hi
game
# PROJECT ROLE You are the lead developer responsible for building a complete commercial-quality indie Steam game currently called: # DON'T LOOK You are responsible for as much of the development process as your environment technically allows. Act as: - Lead gameplay programmer - Unity developer - Multiplayer/network programmer - Technical artist - Blender asset creator - Level designer - UI/UX designer - Audio implementation developer - Optimization engineer - QA engineer - Build engineer - Steam integration developer I am primarily the owner and human playtester. My job should mainly be: 1. Receive a playable build. 2. Play it. 3. Tell you what feels bad, boring, confusing, ugly, broken, or fun. 4. Let you fix and improve it. Do NOT unnecessarily send programming, configuration, debugging, asset placement, or technical work back to me. If your environment gives you access to the required tools, perform the work yourself. --- # MAIN GOAL Build a polished, lightweight, fun Steam game that can realistically be developed by one person using AI assistance. The game must prioritize: 1. Fun 2. Unique gameplay 3. Replayability 4. Multiplayer fun 5. Performance 6. Small storage size 7. Stability 8. Good visuals 9. Expandable maps/content 10. Maintainable code Avoid unnecessary complexity. This is NOT supposed to become a giant AAA project. --- # CORE GAME IDEA DON'T LOOK is a first-person 3D perception/puzzle/comedy game. The main rule is: # OBJECTS CAN CHANGE ONLY WHEN NOBODY IS LOOKING AT THEM. Example: The player sees a chair. They turn around. When they look back, the chair may have: - moved; - rotated; - changed size; - changed material; - turned into another object; - disappeared; - appeared elsewhere; - revealed something; - blocked something; - opened a route; - become part of a puzzle. The player learns to intentionally manipulate what they are looking at. The game should NOT simply be: "Find the anomaly." The player must actively USE visibility as a gameplay mechanic. --- # IMPORTANT GAME IDENTITY Do not turn this into a generic Exit 8-style anomaly game. The goal is NOT simply: "Look around and identify what changed." Instead: The player controls changes by controlling observation. Looking at something prevents it from changing. Looking away allows it to change. This mechanic must be used for: - puzzles; - cooperation; - environmental manipulation; - secrets; - funny situations; - progression; - multiplayer communication. --- # TECHNOLOGY Use: - Unity - C# - Blender where useful - Git - Steamworks Choose an appropriate stable Unity version suitable for releasing a Windows Steam game. Do not change engines unless there is an extremely strong technical reason. --- # AI DEVELOPMENT GOAL Automate as much development as technically possible. Where tools permit, you should: BUILD IT. CREATE ASSETS. WRITE CODE. RUN IT. COMPILE IT. TEST IT. DEBUG IT. PROFILE IT. OPTIMIZE IT. BUILD THE EXECUTABLE. DOCUMENT IT. Then give me something playable. Do not simply tell me: "Create this script." "Add this GameObject." "Drag this into the inspector." if you have the ability to perform those actions yourself. --- # DEVELOPMENT DECISIONS Do not repeatedly stop development to ask minor questions. For minor decisions, choose the option that best satisfies: - simplicity; - performance; - maintainability; - visual quality; - fun; - small storage; - future expansion. Ask me only when a decision significantly changes: - the main gameplay idea; - monetization; - project scope; - legal/licensing matters; - major visual direction; - important player experience. --- # ART DIRECTION Use a clean, attractive stylized 3D art direction. The game should look intentionally designed. It should NOT look like: - default Unity; - random asset-store content; - unfinished prototype; - inconsistent AI-generated assets; - cheap mobile game; - excessive visual noise. Prefer: - clean geometry; - atmospheric lighting; - coherent materials; - good shadows; - subtle post-processing; - strong environmental composition; - visually readable objects. Do NOT chase extreme photorealism. Stylized graphics are preferred because they: - reduce development time; - reduce storage; - improve performance; - make asset creation easier; - allow a stronger visual identity. --- # BLENDER Use Blender whenever useful for creating or modifying: - furniture; - props; - rooms; - doors; - windows; - decorative objects; - simple characters; - puzzle objects; - environmental pieces. Where practical, automate Blender using scripts. Prefer modular assets. Example: Instead of building ten unique walls, create reusable: - wall pieces; - corners; - doors; - trims; - windows; - floors; - ceilings. Reuse them intelligently. --- # THIRD-PARTY ASSETS You may use third-party assets if they are genuinely helpful. However: Do NOT randomly download copyrighted models. Only use assets that allow commercial game use. Maintain an asset/license record containing: - asset name; - source; - creator; - license; - attribution requirements; - modifications made. Never assume downloadable means commercially usable. Prefer self-created or clearly licensed assets. --- # STORAGE SIZE Keep the game reasonably small. Storage efficiency is important. Prefer: - shared textures; - texture atlases where useful; - modular assets; - compressed audio; - reasonable texture resolution; - shared materials; - reusable meshes; - procedural variation; - asset reuse. Avoid: - unnecessary 4K/8K textures; - huge audio files; - duplicated assets; - unused dependencies; - gigantic asset packs; - unused Unity packages. Track actual build size. Before adding a large dependency, evaluate whether its benefit is worth its size. --- # FIRST-PERSON CONTROLLER Create polished first-person controls. Support: - WASD - mouse look - interaction - pause - configurable sensitivity - configurable FOV - invert Y option Jump should only exist if gameplay needs it. Crouch should only exist if gameplay needs it. Do not add mechanics merely because other games have them. Movement should feel: - responsive; - smooth; - predictable; - comfortable. Avoid excessive camera bob. Avoid motion sickness. --- # CORE VISIBILITY SYSTEM This is one of the most important systems in the project. An object must NEVER visibly transform while the player can reasonably see it. Do NOT simply test whether the object's center point is inside the camera. Use appropriate techniques involving: - camera frustum; - renderer bounds; - distance; - occlusion; - walls; - line of sight; - multiple renderers; - screen edges; - grace periods. Objects close to the edge of the player's screen should not obviously pop/change. An object should become eligible for transformation only when it is safely outside player perception. --- # VISIBILITY PERFORMANCE Do NOT run expensive checks against every object every frame. Use efficient strategies such as: - staggered checks; - distance filtering; - active-region filtering; - cached visibility; - event-driven updates; - spatial grouping; - visibility intervals; - broad-phase checks before raycasts. Only use complicated spatial partitioning if measurements show it is necessary. Measure first. Optimize second. --- # CHANGEABLE OBJECT ARCHITECTURE Create a reusable object transformation system. Example architecture: ChangeableObject ChangeState ChangeProfile ChangeManager VisibilityManager WorldStateManager Names may differ if you design something cleaner. Objects should use configuration/data rather than requiring completely custom code for every object. --- # OBJECT STATES A state may define: - model; - position; - rotation; - scale; - material; - active state; - animation; - sound; - collision; - interaction; - connected object state; - puzzle behavior; - rarity. Example: CHAIR_STATE_NORMAL CHAIR_STATE_UPSIDE_DOWN CHAIR_STATE_TINY CHAIR_STATE_CEILING CHAIR_STATE_TABLE CHAIR_STATE_GONE --- # LARGE VARIETY WITHOUT LAG The game should support huge amounts of possible variation. However: # DO NOT CREATE UNLIMITED ACTIVE OBJECTS. "Unlimited changes" means combinations of existing systems, NOT unlimited runtime object spawning. Use: - object states; - procedural combinations; - seeded randomness; - weighted randomness; - prefab reuse; - object pooling; - lightweight state data. Avoid uncontrolled: Instantiate() Destroy() Instantiate() Destroy() loops. Do not continuously create GameObjects. --- # RANDOM EVENTS Create rarity levels. For example: COMMON UNCOMMON RARE VERY RARE SECRET Some players should encounter strange things that other players may not see for many sessions. This creates discussion and replayability. However, randomness must not replace gameplay. --- # GAMEPLAY PHILOSOPHY Randomness creates surprise. Puzzles create purpose. Secrets create curiosity. Multiplayer creates communication. Rare events create replayability. Atmosphere creates identity. Combine these. Do not make the entire experience: "Turn around repeatedly until something funny happens." --- # PUZZLES Build puzzles around observation. Examples: One object must remain visible while another changes. The player must position themselves so: Object A = visible Object B = hidden Object C = hidden A door may only appear after another object changes. A key may keep changing location until watched correctly. A hallway may transform depending on what the player observes. Mirrors may count as observation. Security cameras may count as observation. Windows may allow visibility into another room. Lights may affect whether certain objects behave. --- # MIRRORS Mirrors can eventually become an advanced mechanic. If technically feasible and performant: An object visible through a mirror may count as being observed. Do not build an extremely expensive reflection system just for this mechanic. Choose a performant solution. --- # SECURITY CAMERAS Security cameras may act as artificial observers. Example: A player watches Camera 1. Camera 1 sees the chair. The chair cannot change. This can create advanced puzzles. --- # MAP SYSTEM Maps must be modular. Do NOT hard-code the game around one apartment. Each map should define things such as: - name; - thumbnail; - spawn points; - environment; - changeable objects; - puzzle rules; - special mechanics; - lighting; - sounds; - event pool; - completion requirements; - multiplayer configuration. --- # FUTURE MAP EXPANSION Adding another map should be easy. The ideal workflow should be: 1. Create/open map scene. 2. Build environment. 3. Place reusable objects. 4. Configure change states. 5. Configure puzzles. 6. Configure map-specific events. 7. Register map. 8. Test. 9. Build. Do NOT require rewriting the core visibility/networking systems. --- # INITIAL MAP Do NOT immediately create many maps. Start with one polished vertical-slice map. Suggested setting: A strange apartment/house. The map should be compact but visually polished. Use this map to prove: - core mechanic; - performance; - puzzle design; - object system; - multiplayer; - graphics; - menus; - sound; - replayability. Only build many maps after this one works. --- # MULTIPLAYER The game must support: - Singleplayer - 2-player co-op - 3-player co-op - 4-player co-op Do NOT design for 20, 50, or 100 players. This is small co-op. --- # ZERO MONTHLY SERVER COST This is a strict requirement. Monthly multiplayer infrastructure cost target: # $0 Do NOT depend on: - rented VPS; - dedicated servers operated by us; - paid matchmaking; - paid databases; - Firebase; - Supabase; - AWS; - Azure; - paid Photon hosting; - PlayFab paid infrastructure; - custom backend servers. Use Steam-provided infrastructure wherever appropriate. --- # STEAM MULTIPLAYER Use: - Steamworks; - Steam Lobbies; - Steam networking/P2P; - Steam relay infrastructure where appropriate. One player's PC acts as the HOST. Other players connect to that host. Avoid requiring manual port forwarding. Steam identity should be used. Do not create our own account/password system. --- # MULTIPLAYER ARCHITECTURE Use a host-authoritative architecture. The host controls important shared game state. Clients should NOT independently decide gameplay-critical world changes. Example: BAD: Client 1 decides chair becomes lamp. Client 2 decides chair disappears. GOOD: Host decides: Chair_17 State 2 → State 5 All clients receive the same result. --- # NETWORK EFFICIENCY Do NOT synchronize every object every frame. Prefer synchronized events. Examples: ObjectChanged DoorOpened PuzzleCompleted PlayerInteracted MapStarted MapCompleted SecretTriggered Only player movement and genuinely dynamic objects should need frequent updates. Static environment objects should consume effectively zero continuous network traffic. --- # MULTIPLAYER VISIBILITY The core multiplayer rule is: # AN OBJECT MAY CHANGE ONLY WHEN NO ACTIVE PLAYER CAN SEE IT. Example: Dan cannot see chair. Alex can see chair. Chair stays unchanged. Later: Dan cannot see chair. Alex cannot see chair. Chair may now change. The host makes the authoritative decision. --- # MULTIPLAYER VISIBILITY PERFORMANCE Do NOT send information about every object from every client every frame. Build an efficient solution. Use techniques such as: - distance filtering; - nearby/relevant object lists; - staggered checks; - compact visibility data; - host-side validation; - event-based updates; - bitmasks where useful; - spatial regions if necessary. The solution must scale comfortably to four players. --- # MULTIPLAYER GAMEPLAY Multiplayer must provide gameplay value. Do NOT simply add player models to singleplayer. Examples: Player A must watch an object. Player B walks through a door that would disappear if unobserved. Two players must look away simultaneously. Players communicate from separate rooms. Player A watches a security camera. Player B manipulates an object. Players accidentally prevent changes because one person keeps looking. The game should naturally create funny voice-chat moments. --- # PLAYER MODELS Create lightweight multiplayer player representations. Support: - movement; - looking direction; - Steam name; - basic animations; - interaction animation where useful. Do not use expensive character systems unnecessarily. --- # PLAYER COLLISION Avoid players trapping each other. If normal collision creates frustrating situations: Allow simplified collision or player pass-through. Gameplay matters more than realism. --- # MULTIPLAYER LOBBY PLAY menu should include: SINGLEPLAYER HOST GAME JOIN GAME Hosting creates a Steam lobby. Lobby should display: - player names; - Steam avatars where practical; - host indicator; - ready states; - selected map; - player count; - lobby privacy; - invite button; - leave button; - start game button. --- # LOBBY PRIVACY Support appropriate options such as: PUBLIC FRIENDS ONLY INVITE ONLY --- # STEAM INVITES Allow players to invite Steam friends. Support joining through Steam invitations where supported. Handle: - lobby full; - lobby deleted; - host disconnected; - wrong game version; - connection failure; - player leaving; - invitation failures. Do not display raw developer/network errors to normal players. --- # MULTIPLAYER MAP LOADING The host selects the map. When START is pressed: 1. Lock game start appropriately. 2. Inform clients which map is loading. 3. Load map. 4. Wait for required clients. 5. Synchronize initial world state. 6. Spawn players. 7. Begin gameplay. Do not allow the world to start changing while another required player is still loading unless specifically handled. --- # RANDOMNESS IN MULTIPLAYER Gameplay-critical randomness must stay synchronized. Do NOT allow every client to independently roll random transformations. Use: - host-selected results; - synchronized random seeds; - deterministic systems where useful. Everyone must see the same gameplay world. --- # JOIN-IN-PROGRESS Do NOT prioritize join-in-progress for version 1. Players should normally join before the map begins. Later, if join-in-progress is implemented, synchronize: - map; - players; - object states; - doors; - puzzles; - progression; - event state; - random seed. --- # DISCONNECTS If a CLIENT disconnects: - remove them cleanly; - remove their visibility contribution; - ensure puzzles cannot become permanently broken. If the HOST disconnects: For version 1, display: Host disconnected. Then safely return remaining players to lobby/menu. Host migration is NOT required initially. Do not spend huge development effort on it before the core game works. --- # MAIN MENU Create a polished main menu. Include: PLAY MAP SELECT SETTINGS CREDITS QUIT PLAY should lead to appropriate singleplayer/multiplayer choices. The menu should feel like part of the game's visual identity. Use: - subtle animation; - transitions; - hover effects; - sound; - atmospheric background. Do not use obvious default Unity UI. --- # PAUSE MENU ESC opens pause menu. Include: RESUME RESTART MAP SETTINGS RETURN TO MAIN MENU EXIT TO DESKTOP Exit must actually work correctly in builds. If an action destroys meaningful progress, use confirmation where appropriate. --- # SETTINGS Create a real settings menu. Do not create fake options. Every setting shown must actually work. --- # GRAPHICS SETTINGS Include appropriate options such as: - display mode; - resolution; - VSync; - FPS limit; - quality preset; - shadows; - anti-aliasing; - texture quality; - FOV; - relevant visual effects. Suggested FPS limit options: 30 60 120 144 165 240 Unlimited --- # AUDIO SETTINGS Include: MASTER MUSIC SFX UI Optional: AMBIENCE All audio settings must save. --- # CONTROL SETTINGS Include: - mouse sensitivity; - invert Y; - useful key rebinding if practical. Do not build a giant input configuration system unless necessary. --- # SETTINGS STORAGE Settings must persist after restarting the game. Handle invalid/missing configuration gracefully. --- # SAVE SYSTEM Save appropriate information such as: - settings; - unlocked maps; - progress; - secrets; - achievements/statistics; - relevant collectibles. Use a versioned save format. Future updates should not unnecessarily destroy old saves. Handle corrupted saves gracefully. --- # AUDIO Implement: - ambience; - footsteps; - environment sounds; - interaction sounds; - transformation sounds; - menu sounds; - puzzle sounds. Do not make every change loud. Some transformations should happen silently. Silence can make discoveries more effective. --- # MUSIC Use music sparingly. The atmosphere should not constantly overwhelm the player. Prefer environmental sound where appropriate. --- # PERFORMANCE TARGET Target stable gameplay on ordinary gaming PCs. Aim for: 1080p 60 FPS on reasonable mid-range hardware. Do NOT optimize only for expensive computers. --- # PERFORMANCE RULES Avoid: - every-object Update(); - huge numbers of raycasts; - continuous Instantiate/Destroy; - memory leaks; - runaway coroutines; - endless async tasks; - duplicate managers; - unnecessary realtime lights; - giant textures; - expensive physics everywhere; - huge garbage collection spikes; - repeated log spam. --- # PROFILING Use available profiling tools. Measure: - CPU; - GPU; - memory; - garbage collection; - draw calls; - batches; - network traffic; - load times; - build size. Optimize based on actual measurements. --- # OBJECT POOLING Use pooling where objects are frequently reused. Do NOT pool everything blindly. Only use it where useful. --- # LIGHTING Create attractive lighting while preserving performance. Prefer baked/mixed lighting where appropriate. Avoid excessive realtime shadow-casting lights. Atmosphere is important, but performance is more important than technically impressive lighting. --- # LOADING Create clean loading transitions. Avoid frozen-looking screens. Maps should unload correctly. Unused resources should not remain permanently loaded. Repeated: Map → Menu → Map → Menu must not continuously increase memory. --- # STEAM INTEGRATION Eventually implement appropriate: - Steam initialization; - Steam lobby integration; - Steam invites; - Steam networking; - Steam achievements; - Steam overlay compatibility; - statistics if useful. Steam failure should be handled gracefully. The development version should still support reasonable offline/local testing where practical. --- # STEAM ACHIEVEMENTS Use achievements carefully. Examples could reward: - completing maps; - discovering secrets; - triggering rare events; - solving unusual puzzles; - multiplayer actions. Avoid meaningless achievements every thirty seconds. --- # OFFLINE DEVELOPMENT Do not make development completely dependent on Steam being available. Separate Steam-specific code behind clean interfaces. Allow: - singleplayer local testing; - test networking where practical; - simulated clients where useful. --- # CODE ARCHITECTURE Keep systems modular. Do NOT create one enormous script controlling everything. Separate responsibilities. Possible systems: GameManager VisibilityManager WorldStateManager ChangeManager PuzzleManager MapManager SaveManager SettingsManager AudioManager SteamLobbyManager NetworkSession PlayerNetworkController VisibilityAuthority MapSessionManager Names may differ if a better architecture is created. --- # CODE QUALITY Use: - clear naming; - sensible folders; - appropriate interfaces; - reusable components; - data-driven systems; - events where appropriate. Avoid: - unnecessary static globals; - giant scripts; - duplicate logic; - unnecessary packages; - abandoned experiments; - deeply coupled systems. --- # EVENT CLEANUP Properly: - unsubscribe events; - stop coroutines; - dispose resources; - destroy network state; - unload scenes; - close Steam/session callbacks where necessary. Do not allow repeated sessions to accumulate hidden listeners. --- # LOGGING Development logs may be detailed. Release logs should not spam. Never allow an error to print thousands of times per second. Rate-limit repeated non-critical logging. Fix root causes instead of hiding important errors. --- # DEBUG TOOLS Create developer-only tools. Examples: FORCE OBJECT CHANGE DISPLAY VISIBILITY STATE SHOW OBJECT ID FREEZE CHANGES SET RANDOM SEED TELEPORT TO PUZZLE UNLOCK MAP SHOW FPS SHOW MEMORY SHOW NETWORK TRAFFIC SHOW ACTIVE CHANGEABLE OBJECT COUNT These tools must not appear in normal gameplay. --- # AUTOMATED TESTING Create automated tests for systems that can reasonably be tested. Examples: - save/load; - settings; - object states; - deterministic randomness; - map registration; - puzzle state; - visibility calculations; - multiplayer state serialization; - invalid configuration; - scene transitions. --- # MULTIPLAYER TESTING Test multiplayer as extensively as technically possible. Test: - creating lobby; - joining lobby; - leaving; - invitations where testable; - two players; - three players; - four players; - loading maps; - synchronized spawning; - movement; - visibility; - object changes; - puzzle synchronization; - simultaneous interactions; - client disconnect; - host disconnect; - restart; - return to menu; - repeated lobby creation; - repeated map loading; - network delay; - packet loss where tooling permits. --- # MULTIPLAYER FAILURE CHECKS Specifically look for: - duplicate players; - duplicate objects; - desynchronization; - wrong object states; - objects changing while visible; - RPC/message spam; - reconnect loops; - memory growth; - console spam; - incorrect player visibility; - mismatched random events; - stuck loading screens. --- # SECURITY This is cooperative multiplayer. Use sensible host-side validation. Do not blindly trust clients. Examples: Client should not simply be able to send: COMPLETE_MAP CHANGE_ALL_OBJECTS UNLOCK_EVERYTHING without host validation. However: Do NOT waste development time building enterprise-grade anti-cheat for a small cooperative indie game. --- # BUILD PROCESS Produce Windows builds. Before handing me a build: 1. Compile. 2. Fix compile errors. 3. Check console. 4. Fix important warnings. 5. Run automated tests. 6. Launch game. 7. Test menus. 8. Test gameplay. 9. Test multiplayer where possible. 10. Profile. 11. Fix major problems. 12. Create build. --- # NO FALSE CLAIMS Never say: "Everything is fully tested." unless it genuinely was. Clearly separate: AUTOMATICALLY VERIFIED TESTED IN RUNNING GAME CODE-REVIEWED ONLY REQUIRES HUMAN TESTING Do not claim zero bugs. --- # HUMAN TESTING My testing should primarily focus on things AI cannot judge perfectly: - Is the game fun? - Is movement comfortable? - Are graphics attractive? - Is the atmosphere good? - Is something boring? - Is a puzzle confusing? - Is a transformation surprising? - Is multiplayer funny? - Would I keep playing? - Would I buy this? --- # DEVELOPMENT MILESTONES Do NOT attempt everything simultaneously. --- # MILESTONE 1 — TECHNICAL PROTOTYPE Build: - Unity project; - Git repository; - project structure; - first-person controller; - simple test room; - several objects; - visibility system; - object state system; - basic interactions; - random seed system; - debug tools; - automated tests. Goal: Prove the mechanic works. --- # MILESTONE 2 — FIRST REAL MAP Create a polished vertical slice. Approximately: 10–20 minutes of gameplay. Include: - proper environment; - good lighting; - several puzzles; - interesting transformations; - sounds; - secrets; - rare events; - basic ending; - menu; - settings; - pause system; - saves. Goal: Make it feel like a real game. --- # MILESTONE 3 — MULTIPLAYER FOUNDATION Make existing architecture network-ready. Implement: - Steam integration; - lobby creation; - lobby joining; - host/client; - two-player networking; - synchronized movement; - synchronized world states; - synchronized transformations; - multiplayer visibility. Goal: Prove the mechanic works correctly with two players. --- # MILESTONE 4 — FOUR PLAYER CO-OP Expand to: - 3 players; - 4 players; - Steam invitations; - ready system; - lobby UI; - map selection; - disconnect handling; - multiplayer-specific puzzles. Goal: Stable 1–4 player gameplay. --- # MILESTONE 5 — OPTIMIZATION REVIEW Profile everything. Focus on: - Update usage; - visibility queries; - raycasts; - network traffic; - allocations; - GC; - memory; - loading; - textures; - audio; - build size; - draw calls. Fix measured issues. --- # MILESTONE 6 — CONTENT EXPANSION Only after the first map and multiplayer work: Create additional maps. Use reusable architecture. Each map should introduce something interesting instead of simply changing wall textures. Potential environments could include: - apartment; - office; - hotel; - supermarket; - warehouse; - school; - museum; - strange laboratory. Do not create all of these automatically. Choose maps that actually produce good gameplay. --- # MILESTONE 7 — REPLAYABILITY Improve: - random events; - rare events; - secrets; - alternate puzzles; - transformation combinations; - map variations; - achievements. Aim for replayability without enormous asset requirements. --- # MILESTONE 8 — POLISH Improve: - animations; - UI; - audio; - transitions; - graphics; - feedback; - accessibility; - loading; - controller feel; - multiplayer feedback. --- # MILESTONE 9 — STEAM RELEASE PREPARATION Prepare: - Steam integration; - achievements; - production build; - credits; - license list; - save migration; - final settings; - release configuration; - clean build; - final QA. Do not publish or make irreversible Steam/account actions without my approval. --- # MAIN MENU QUALITY The main menu should feel polished enough that the game feels legitimate immediately after launch. Possible background: A room where subtle things change whenever the menu camera looks away. This could demonstrate the game's mechanic before gameplay even begins. Keep it performant. --- # EXIT TO DESKTOP QUIT / EXIT TO DESKTOP must: - close the game cleanly; - disconnect network sessions; - save settings; - release resources; - close correctly in the Windows build. Do not test Application.Quit only inside the editor and assume it works. Test the actual build. --- # CRASH RESISTANCE Handle reasonable failure conditions: - missing save; - corrupt setting; - network disconnect; - invalid map; - missing optional asset; - failed Steam initialization; - lobby disappearance. Do not let recoverable failures unnecessarily crash the game. --- # VERSIONING Maintain project version numbers. Keep network compatibility in mind. Players on incompatible versions should receive a clear message rather than entering a broken lobby. --- # SOURCE CONTROL Use Git properly. Create sensible commits after meaningful stable milestones. Example: core visibility prototype object state system first puzzle settings menu steam lobby prototype multiplayer visibility Do not create useless commits every few seconds. Never commit: - Library; - Temp; - Logs; - generated builds; - secrets; - unnecessary caches. --- # BACKUPS Keep the project recoverable. Before large architectural changes, create a safe Git checkpoint. If a change causes serious regression, investigate or revert instead of stacking fixes onto broken architecture. --- # DEVELOPMENT PRINCIPLE Never continue building large systems on top of known major bugs. Use: # BUILD → RUN → TEST → PROFILE → FIX → CONTINUE not: BUILD → BUILD → BUILD → BUILD → DISCOVER EVERYTHING IS BROKEN. --- # PERFORMANCE PRINCIPLE Do not prematurely create extremely complicated optimization systems. First create a good architecture. Then measure. Then optimize real bottlenecks. --- # CONTENT PRINCIPLE More content does not automatically mean a better game. One excellent map is more valuable than ten boring maps. Do not mass-generate low-quality content simply because AI can create it quickly. --- # GRAPHICS PRINCIPLE Good art direction is more important than maximum graphical complexity. Make the game: clean, recognizable, atmospheric, consistent, performant. --- # PLAYER EXPERIENCE Avoid frustrating mechanics such as: - excessively slow movement; - extremely dark rooms where nothing can be seen; - constant jumpscares; - unfair transformations; - unclear objectives; - puzzles requiring random guessing; - excessively long loading; - transformations that obviously happen on screen; - multiplayer puzzles permanently broken by disconnects. --- # TUTORIAL Teach the core mechanic naturally. Avoid a giant wall of text. Example: Player sees door locked. Player sees strange chair. Text: "Look away." Player turns away. Returns. Chair changed. The game teaches through interaction. --- # ACCESSIBILITY Where practical support: - subtitles; - volume controls; - sensitivity; - FOV; - brightness; - motion reduction; - readable UI. Do not make essential information depend only on color. --- # FINAL BUILD REPORT Before asking me to test a build, provide: ## BUILD STATUS Playable / Not Playable ## VERSION Current version. ## IMPLEMENTED Major completed features. ## AUTOMATED TESTS Passed/failed tests. ## RUNTIME TESTS What was actually launched and tested. ## MULTIPLAYER TESTS Number of clients tested and results. ## PERFORMANCE Actual measurements where available: FPS CPU GPU RAM GC network traffic ## BUILD SIZE Actual Windows build size. ## KNOWN ISSUES Anything known to still be broken or uncertain. ## HUMAN TESTS REQUIRED Specific things you want me to judge. Keep this report concise. --- # IMPORTANT LIMITATION If your environment prevents you from directly: - launching Unity; - controlling Blender; - running the game; - testing Steam networking; - producing a Windows build; do NOT pretend these actions happened. Complete everything that is possible. Clearly state what could not be executed. Design the project so that future agents/tools with access can continue easily. --- # FIRST TASK Begin with: # MILESTONE 1 — TECHNICAL PROTOTYPE First inspect the available machine/environment. Determine: - operating system; - installed Unity versions; - Unity Hub availability; - Blender availability; - Git availability; - .NET/C# environment; - available disk space; - project permissions; - any existing project files. If Unity is not installed and your environment allows installing required development tools, install/configure an appropriate stable Unity version. Then: 1. Create the Unity project. 2. Initialize Git. 3. Create proper project folders. 4. Build the first-person controller. 5. Build a small test environment. 6. Build the visibility system. 7. Build the object-state/change system. 8. Add several test objects. 9. Add debugging tools. 10. Create automated tests. 11. Compile. 12. Run. 13. Test. 14. Fix errors. 15. Profile the basic system. 16. Produce the first playable development build if technically possible. Do NOT begin building many maps yet. Do NOT spend time on Steam multiplayer yet. First prove that the core mechanic is: - reliable; - fun enough to test; - expandable; - performant. Once Milestone 1 is stable, continue to Milestone 2 unless a major design decision genuinely requires my input. Remember: - My primary role is to PLAY and TEST. - Your primary role is to BUILD. - Make a file and call it CURRENT_STAGE, keep locating and saying where you reached in development so If you ever encountered an usage limit, or stopped, so you know where to continue from