • Local runtime inspection at the time reached agent_visual_reviewed: the painted guild threshold was visible, focal order was clear and tested short/130% states did not overlap. This agent-only visual state was later superseded by the Owner's rejection below.
  • DEV-1 used existing approved-direction art and made zero paid image calls. Device, packaging, signing, store and release gates remain open and were not exercised.

2026-08-17 - DEV-1 Entry Chrome Rejected

  • The Owner compared the functional DEV-1 runtime with 05_battle_ui_japanese_scale.png and reported a substantial mismatch.
  • Earliest failed boundary: the runtime used generic translucent rectangles and flat settings controls instead of the reference's deep-navy enamel, ivory control plates, warm-gold layered trim, cut corners, heraldic anchors, icon-forward hierarchy and authored interaction states.
  • The functional DEV-1 and preserved FG-4 evidence remains valid only for its named behavior. DEV-1F reopens visual production at the Title fixed frame; Boot, first launch, Settings and Credits follow only after that grammar works at actual play size. No paid generation is yet justified because approved background art and procedural Godot chrome cover this bounded repair.
  • Human visual/comprehension, device, packaging, store and release gates remain open. DEV-2 and later product implementation remain frozen.

2026-08-17 - DEV-1F Title Chrome Repaired

  • The fixed Title frame now uses a heraldic enamel guild header, ivory utility bridge and bottom command dock with layered gold trim, cut corners and procedural icons. New Game receives initial focus; Continue keeps a distinct disabled state and visible reason.
  • All six DEV-1 headless checks, including real parsed ScreenTouch, pass. A normal-render development matrix produced 20 frames at 760x360, 844x390 and 1280x720 across English/Chinese and 100/130% text with strict diagnostics 0.
  • Development capture root: /private/tmp/abg-dev1-title-dev-capture-2. It guides the visual repair but is not source-bound candidate evidence and does not grant human acceptance.
  • No image-generation calls were made. DEV-1G may extend only this proven UI grammar to Boot, first launch, Settings and Credits after DEV-1F is pushed.

2026-08-17 - DEV-1G Entry Pages Repaired

  • Boot now uses a compact verification deck with named progress; first launch uses an edge-attached setup panel; Settings and Credits use one enamel workspace with ivory selected segments, value plates and explicit toggle, slider, disabled and primary-command states.
  • The first development render exposed a setup GridContainer width regression and incorrect unselected-tab text state. That attempt was rejected; the controls gained real expansion sizing, the state binding was repaired and the same matrix was rerun.
  • Final local checks: all six DEV-1 headless markers, real parsed ScreenTouch and 20 normal-render frames pass with strict diagnostics 0. Development capture root: /private/tmp/abg-dev1-entry-pages-dev-capture-5.
  • The development capture is not source-bound candidate evidence and does not grant Owner visual acceptance. No image-generation calls were made; DEV-2 and all human/device/package/release gates remain open.

2026-08-17 - Revised DEV-1 Candidate Verified

  • Candidate product source: d3b93268776b87b7bfb407c79165011facc438f8.
  • Revised source-bound DEV-1 attempt: /private/tmp/aetherbound-guild-test-evidence/dev1-candidate-revised/20260817T091805Z-82593; PASS with archive SHA-256 980ad2fc2e2b0b667e2d2a46dec84320aaca6959755f1f524ee8e5835c85451e. It covers six headless behavior checks, real parsed ScreenTouch and 20 normal-render frames across 760x360/844x390/1280x720, English/Chinese and 100/130% text with strict diagnostics 0.
  • Revised preserved-system regression: /private/tmp/aetherbound-guild-test-evidence/dev1-fg4-regression-revised/20260817T092203Z-85944; PASS with archive SHA-256 d54963d22411165ad97ea31d256292701670a7cddd339f43d672c88e1b53f56a. All 30 Godot checks, five Python/media/audio checks, Save-v11, five-profile midpoint simulation, parsed touch and 24 legacy normal-render frames remain unchanged with strict diagnostics 0.
  • Source-bound entry frames were inspected at actual play size and reach only agent_visual_reviewed: enamel/ivory/gold hierarchy is coherent, selected, focused, disabled and value states are readable, and required 130% layouts do not overlap. Human visual/comprehension acceptance remains open.
  • No paid image calls were made. Device, packaging, signing, store and release gates remain open and were not exercised.

2026-08-17 - Continuous Initial Company Model Approved

  • The Owner confirmed Aetherbound Guild as a continuous-growth incremental game. New Game's four-offer choose-two draft establishes the persistent initial Expedition Company.
  • A Contract run remains a bounded expedition, but Contract completion retains recruited characters, individual levels/XP, Professions, Traits, equipment and the continuing roster instead of archiving and replacing it.
  • Zero HP removes a Recruit only from the current battle. That Recruit remains owned and returns for the next stage; permanent death and death-funded Replacement Credit are superseded product rules.
  • DEV-2 implements only initial-company creation. Existing FG-4 casualty, run closure, economy and Save-v11 code remains preserved historical behavior for a later player-order conversion and must not be silently presented as already compliant.

2026-08-17 - DEV-2A Continuous-Growth Authority Corrected

  • All six current authorities now define the one-time four-offer choose-two as creation of the persistent initial Expedition Company, followed by a Guild handoff rather than immediate per-run roster creation.
  • A zero-HP Recruit remains owned, leaves only the current battle and is battle-ready again at the next stage with the same stable identity and growth. The exact returned HP amount is intentionally deferred to the later balance conversion because the Owner approved return, not a specific full-heal value.
  • Contract clear or withdrawal retains Recruits, level/XP, Professions, Traits, equipment, history and persistent resources. Old permanent-death, Replacement Credit, Company-collapse and roster-archive detail is explicitly FROZEN_LEGACY_FG4, not current product authority.
  • ABG_PREDEVELOPMENT_DESIGN_SELF_TEST_OK cases=7, ABG_RECRUIT_LOADOUT_LEVEL_OK and git diff --check pass. No runtime code, Save-v11 behavior, media, package or release state changed in DEV-2A.