6741

What could make 6741 boring, unfair, confusing, or unbuildable.

• Critical Risk
  Two players + a comms channel is the death of fear. Horror works on isolation. The moment two friends can see each other and talk, dread collapses into a buddy stroll. The whole design must manufacture separation — or it isn’t a horror game, it’s a co-op errand.
• Critical Risk
  One zombie cannot carry a whole session. A single enemy is either trivially avoidable (boring) or corner-camps the only key room (frustrating). Players will reverse-engineer a waypoint AI in ten minutes, and a monster you’ve solved is furniture.
• Critical Risk
  Fear that only triggers “when the zombie is visible” is off 90% of the time. This makes fear a binary toggle bolted to your rarest entity. The empty house — which players spend most of their time in — has no tension system at all.
• High Risk
  Server-authoritative AI with no NavMesh will feel dumb or unfair. Waypoints + line-of-sight over WebSocket, interpolated on the client, tends to read as a stuttering monster that either teleports onto you or gets stuck on a chair. The zombie’s feel is the single biggest build risk — prototype it in a gray box before any house art.
• High Risk
  “Dark + maze-like + no map” produces disorientation, not dread. Getting lost without progress is the #1 rage-quit cause in horror. You need legibility anchors (landmarks, a home base) so players feel hunted, not confused.
• High Risk
  The win condition is a flat fetch quest. Two keys, one door, no decisions. There’s no moment where the team must choose between safety and progress — and that choice is where co-op horror lives.
• Cannot Confirm
  Fog + dynamic flashlights + post-FX is exactly what tanks WebGL. You may be authoring atmosphere you can’t ship at 60 FPS. Stand up a render-budget test scene early — this is a Cannot confirm until measured in Chrome on the real target laptop.

Higher tension without growing the build — most reuse systems already planned.

• Recommendation
  Make the zombie a slow, relentless stalker — not a sprinter. Telegraphed, unhurried, impossible to permanently outrun but evadable with planning. This is cheaper to build (no twitch netcode), far scarier, and removes the “juke the dumb AI” failure. Dread > reflexes.
• Recommendation
  Detection by sound, not perfect sight. Sprinting, dropping items, and the loud key door make noise that draws it. Now movement itself is a tension dial the player controls — and it ties straight into the fear loop (fear slows you and makes you louder).
• Recommendation
  Put a cost on Key B. Retrieving it should require a loud, committing action — turn on the power, pry open the storage door — that wakes the zombie. One clean decision (“do we make noise to win?”) transforms a fetch quest into a heist.
• Recommendation
  Promote the Director to a real pillar. Whispers, flickers, distant footsteps, a door slamming in an empty room — this is the cheapest scares-per-dollar in the whole design and it makes the house scary when the monster is gone. It’s listed as “atmosphere”; it should be load-bearing.
• Recommendation
  Guarantee one scripted “first contact.” Every session should have at least one authored scare beat that always fires — so a run is never a quiet, empty walk. Cheap to script, huge for consistency.
• Recommendation
  A directional pull to the exit, not a map. A faint compass or a sound cue toward the door fights disorientation without removing the dark. Players should feel lost-but-oriented, never lost-and-stuck.

Shape the run so tension rises, breaks, and climaxes.

• Recommendation
  Structure every run in three beats. Arrival (safe, learn the controls, the house feels wrong) → Escalation (first contact, hunt for keys, noise economy bites) → Exfil (door opens, the house turns hostile, a sprint for the exit). An intensity ramp beats a flat patrol.
• Recommendation
  Sell anticipation, not the reveal. The scariest second is before you see it. Stretch the approach: distant footsteps, then closer, then the flashlight catches a shape. Audio telegraphing is your most valuable cheap asset.
• Recommendation
  Force communication with asymmetric information. One player sees a symbol; another sees where it fits. Make players need each other’s eyes, not just shout “found a key.” This also lays the psychological groundwork for the skinwalker later.
• Recommendation
  Don’t make the exit an instant escape. Opening the door should start a countdown or a last stand — the team must regroup and run together. That single change creates a climax and forces the central revive dilemma at the worst possible moment.
• Recommendation
  Give escaped/dead players something to do. Let them whisper or place a single limited ping for the living. Spectator dead-time kills the mood and tempts rage-quits; a haunted voice from a dead teammate adds to it.

Make the four distinct through interdependence, not just stat sliders.

• Recommendation
  Define each character by what they do under fear. Since fear is the core system, anchor every kit to it: Billy can perform the terrifying interaction (the loud door) while panicked; Joe’s Sprint Burst is his panic button — escape a grab, but die if caught alone; Bob body-blocks and revives faster, the calm in the storm.
• Recommendation
  Build dependence, not dominance. Joe is the only one who can outrun a grab — but he’s fragile, so he needs the team to survive. Bob’s barricade should cost something real (consumes the door, seals a route). Strengths that create obligations make co-op talk.
• Recommendation
  ID teammates by light color in the fog. Give each character a distinctly tinted flashlight plus a clear silhouette. It’s nearly free, fixes “who is that in the dark,” and quietly pre-loads the skinwalker’s best tell (the copy’s light is subtly wrong).
• Phase 2
  Keep PhilBillyBob a goofy risk, not a strict upgrade. Make him a run modifier — faster but louder, draws more aggro — so the victory-lap character is fun chaos, not the obvious min-max pick. Preserves challenge and the joke.

Make it feel alive, navigable, and replayable on a modular budget.

• Recommendation
  “Alive” on a budget = a handful of scripted changes. A hallway you passed is now dark; a door that was open is bricked; the Safe Room moved. Even three or four in-session swaps sell “the house moves” without procedural generation.
• Recommendation
  Give every room one strong identity. One memorable prop, one color of light each. Players must be able to build a mental map even in fog — “the red bathroom,” “the room with the chair.” This is what stops the maze from becoming a frustration generator.
• Recommendation
  Make the Safe Room the emotional anchor. A hub players retreat to, where fear drops — so leaving it always costs something. The push-pull between the safe room and the dark is your core tension engine.
• Recommendation
  Pre-author 5–6 reusable “scare nodes.” A shape down a long sightline, a thing in the mirror, a silhouette that’s gone when you look again. The Director picks from them at random. Cheap, modular, memorable.
• Recommendation
  Replayability for almost free. Randomize key spawns among a set of valid spots and randomize which scare nodes fire. Same blockout, a different night every time.

Decouple fear from the zombie and make it change decisions, not just visuals.

• Reframe
  Fear needs multiple inputs. Time in darkness, distance from teammates, proximity to the zombie (even unseen), and Director events should all raise it. Now the whole house is scary and you have real gameplay levers — not a light switch tied to one monster.
• Recommendation
  High fear must feed back into the loop. Slower interactions, shorter sprint stamina, louder footsteps, an unreliable flashlight. Fear makes you louder → the zombie hears you → more fear. A death spiral the team has to actively break is the best kind of pressure.
• Recommendation
  Let players spend agency to lower it. Standing in light, hugging a teammate’s position, or reaching the Safe Room calms fear. Now it’s a resource to manage, not just a debuff to endure — and managing it together is teamwork.
• Recommendation
  Make the UI itself afraid. No clean meter — a heartbeat thump at the screen edges, ragged breathing, HUD elements that flicker or briefly lie at high fear. Diegetic dread beats a number every time.
• Cannot Confirm
  Vision distortion risks motion sickness in WebGL. Cap the distortion hard and ship an accessibility toggle from day one. Comfort can’t be confirmed until playtested on the real build.

The signature feature. It must be fair, learnable, and social — not a coin flip.

• Design
  Author a fixed vocabulary of tells. Repeats old chat verbatim; can’t answer something that just happened; the name flickers; one beat of wrong animation; the flashlight color is subtly off; casts no reflection. Players need a stable set of tells they can learn — random unfairness kills the mechanic.
• Design
  Turn paranoia into a mechanic: the verification ritual. Give players a cheap test — “flash your light twice,” or ask a fact only the real teammate witnessed. The copy fails it. This converts a guessing game into a skill, which is the difference between tense and infuriating.
• Design
  Its weapon is leading you, not hitting you. The mimic rarely attacks directly — it lures you away from the group, into the dark, toward the wrong door. A social predator is far scarier than another melee enemy.
• Fairness Rule
  Never let a wrong accusation instantly kill a real teammate. If accusing is lethal and friendly fire is on, you’ve built a griefing machine. Accusation must carry a cost and a consequence both ways — wrong calls hurt the accuser too.
• Design
  Spawn it inside legitimate confusion. The best beat: the copy appears right after a real separation, wearing the teammate you just lost sight of. Tie spawns to genuine sightline-loss windows, never to a random timer.
• Scope
  Ship the text-mimic version; treat voice replay as the dream, not the gate. Voice replay is the killer feature and the riskiest (consent, privacy, latency). Build the fully-shippable text/movement/appearance mimic first; the appearance descriptor being networked from day one is the right call.

What should not be in the first playable build.

• Blocker
  The two briefs disagree on the engine — Godot/GDScript vs Unity/C#. These are mutually exclusive and decide everything downstream. No code should be written until one is locked. This is the highest-priority open item in the whole document.
• Cut from MVP
  Voice chat. The briefs also conflict here — one puts WebRTC + positional audio + STUN/TURN inside the MVP. That alone is weeks of work and live ops. Strongly recommend a text-only MVP; voice is its own milestone.
• Cut from MVP
  Four distinct character kits. Ship the MVP with two clear opposites (Joe + Bob) — or even one readable body. Abilities should come only after movement and AI feel right; tuning four kits on top of unproven core loop is wasted effort.
• Defer
  Drop-in / drop-out with full-state snapshots. Safe-spawn rules, mid-session joins, and orientation grace periods are a lot. Consider join-at-start for the first playable and add true drop-in once the loop is fun.
• Correctly Phase 2
  Keep these out, as planned. Skinwalker voice replay, flashbacks, minigames, secret ending, weapons, crafting, accounts, matchmaking. The brief already fences these — hold the line.
• Cannot Confirm
  PS1 post-FX at 60 FPS in WebGL is unproven. Dithering, color quantization, screen distortion, and fog stacked together may not hold frame-rate. Prototype the render target before you depend on the look.

Only the questions that must be resolved before implementation begins.

1. Engine: Unity/C# or Godot/GDScript?The two briefs conflict. Nothing can start until this is locked.
2. Is voice chat inside the MVP gate, or a later milestone?Also conflicts between briefs. Changes the scope by weeks and adds live ops.
3. Target session length — 10, 20, 30 minutes?Drives map size, key spacing, and the fear-pacing ramp.
4. Is the true design target 2 players, or 3–4 with 2 as a fallback?Changes revive math, separation pressure, and the skinwalker’s viability.
5. Can players ever harm each other (friendly fire)?Decides whether skinwalker accusations can be lethal — and whether griefing is possible.
6. What does failure feel like — full wipe and restart, or a softer retry?Sets how punishing the loop is and how players treat risk.
7. What is the performance floor — minimum laptop/GPU at 60 FPS? Is mobile browser ever in scope?Caps the entire art and post-FX budget.
8. How far does the horror go — gore, jump scares, audience/age rating?Affects content, store eligibility, and whether an age gate is needed.
9. Is browser localStorage the only persistence forever, or accounts later?Shapes the privacy stance for future skinwalker voice data.
10. Who owns the art pipeline after MVP, and what’s the budget?The modular kit and characters need an owner before custom art begins.