agentleFS
EL4CTEOEL4CTEO/roblox-skills/skills/roblox-architecture/SKILL.md

roblox-architecture

Skill92 starsChanged 47 days agoMITView on GitHub

Roblox project structure and client/server model - replication, where code and assets belong (ServerScriptService, ReplicatedStorage, StarterPlayer), RunContext, service/controller bootstrapping, player/character lifecycle, CollectionService components, attributes, streaming-safe client code, library choices. Use when starting a project, deciding where code goes, or fixing replication/streaming bugs.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Every Roblox game is a server (authoritative, runs Scripts, sees everything, owns DataStores) plus one client per player (runs LocalScripts …

    line 10
  2. Distinctive:

    Always handle players/characters that already exist when your code starts.

    line 138
  3. Very specific:

    Use ModelStreamingMode = Atomic for models whose parts scripts need together.

    line 203
  4. Very specific:

    Treat GetChildren, raycasts, and property reads of streamed-out parts as partial/stale.

    line 205

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Roblox game architecture

The model in one paragraph

Every Roblox game is a server (authoritative, runs Scripts, sees everything, owns DataStores) plus one client per player (runs LocalScripts / client Scripts, renders, reads input). The server's DataModel replicates to clients automatically. Client changes do not replicate back, except: physics of parts the client has network ownership of (its character by default), its own Humanoid state/animations, and remote calls. Assume every client is compromised (see roblox-security). Design every feature as: client requests → server validates and decides → state replicates → clients render.

Where things go

ContainerRuns/visible onPut here
ServerScriptServiceServer onlyServer Scripts and server-only ModuleScripts (game logic, data, anti-cheat).
ServerStorageServer onlyServer-only assets: maps, NPC templates, tools to clone, secrets-adjacent config.
ReplicatedStorageBothShared modules (types, config, pure utils), RemoteEvents, assets clients clone.
ReplicatedFirstClient, firstLoading screen only. Replicates before everything else.
StarterPlayer.StarterPlayerScriptsClient (copied once to PlayerScripts)Client entry point and controllers.
StarterPlayer.StarterCharacterScriptsClient (copied per spawn into character)Per-character client scripts (rarely needed).
StarterGuiClient (copied to PlayerGui)ScreenGuis. Set ResetOnSpawn = false unless you want reset on death.
WorkspaceBoth (streamed to clients)The 3D world only. Not a place for code or templates.
Lighting, SoundService, Teams, TextChatServiceBothTheir respective config objects.

Rules:

  • Code in ReplicatedStorage is downloadable by exploiters. Never put server logic, secrets, admin lists, or anti-cheat thresholds there.
  • Script.RunContext: Legacy (default: behavior depends on the container), Server, Client, or Plugin. A non-Legacy RunContext makes a Script run regardless of container (e.g. a Client Script in ReplicatedStorage runs on every client) — so templates you clone must not contain such scripts unintentionally. Legacy Scripts don't run in ReplicatedStorage/ServerStorage.

Project layout (Rojo-style, recommended)

src/
  server/            -> ServerScriptService.Server   (init.server.luau = entry point)
    Services/        -> one ModuleScript per system (DataService, CombatService, ...)
  client/            -> StarterPlayer.StarterPlayerScripts.Client (init.client.luau)
    Controllers/     -> one ModuleScript per client system (InputController, UIController, ...)
  shared/            -> ReplicatedStorage.Shared (types, config, pure logic, network definitions)
Packages/            -> ReplicatedStorage.Packages (Wally/pesde dependencies)
ServerPackages/      -> ServerScriptService.ServerPackages

One entry script per side, everything else is a ModuleScript. This gives deterministic load order, testable modules, and a single place to wire dependencies. Roblox's own recommendation is equivalent: one Script (RunContext = Server) in ServerScriptService and one Script (RunContext = Client) in ReplicatedStorage, each requiring modules and calling their start(). Tooling setup: roblox-tooling.

Bootstrapping services (no framework needed)

--!strict
-- ServerScriptService/Server/init.server.luau
type Service = {
	init: ((self: Service) -> ())?, -- sync setup; may reference other services; must not yield
	start: ((self: Service) -> ())?, -- runs after every init; may yield / connect events
}

local servicesFolder = script:WaitForChild("Services")
local services: { [string]: Service } = {}

for _, module in servicesFolder:GetChildren() do
	if module:IsA("ModuleScript") then
		services[module.Name] = (require :: any)(module)
	end
end

for name, service in services do
	if service.init then
		debug.setmemorycategory(name)
		service:init()
		debug.resetmemorycategory()
	end
end

for name, service in services do
	if service.start then
		task.spawn(function()
			debug.setmemorycategory(name)
			service:start()
		end)
	end
end
  • Services reference each other with plain require (no cycles) or receive dependencies in init.
  • The client mirrors this with Controllers. Keep names symmetrical (ShopService / ShopController).
  • Frameworks like Knit are unnecessary for this; if a project already uses one, follow its conventions.

Player and character lifecycle

--!strict
local Players = game:GetService("Players")

local function onCharacterAdded(player: Player, character: Model)
	local humanoid = character:WaitForChild("Humanoid") :: Humanoid
	humanoid.Died:Once(function()
		print(`{player.Name} died`)
	end)
end

local function onPlayerAdded(player: Player)
	-- Load data first (see roblox-data-stores), then set up character handling.
	if player.Character then
		task.spawn(onCharacterAdded, player, player.Character)
	end
	player.CharacterAdded:Connect(function(character)
		onCharacterAdded(player, character)
	end)
end

local function onPlayerRemoving(player: Player)
	-- Save data, release session locks, clear every table keyed by this player.
	print(`{player.Name} left`)
end

Players.PlayerAdded:Connect(onPlayerAdded)
Players.PlayerRemoving:Connect(onPlayerRemoving)
for _, player in Players:GetPlayers() do -- players who joined before this script connected
	task.spawn(onPlayerAdded, player)
end

game:BindToClose(function()
	-- Server shutdown: flush saves for remaining players (max ~30 s total).
end)
  • Always handle players/characters that already exist when your code starts.
  • Characters respawn: anything bound to a character must be re-created on CharacterAdded.
  • On the client, character descendants may not exist yet (streaming): WaitForChild them.

Components with CollectionService tags

Tag instances (Studio Tags section or CollectionService:AddTag) and attach behavior in code. This scales better than copying scripts into every model and works with streaming because the added/removed signals fire as instances stream in and out.

--!strict
local CollectionService = game:GetService("CollectionService")

local TAG = "KillBrick"
local cleanups: { [Instance]: () -> () } = {}

local function onAdded(instance: Instance)
	if not instance:IsA("BasePart") then
		return
	end
	local connection = instance.Touched:Connect(function(hit)
		local humanoid = hit.Parent and hit.Parent:FindFirstChildOfClass("Humanoid")
		if humanoid then
			humanoid.Health = 0
		end
	end)
	cleanups[instance] = function()
		connection:Disconnect()
	end
end

local function onRemoved(instance: Instance)
	local cleanup = cleanups[instance]
	if cleanup then
		cleanup()
		cleanups[instance] = nil
	end
end

CollectionService:GetInstanceAddedSignal(TAG):Connect(onAdded)
CollectionService:GetInstanceRemovedSignal(TAG):Connect(onRemoved)
for _, instance in CollectionService:GetTagged(TAG) do
	task.spawn(onAdded, instance)
end

Configure components with attributes (instance:GetAttribute("Damage"), GetAttributeChangedSignal) instead of IntValue/StringValue children. Attributes replicate server → client, are cheaper, and show in the Properties panel.

Sharing state

NeedUse
Small per-instance/per-player values visible to clientsAttributes on the instance / Player (server sets).
Events and requestsRemoteEvent / UnreliableRemoteEvent (see roblox-networking).
Large or per-player private state (inventory, quests)Server-owned tables + remotes that send snapshots/diffs to that player only.
Same-side communicationModule function calls or a typed Signal module — not BindableEvent, _G, or shared.
Persistent dataroblox-data-stores. Cross-server: MessagingService / MemoryStore.

Streaming (on by default for new places)

With Workspace.StreamingEnabled, clients only have nearby parts. Client code must:

  • WaitForChild (with a timeout when it may never arrive) or nil-check anything in Workspace.
  • Use ModelStreamingMode = Atomic for models whose parts scripts need together; Persistent only when a model must always be present (costs memory).
  • Treat GetChildren, raycasts, and property reads of streamed-out parts as partial/stale; do full-world queries on the server.
  • Pre-stream destinations on the server with player:RequestStreamAroundAsync(position) before teleporting a character.
  • Scopes, racing, free cameras: player.FrustumStreaming = Enum.FrustumStreamingMode.Automatic streams what the camera sees instead of raising the radius.

Full pattern catalog and recommended Workspace settings: references/streaming.md.

Anti-patterns

  • Dozens of loose Scripts inside parts/models (use tags + one service).
  • Polling loops (while task.wait(0.1) do check() end) instead of events or Heartbeat with deltaTime.
  • Trusting the client for game state, or firing remotes every frame.
  • _G/shared globals, circular requires, yielding at module top level.
  • Cloning maps into Workspace from the client (client-only copies never get server updates).
  • Storing per-player data in the character (it's destroyed on death).

Library ecosystem (signals, cleanup, networking, UI, ECS, data): references/ecosystem.md.

Related skills

roblox-networking, roblox-security, roblox-data-stores, roblox-tooling, roblox-performance, roblox-game-systems.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.