# agentSkill A skill packages what an agent needs to do one job well: the instructions, the assets and toolboxes the job runs through, and — for jobs with distinct branches — subskills that go deeper. Skills advertise themselves. Every registered skill's name and one-line description rides the agent's tool responses, so an agent discovers what this world knows how to do without being told to look. Invoking one returns the instructions plus the exact identities and tool names the job needs. ## Shape ``` <name>.agentSkill/ skill.yaml # the manifest: description, when, checks, dependencies, order instructions.md # the prose an agent receives on invoke README.md # what this skill is, for a human reading the tree .metadata # tags <sub>.agentSkill/ # optional subskills, reachable as <name>/<sub> ``` ## The manifest ```yaml $schema: zero/agent-skill/v1 description: Making a character a person can actually play. when: > Reach for this when the player should look like something else, or when you are about to build a camera and a body and wire them together by hand. checks: - check: Holding a movement key in play moves the body across the ground. observe: Read its position before and after holding the key; the numbers differ. - check: The body animates while it moves. observe: A time-sampled collage from a camera outside the player, while the key is held. dependencies: - "@builtin::tools.scene" - "@builtin::tools.scene.player" - "@builtin::components.PlayerPrototype" - "@builtin::docs.core.multiplayer" order: [spawn-points, camera-rig] ``` `description` is the line the roster carries, and the whole basis on which a skill is chosen. Name the job the skill makes a reader competent at, in the outcome they are after — "getting a scene's player and view right", "making content the engine will load". A reader scanning the roster is matching it against the job in front of them, so the closer it reads to their own sentence about that job, the sooner they open it. `when` is the longer trigger — the situations that should send a reader here, in the words they would use for their own task. `checks` is what the reader is answerable for at the end, and it leads the rendered skill, ahead of the procedure. Each entry pairs a **condition that must hold** with **where it is seen**, and both halves are required — a condition alone is satisfied by any glance, and the wrong vantage is how a job passes while every one of its faults sits outside the frame. Write them as observations someone else could repeat and disagree with: a number that changed, a thing visible in a named shot, a count that came back as one. A skill publishes only once it carries at least one. `stages` models a skill whose passes run in order, when the order is load-bearing. Each stage is `{ name, detail?, checks? }`, and a check that sits in a stage is answered after that pass has run — which is the difference between a criterion that means something and one that misleads. A skill with no meaningful order leaves `stages` out and lists its checks flat; a skill that has one may do both, and the flat checks then read as the final gate. `verdicts` names the fixed set of outcomes a reader reports — at the end of each stage when the skill has stages, and once at the end when it does not. Each is `{ verdict, means }`: the word, and what reporting it asserts. Both halves are required for the reason both halves of a check are. A verdict is a check one level up, deciding a whole pass rather than one condition, and a bare `refine` is a label each reader fills in privately — so two runs report the same word about different states and nothing downstream can tell them apart. Saying what it means settles the boundary in the skill instead of in whoever happens to be reading it. A check marked `perRun: true` is one whose content is written for the job at hand: the manifest can say *keep a list of every feature this object must have and answer against it*, but not what is on that list, because that is per run. It renders with that instruction attached. ```yaml verdicts: - verdict: continue means: Every check in the stage held, and the next stage can build on it. - verdict: refine means: A check failed on something this stage can still fix. Run it again. - verdict: stop means: A check failed on something an earlier stage decided. Go back to it. stages: - name: blockout detail: Mass and proportion only. checks: - check: Every part on the identity list has a stand-in at roughly the right size. observe: A capture beside the reference, at the same angle. perRun: true - name: material checks: - check: Metal reads as metal and cloth as cloth. observe: The metallic and roughness passes, after the lighting stage has run. ``` `dependencies` is one flat list of asset **identities**: a toolbox, a single tool, a guide, a component, a module, another skill. Each entry is resolved when the skill is opened and reported with what it turned out to be, so a skill never declares "this one is a toolbox" — the asset says so, and the reader groups by that. Identities are what make a skill portable. They pin like any other reference, so installing a skill into another world brings the toolboxes and guides it runs through along with it. A bare name (`scene`) or a guide path (`core/multiplayer`) resolves to nothing and is refused before publish. A toolbox entry lists the tools it carries, and a tool entry carries its own signature, so the instructions never restate a call contract that can drift. An identity this world lacks is reported as missing rather than dropped. Write instructions that name the toolboxes and tools a job runs through, and leave the call form to the reader — an agent may be holding the MCP tool surface, the shell, or Luau, and the signatures are listed for it either way. ## Subskills A `<sub>.agentSkill/` folder nested inside a skill is a subskill. It has the same shape as its parent and is invoked as `parent/sub`. Subskills are reachable only through their parent: they are absent from the roster and from `skills.list`, and appear in the parent's invoke result. A skill can therefore carry a dozen narrow procedures while costing the roster one line. `order` in the manifest names the subskills that lead. Any subskill it omits still appears, sorted by name, after the ordered ones — so adding a subskill folder is the only step needed to publish it. ## Publishing one Make a skill and it is live: it joins the roster agents read and it opens by name from the moment it exists. Edit its description and that is the line agents see next. A skill lives anywhere under `/zero/source/`. `asset.create` hands you a valid one to fill in: ```luau asset.create("agentSkill", "waveSpawning", { folder = "skills" }) ``` A skill's scope — `builtin`, `library`, or `world` — comes from where it lives rather than from anything it declares, so a world's own skill always reads as the world's. ## Opening one ``` agent_skill {} -- the roster agent_skill { name = "scenes" } -- one skill's instructions + deps agent_skill { name = "scenes/player-setup" } -- a subskill ``` The same three through the toolbox — `skills.list`, `skills.invoke` — and from Luau via `tools.use("skills", "invoke", "scenes")`. `asset.inspect` on a skill ref returns the structured record all of them render. Skills authored in a world are the world's own, and are labelled as world content wherever they surface beside built-in ones. ## In the Inspector Skill: what the skill is for and when it applies, how many checks and verdicts it declares, its stages, a link to each asset it depends on (and a problem for each that does not resolve), its subskills, and **Open** for instructions.md.

# profiler Find where the frame time goes — autonomously, in any scenario, without a human reading a graph. The engine's raw profiling surface is flat (`getProfilingDataTable`, `getLuauVmDataTable`, `profiler.stopCapture`). These tools compose it into the honest, self-accounting hierarchy a profiler should give you: every node shows its **self** time (own cost, excluding children), its **total**, call count, and share of the frame; the frame time no schedule covers shows as an explicit **present / idle** (vsync) bucket instead of vanishing; and each heavy component's `update(dt)` is attributed down to the exact **component + entity** so the cost lands on something you can open and fix. The intended call site is the `use_tool` MCP surface — `use_tool { toolbox = "profiler", tool = "...", args = [...] }`. ## The tools - **`summary`** — the engine's own one-line performance summary, the string that rides a tool response. A *different measurement* from everything below: this is the published smoothed frame, while the rest report the profiler's effective frame time (wall clock minus the cost of the call doing the asking). Read it when you want the figure other callers are being told; read `frame` when you want to know where the time goes. - **`frame`** — the full attributed tree (self / total / calls / % per node). The deep view; read from the top row down. - **`hotspots`** — a flat, ranked list of the costliest nodes by self time. The fast "what's slow" triage. - **`scripts`** — per-component `update(dt)` attribution: last / avg / max / calls, by `component @ entity`. The "which of *my* content is slow" view. - **`record`** — start / stop / status a background recording that spans a play session. Play across as many turns as you like, then stop to get avg / p50 / p90 / p99 / max frametime and the **spike-cluster distribution**. The path for "it hitches while I play", which a single live frame can't catch. - **`ring`** + **`retro`** — the latency-immune path. `ring on` keeps an always-recording history of the last N seconds; `retro` reads it back AFTER the fact. Because the data is historical, it does not matter that a tool call lands seconds after the spike — the flaw that makes `frame`/live sampling unreliable for an agent (which can't control when its calls land). - **`watch`** + **`hits`** — the conditional path. `watch <ceilingMs>` arms a frame-time watchdog that either records every bad frame (review with `hits`) or pauses gameplay the instant one hits (freeze + inspect with `retro`). Runs in the engine's per-frame hook, so it catches stalls an agent poll would miss. - **`gpu`** — GPU pass timings from timestamp queries, label-aggregated (`compute.<shader>`, `scene.*`, `post.<effect>`, `feature.*`). The other side of the frame: when the CPU tree shows everything under **present / idle**, the frame is GPU-bound and this is the view that says which passes the GPU spends it on. Everything an agent reads is on **effective** frame time — wall-clock minus the agent's own injected `execute` cost — so your profiling and authoring calls never masquerade as game hitches. Spike reports are **distributions** (every spike frame grouped by its dominant hotspot), so one capture shows the whole picture rather than a single anecdotal worst frame. The ring and watchdog are **editor-profile only** and off until enabled, so a shipped runtime never pays for them. ## The autonomous loop 1. `hotspots` or `frame` — baseline, what's expensive now (best when the scene is steady; live sampling races the ~seconds of tool-call latency). 2. For anything that hitches: `ring on` → play / test → `retro` — latency-immune, and the clustered distribution of every spike. Or `watch <ceilingMs>` in `pause` mode to freeze the exact bad frame. 3. `scripts` — attribute script cost to the component + entity. 4. Fix, then `retro` / `record` again and `compare`. Every read tool accepts a recording label as its `source` — including a `retro` or `hits` snapshot — so anything you can see live you can also drill into for a recorded window's clusters, worst, or typical frame.

# renderDiagnostics module Publishes the `overdraw` and `zfighting` capture views, so `capture pass=<name>` resolves to them. Requiring the module registers both. ```luau local diagnostics = require("@builtin::systems.renderDiagnostics.renderDiagnostics") diagnostics.register() -- idempotent; already done by the require diagnostics.unregister() -- withdraw both views and drop their features ``` Registration is cheap and permanent: each view's passes are gated to the debug channel the registration hands out, so they are skipped for every camera that is not a capture selecting that view. The feature that draws a view is created the first time it is selected, through the registration's `ensure`. `renderer.captureView.list()` enumerates every registered view, and the package readme has what each colour band means.

# inputSim Every simulated input, across every device a player might hold. Drive the world as a keyboard-and-mouse player, as a phone, or as someone with a controller — and read back what actually reached the game. ## Why all three, always A world checked only from a keyboard is a world whose keyboard experience is the only one anybody verified. That is how a game ships playable on a desktop and dead on a phone: not because the touch path was written badly, but because nothing ever drove it. So this toolbox covers the three classes symmetrically, and `device` makes a desktop session present as any of them. ## Driving the game: actions A game reads its input through the controls its input maps declare — `move`, `jump`, `sprint`. `action` drives one by name with the value the game should read, which is how a test or an agent plays a game: the same call works whatever device the player would hold, and nothing about a device — which key a map binds, where the pointer is — has to be worked out. ```lua tools.use("inputSim", "bindings") -- the live controls and their kinds tools.use("inputSim", "action", "jump") -- tap a button tools.use("inputSim", "action", "move", { 0, 1 }, 2) -- walk forward for 2 s tools.use("inputSim", "action", "sprint", true, nil, true) -- hold until `release` tools.use("inputSim", "fired") -- reported under class "driven" ``` A button takes `true`, a 1-D axis a number, a 2-D axis `{ x, y }`. The value joins the control like another device: its shaping applies, its press and release fire, and a context or a suppression that stands the control down stands it down too. `release` lets go of every driven control. The device verbs below prove the devices themselves: that a key, a pad or the touch overlay reaches the control a player would press it for. ## The loop ```lua tools.use("inputSim", "device", "touch") -- present as a phone tools.use("inputSim", "bindings") -- what controls are live? tools.use("inputSim", "layout") -- what is on screen? tools.use("inputSim", "tapButton", "Jump") -- tap it where a player would tools.use("inputSim", "fired") -- did it reach the game? ``` `bindings` first, always. Nothing is live until a component activates an input map, so an empty list means no awake component asked for input — which is a different problem from a control that fires and does nothing. `fired` is what turns a simulated input into an assertion. A capture can look identical whether or not the input landed; this cannot. Its `count` is the number of engine frames the control was active, so a tap that holds a button across two frames reads 2 for one press. To count presses, subscribe with `control:onPressed`. ## One argument shape Every verb here takes its arguments POSITIONALLY, in one order across the whole toolbox: **what** first — the button, key, stick, trigger, label or contact id — then **where** as flat `x, y` (or `dx, dy`), then **how long** in seconds or **how far spread** in frames, then **which pad slot**. Every optional argument trails. No verb wraps its arguments in an options table, so a verb's shape is known from having called its neighbour. ```lua tools.use("inputSim", "click", 0, 400, 300) -- what, where tools.use("inputSim", "mouseDown", 0) -- the same what tools.use("inputSim", "touch", 400, 300, 0.5) -- where, how long tools.use("inputSim", "padStick", "left", 0, -1) -- what, where ``` Every surface that names arguments resolves those names against the same positional list, so one call is written four ways and reaches the same slots: `click(0, 400, 300)` in Luau, `{ button = 0, x = 400, y = 300 }` through `tools.use`, `args` over `use_tool`, and `--button 0 --x 400 --y 300` in the shell. A verb whose every argument is optional takes `{}` for "no arguments at all", and a name the verb does not declare is refused with the ones it does. ## Who the input is for What this toolbox simulates is the game's input: a key, click, motion, wheel or touch reaches the scene alone. It reaches the scene whatever a person at the window has focused — a panel they clicked, a terminal they are typing a message into — and it reaches no UI widget, so a run driving the game never types into the person's panel and the person's typing never stands the run down. Every key, pointer and touch verb takes a trailing `target`: `"game"`, the default, or `"ui"`, which reaches the UI the way a person's device does. Driving a UI by its elements — clicking a button, typing into a field — is the `gui` toolbox's work. ```lua tools.use("inputSim", "key", "KeyW", 1) -- walks the player tools.use("inputSim", "key", "KeyA", nil, "ui") -- types into the focused field ``` A mouse button is written either way, index or name: `0` / `"left"`, `1` / `"right"`, `2` / `"middle"` — the same pair `Zin.state.mouseButtonDown` reads. A pad button is always a canonical position name (`"south"`, `"dpad_up"`, …), because a pad has no index a player could recognise. ## Where a point is measured from `mouseMove` and `click` take their point in physical pixels, measured in the space the input's receiver reads positions in, so `target` decides it. With `target` `"game"` (the default) the point is in the space `capture` reports: the scene's own pixels, measured from the top-left of the picture a capture returns. That is the space `Zin.state.get().mouse_position` reads a pointer back in and the space `worldToScreen` answers in, so a point read off any of the three is a point the other two take. ```lua tools.use("inputSim", "mouseMove", 300, 200) Zin.state.get().mouse_position -- 300, 200 ``` The window can hold the scene inside a rect with editor chrome above and to the left of it; the offset is those two verbs' to carry, not the caller's. With `target` `"ui"` the point is on the window: the space every UI widget is recorded in, wherever its screen draws (the window itself, or the picture a Scene panel presents). A widget's `gui.bounds` rect is in logical points, and times `ui.pixelRatio()` it is the point to aim at: ```lua local b = tools.use("gui", "bounds", "play-button") local r = ui.pixelRatio() tools.use("inputSim", "click", 0, b.centerX * r, b.centerY * r, nil, "ui") ``` `gui.clickElement` does the same by id and refuses a widget the last frame did not draw where its rect claims, so it is the call for pressing a named button. A `macro`'s ops are those same verbs, so an absolute point written for one is the point the other takes. `touch`, `touchDown` and `touchMove` place their contact on the window. `camera.viewData().viewport` states where the scene's rect sits on that window. ## Keyboard and mouse | Tool | What it does | |---|---| | `key` | Tap a key — down, optional hold, up | | `keyDown` / `keyUp` | Hold and release | | `click` | A full click, optionally at a point | | `mouseDown` / `mouseUp` | Hold and release a button | | `mouseMove` | Move the cursor to a point; the delta is what look bindings read | | `mouseMoveBy` | Move the pointer by a motion, over `steps` frames — repeat it to hold a look axis | | `scroll` | Turn the wheel | | `lockPointer` | Lock or release the pointer — mouse-look is usually gated on it | `keyDown` and `mouseDown` hold until something releases them, and a hold outlives the call that made it — including a driver task that was cancelled mid-press. `held` says what is down; `release` puts all of it down at once. ## Touch | Tool | What it does | |---|---| | `touch` | A whole tap at a point | | `touchDown` / `touchMove` / `touchUp` | One finger, by contact id — several at once | | `stick` | Push the on-screen movement stick | | `drag` | Drag across the look zone, spread over frames | | `pinch` | Pinch or spread two fingers | | `tapButton` | Tap an on-screen button **by the label a player reads** | `tapButton` aims at the button's real centre and raises when no live button carries that label, listing the ones that do. That is deliberate: a tap that silently lands on nothing looks exactly like a control that works. ## Gamepad | Tool | What it does | |---|---| | `pad` / `padOff` | Connect or disconnect a simulated pad | | `padPress` | Press a button — down, optional hold, up | | `padDown` / `padUp` | Hold and release | | `padStick` | Push a thumbstick | | `padTrigger` | Squeeze a trigger to an analog travel | Buttons are named by POSITION, never by the letter printed on them: `south`, `east`, `west`, `north`, `left_shoulder`, `right_shoulder`, `left_trigger`, `right_trigger`, `left_stick`, `right_stick`, `select`, `start`, `guide`, `dpad_up`, `dpad_down`, `dpad_left`, `dpad_right`. The same call works whatever pad is connected. The name passed to `pad` decides which legends a prompt draws, so connecting `"Sony DualSense Wireless Controller"` is how a scheme's PlayStation prompts get checked. ## Reading back | Tool | What it answers | |---|---| | `device` | Which class the session presents as — and forces one | | `bindings` | Which controls are live, and which map contributed each | | `layout` | What the on-screen controls are actually drawing | | `fired` | Which controls were active since last asked, for how many engine frames, and from which device | | `held` | What the session is holding down right now, across every device | `layout` is how a phantom control is caught: a label on screen that no entry in `bindings` explains is a button nothing consumes. `held` is how a leftover hold is caught: keys, mouse buttons, fingers, the stick and connected pads in one answer, with `atRest` true when the session holds nothing. A control that keeps firing after a take ended reads here as the key that is still down, so recovering from it is a `release` rather than a guess at what was pressed. ## Replaying a sequence A run of events worth repeating becomes a `.inputMacro` asset, so the same sequence drives the same world on every later check. | Tool | What it does | |---|---| | `recordMacro` | Capture real input for `duration` seconds | | `saveMacro` | Write an event list to a `.inputMacro` asset | | `macro` | Play an event list back through the input surface | | `listMacros` | Every `.inputMacro` asset in the project | An event is `{ t, op, arg }`. `op` names one of the verbs above that drives input and `arg` carries that verb's arguments — the value itself when the verb takes one, a table keyed by the verb's parameter names when it takes several — so every device class a verb reaches is a device class a macro reaches. A `touchDown` event's `arg.id` is the name its contact goes under, and the `touchMove` and `touchUp` events carrying that `id` drive it. ```lua tools.use("inputSim", "macro", { { t = 0.0, op = "pad" }, { t = 0.4, op = "padDown", arg = "south" }, { t = 0.6, op = "padUp", arg = "south" }, { t = 0.8, op = "touchDown", arg = { x = 714, y = 984, id = 1 } }, { t = 1.2, op = "touchUp", arg = 1 }, }) ``` `recordMacro` reads what every device is HOLDING each frame — keys, mouse buttons, pad buttons, the pad itself, fingers on the glass — and what each one is MEASURING every `sampleInterval` seconds — mouse motion, the wheel, thumbsticks, triggers, a sliding contact. `release` is the reset between takes — it lifts every contact, drops the on-screen stick, releases every held key and mouse button, and disconnects every connected pad, so one take never leaks a held control into the next. It reports how many of each it let go, and `held` reads the same set before or after.

# diagnostics Ask the engine what's wrong with a scene — triage broken materials, missing assets, and component errors before guessing. Three read-only inspectors over the active world's canonical state: the current component and entity errors (all of them, or narrowed to one entity), an at-a-glance summary of the scene (entity count, component-type breakdown, error count, cameras, culling stats), and a validation pass that checks scene integrity — broken materials, missing model assets, entity errors. Nothing here mutates the scene; it only reports.

# capture Render a screenshot of the scene from anywhere — an explicit position, a named camera station around an entity, a specific camera, or a mirror of the active viewport — and a grid of them in one image; or read back the frame the engine actually presented, exactly as the player saw it. The picture of the machine's own camera (a webcam, a phone camera) is `cameraInput`, covered by the `topics/camera-input` guide; this toolbox photographs the scene. Every source but `presented` and `window` (see [The presented frame](#the-presented-frame)) spawns an ephemeral offscreen camera, triggers one render, and returns the picture as an image value — returned from a tool, a probe or `execute`, it reaches you as a picture, with what the capture has to say about it (its size, the frame it is of, the cells of a grid) beside it. A PNG is written as well when you pass a `path`. The convenience wrappers render from a world-space position (with optional look-at), frame a target entity, mirror the active viewport camera including its layers, or render a specific camera by name from its exact authored pose and settings (`fromCamera`, with optional per-capture overrides that never touch the camera); they run as a background task that waits for the GPU readback and cleans up — which only works in play mode, since the task scheduler ticks there. The low-level oneshot spawns the camera, triggers one render, and returns the handles synchronously — the only path that works from edit mode, and the one the `capture` MCP tool is built on. A cleanup tool clears the capture output directory. ## Where a shot lands A capture given a `path` writes the PNG there and says where those bytes stand, beside the path it wrote: `durable` is true when the PNG is on the path it names and the next session reads it back, and false when the play shadow took it (live in this session and discarded on a guarded play-exit unless kept, with `survivesRestart` saying whether a restart keeps it and `playShadow` naming the routes that keep it) or when the path is in `/source/tmp/` scratch (gone when the engine stops or restarts, with a `warning` opening `SCRATCH`). Every capture carries it on the picture it returns as `durability`. Shooting gameplay means shooting in play, which is exactly when a `/source` destination is the shadow's, so this is the field that says whether a shot survives leaving play — `vfs.promotePlayShadow(path)` keeps it with play still running. ## The presented frame `source = "presented"` reads back the frame the engine presented to the player instead of rendering one. Nothing is drawn for it: it is the frame the viewport drew, copied out as it was finished, after the post chain, the upscale, the display transform, the content UI and every post-scene pass, at the size it was composed at. So it is the one capture whose temporal antialiasing, motion blur, camera motion, auto exposure and render scale are the real ones, and the source to verify any of them with. `source = "window"` is the whole presented surface (the window, the browser canvas or the headless framebuffer) with everything painted on the window: the editor's chrome and every window-level UI pass. Where no Scene panel shows the viewport, the two are the same image: the scene fills the window and the editor's chrome is painted over it, and each such frame says so in its `note` (a cell of a `presented` capture, a frame `renderer.presentedFrames` returns) for as long as that chrome is on the window. Show the Scene panel (the editor toolbox's `layout` tool) to read the game view alone. `frames = n` (1-32) takes n presented frames, one every `interval` frames (default 1: consecutive), at real gameplay speed, and lays them out as a grid in the order they were presented. Each cell states what its frame was drawn with: `frame` (the renderer's count, so consecutive frames carry consecutive numbers), `rasterWidth` / `rasterHeight` (the size the scene rasterized at), `compositeWidth` / `compositeHeight`, `renderScale` and `scaleMode`, `jitter` (the projection's sub-pixel offset in raster pixels), `exposure` (what the display transform multiplied the scene by) and `historyReset` (the viewport camera's first frame, a cut, or a hand-over to another camera). A single frame comes back at the size it was presented at; `width` / `height` resample it, or size the sheet a grid is laid into. A presented frame is drawn through the camera, layers, lens and chain that drew the screen, so the options that aim, light or shape a rendered frame (`camera`, `entity`, `position`, `viewpoint`, `fov`, `pass`, `renderLayers`, `postProcessing`, `environment`, `mode = "collage"` and the collage axes) are refused for it; aim the camera holding the viewport instead, or take an offscreen capture. A source that renders refuses `frames` and `interval` the same way, since they count presented frames. `deterministic` and `clearAir` hold the clock and the air for the whole engine, so they reach the presented frame too. A diagnostic buffer of the viewport itself, under its real camera motion and history, is read the same way: put the camera holding the viewport on that pass (the camera toolbox's `set` with `debugPass = "motion_vectors"`) and take `source = "presented"`. The viewport's velocity buffer is the frame's travel against the frame before it, so a turning camera reads in it, where an offscreen `pass = "motion_vectors"` holds the objects' own motion alone. The `motion_vectors` picture is a visualization: each channel is the square root of 100 times the stored displacement's size on its axis (R horizontal, G vertical, B the length), the sign is dropped, and R reads 255 from `width / 200` px of horizontal travel, G from `height / 200` px of vertical travel (raster pixels). The rendering guide's "What the `motion_vectors` picture encodes" states it in full. The same frames, with their pixels as `TextureCpuHandle`s, come from `renderer.presentedFrames({ scope = "viewport" | "window", frames, interval })`, and `velocity = true` adds the viewport's velocity buffer from each frame as numbers (`frame.velocity`, a `VelocityField`): each pixel's travel in pixels per frame, x right and y down, signed and unclamped, read with `field:at(x, y)`, `field:row(y)` and `field:region(rect)`. ## The feed Every capture publishes the frame it made to `@builtin::tools.capture.feed`, so a viewer can show the newest one without taking a shot of its own. `Feed.latest()` is that frame: a `src` an image widget draws (the render target's guid while the engine still holds it, the written path when the capture produced a file), with the size, the source and the pass beside it. `Feed.watch(fn)` calls `fn` as each new frame arrives. The editor's Capture panel is built on it, and the `watch` layout puts that panel under the scene. A viewer calls `Feed.shown()` to say it is drawing frames from the feed. From then on the feed holds the newest capture's render target so the picture stays up after the capture that made it is finished with it, and lets it go when a later frame replaces it; `Feed.shown(false)` stops. With nobody drawing, a capture's target is released the moment its capture is done, and nothing is held. ## Viewpoints A **viewpoint** names where the camera stands and which way is up in the image: `front`, `back`, `left`, `right`, `top`, `bottom`, `iso`. It replaces reverse-engineering a `{yaw, pitch}` pair, and names the seven stations that answer "is this built correctly". `basis` says which axes the name is measured against, and it is **required** alongside a viewpoint when a subject is being framed: - **`"local"`** — the subject's own axes. `front` is the side it faces however it is turned, and the frame is fitted to its own extents rather than to the world-axis box around them. - **`"world"`** — the world axes. `front` is whatever faces world +Z. For a crate turned 40 degrees those are different pictures, so the call says which one it wants. Framing a world position rather than an entity, both values mean the world axes and either is accepted. ## Passes `pass` names **what** the frame holds, by name. `"final"` is the lit picture; a built-in diagnostic pass (`albedo`, `normal`, `depth`, `motion_vectors`, …) replaces the shaded image with the buffer behind it; and any capture **view** a render feature has published (`renderer.captureView.list()` enumerates them) is selectable by its own name. A name that is neither is refused against the list of what is registered. The name carries two decisions: which channel the camera renders, and which render layers that channel is legible under. A diagnostic buffer is read flat, so the sky behind it and the post-process chain over it come off; a view that draws its own geometry declares the layers it needs and is rendered under them. Stating `renderLayers` answers that second question yourself, for the one call. `passes` spreads the same vocabulary across the cells of a collage, and each cell reports the layers it was read under. ## Projection `projection = "orthographic"` covers a fixed world-space height at every distance. Parallel edges stay parallel and two equal-size objects at different depths cover equal pixels, which is what makes it the projection to check a shape or compare sizes under — a perspective frame's convergence hides both. `orthoHeight` sets that height; omit it and it is fitted to what is being framed. Any camera can be orthographic, not only a capture: it is a field on the `Camera` component. A capture through a camera the scene holds (`fromCamera`, `viewport`, and `source = "camera"`, `"screen"` or the scene camera) takes that camera's `projection` and `orthoHeight` with the rest of its lens, and its lighting `environment` with an `isolated` camera's background and ambient; a `projection` or `orthoHeight` stated for the call stands over the camera's. ## Which entity a capture framed A capture names an entity, or a camera to render from, by its id or by its name. A name exactly one entity carries means that entity. A name several entities carry means none of them on its own, so the capture is refused, and the refusal says how many entities carry it and names each by name and id; pass the id of the one you mean. A player is drawn as its body, so a capture of a player identity (an entity carrying `UserIdentity`, named `user`) frames the body its `avatar` names. A player prototype is absent from the live scene while it plays, so its name reaches no entity there: a capture of that name is refused, and the refusal names each player spawned from the prototype, by its identity and its body, with their ids. Every cell of a capture aimed at an entity reports what it framed under `subjects`: one `{ id, name }` per subject, carrying `playerIdentity = { id, name }` when the picture is the body of the identity the capture named. ## Isolate Framing a subject fits the **frame** to it; it does not decide what is drawn in that frame. Everything else standing there is still rendered — and a scene's default player spawn sits at the world **origin**, which is exactly where a prop usually gets built, so a tight shot of a new prop at the origin can come back showing the player instead. `isolate = true` draws the subject without the other geometry. It changes exactly that — lighting, sky and post-process are untouched, so a final frame stays a final frame, and a flat read of the geometry is a `pass`. Excluded geometry still casts shadows onto the subject and still bounces light into it, the same as any render-layer exclusion. In a `setups` collage each cell shows the entity its own set-up frames alone, and a set-up can state `isolate` for its own cell; a set-up aimed from a position or a camera frames no entity and is refused when it is to be isolated. ## The tone the frame carries Every capture reports `luma` on its cell record: the tone of the pixels it just encoded, in code values on the 0-255 scale the image is delivered at. ``` luma = { pixels, min, max, mean, p1, p5, p50, p95, p99, span, spread, crushed, clipped } ``` `span` is `max - min` — the whole range, a single stray pixel included. `spread` is `p95 - p5` — the range the body of the picture occupies, which is the reading that answers whether a shot is legible. `crushed` and `clipped` are the shares of the picture sitting at code 0 and at code 255. A frame's mean cannot answer that question: half a picture at 40 and half at 200 carries the same mean as a flat field at 120, and only one of them has anything in it. A subject can be modelled, shaded, lit and drawn and still arrive inside a handful of code values, in which case the image looks like a wash and every other field here reports success. `spread` is the number that says so. The same reading is available anywhere pixels are: `cpu:tone()` on a `TextureCpuHandle` — from `renderer.texture.readback(rtHandle)` on a capture handle, from a render-to-texture camera's target, from a CPU canvas — and `cpu:histogram()` under it, the per-code-value counts for luminance and for each of the three channels, which every reading above is a sum over. `cpu:histogram(x, y, w, h)` counts one rectangle of the picture alone, so a reading per region of the frame — how much of each cell is dark, lit or a given colour — is one call per region. `cpu:flip(reference)` holds the picture against another of its size by NVIDIA FLIP, the comparison a golden is judged by; `@builtin::modules.postMetrics` carries it with the golden runner, the pixel metrics (banding, temporal variance, halo overshoot, ring energy, hue drift) and the scopes, each reading a presented frame or a capture result as it comes. ## The grade an offscreen frame carries A capture spawns its own camera and renders offscreen, and the post-process chain runs over that frame the way it runs over the viewport's — for a station, for an orbit around an entity, for a named camera, for a render-to-texture camera alike. Inside an effect, `engine.view_proj` / `engine.prev_view_proj` / `engine.inv_view_proj` are the camera THAT render was drawn from and `engine.resolution` is that target's own size, so a grade that reconstructs world space from scene depth reconstructs against the station and lens the capture asked for. That makes an offscreen capture an oracle for an authored look: it photographs a chosen station and lens without taking the on-screen camera away from whoever else is driving the scene, and moving that camera somewhere else does not move the frame this one returns. An offscreen render keeps no view history of its own, so `engine.prev_view_proj` there holds that same matrix rather than the frame before it, and a pass taking camera motion from the two reads none. On the viewport the pair is a frame apart, so a camera's motion (and the blur and temporal reconstruction built on it) is read from the presented frame: `source = "presented"`. A render feature's passes reach an offscreen render the same way, against the same camera — the `renderFeature` assetType README carries that half, including `@frame.camera`, its own per-target camera input. `render(ctx)` itself runs once a frame for the main viewport, so `ctx.viewport` is the window's size rather than the captured target's; a feature sizes its scratch with `screen = true` and lets the engine track the drawn target. `postProcessing = false` takes the chain off the frame a capture returns, and with it the film a render feature draws — the flare and grain, a short focus's bokeh, the shutter's smear. It is a `Camera` property rather than a render layer, so `renderLayers` does not reach it. A capture naming a diagnostic `pass` drops the chain by default, so the buffer comes back as the value it encodes. ## Collage `collage` renders a grid of frames into one image. What varies across its cells is what you pass — `setups` for a cell per whole camera set-up, `viewpoints` for a cell per view of one subject (or `"sides"` for all six), `passes` for a cell per render pass, `duration` for a cell per sample over time. A collage with no axis is refused rather than quietly sampling time. `setups` is the shot list: each entry is written the way a single capture is aimed (`source`, `camera`, `entity`/`entities`, `position`, `lookAt`, `rotation`, `distance`, `angle`, `margin`, `viewpoint`, `basis`, `projection`, `orthoHeight`, `fov`, `near`, `far`), plus `label` to name the shot on the result, and a field an entry leaves out comes from the collage's own options. `setups` and `viewpoints` are one axis — both say where the camera stands — so a request names one of them; either crosses with `passes` into a 2D grid, a row per station and a column per pass. The shot most of this exists for is one call: six sides of one object, each an orthographic elevation, nothing else in frame. Authoring the cameras themselves — creating, listing, aiming, framing, and setting their parameters — is the `camera` toolbox.

# debug Surface recent problems, inspect entity component state, and control the ambient debug-visualization overlay. - **problems** — return the errors and warnings the engine has logged since your last call (contract violations, `awake`/`update` throws, auto-disabled components, unresolved requires, bridged renderer/asset errors). Advances a read-cursor, so repeated calls report only what is new. **When something you built does not behave, call this first.** - **inspect** — component state for every entity matching a target (id, name, glob, proxy, or an array): script components as serialized public data, native ECS components as field values, or both — selected by `include`, filterable by `types`. In the editor the overlay draws gizmos for the whole scene — mesh bounds, camera frustums, skeleton bones, physics colliders, icons for meshless entities (audio sources, cameras, lights), and mesh wireframes. It is ambient (no per-entity setup) and agent-driven: - **set** — show or hide a category (`bounds`, `frustum`, `bones`, `colliders`, `icons`, `wireframe`), or `all`. - **scope** — focus the overlay on specific entities (by name or id); call with nothing to clear the focus and draw for everything. - **playSession** — show the overlay in play mode for a temporary, never-saved session, so bones/colliders/bounds stay visible on a running, animating scene. - **state** — report which categories draw, the current scope, the play session, and the engine mode.

# worldValidation Validate a world's authored content — the cargo-check equivalent for scenes, scripts, and assets — and report what's broken. Run a full pass over the world plus every imported library, or narrow it: the world's own content only (everything under source except the libraries), every imported library, or one named library. Generate a report with structured filter and format options — scope, minimum severity, category, code, path pattern, source, placeholder inclusion, a problem limit, and a render format (human, markdown, JSON, or summary) — and optionally write it to a file. Or ask for just the one-line health summary with no problem list. Each pass returns the validation report alongside a one-line summary.

# topics/rendering guide Reference docs surfaced through `guides { path: "..." }` and `man topics/rendering`. The content lives in `guide.md`; this README is just the asset summary.

# wld Manage the live world while you work — flip between edit and play, and compose additive scene layers. Flip into play mode — which snapshots the edit state, then starts scripts and physics — and back to edit, which restores that snapshot and stops scripts; query the current mode any time. Load a scene as an additive layer alongside the current one, unload it, or show and hide layers without losing their entities. List the worlds saved on disk, and read the current world state: name, mode, active layer, and available scenes. To list the loaded scene layers or the scenes available to load, use `scene.loadedLayers` and `scene.list`. Saving and loading scene snapshots is the `scene` toolbox (`scene.save` / `scene.load` / `scene.list`).

# tests Run the engine test suite from inside a live engine and triage the results — the baseline-then-regression loop, without leaving the session. List every registered test suite (optionally with test counts) to see exactly what the runner and CI will execute. Run everything, one suite, or a subset — each suite runs through its own cooperative coroutine so multi-frame tests yield correctly — and get back structured results plus Markdown and JSON reports written to the VFS. Compare two JSON reports to surface the failures that are new since the baseline, which is how you catch regressions a change introduced. The lower-level framework for authoring suites lives in the test module and the `testSuite` asset type.

# scene Work with the live scene across its whole lifecycle: start one from a template, understand its structure, reshape its player, clean it, and set it as the world's entry point. ## The player model A scene is either **playerless** (`"none"`) or **spawns a player per joining user** (`"spawns"`). The player is defined by a **PlayerPrototype** subtree — a body (the avatar a joining user becomes) and a camera rig. Zero is like Roblox here: the default character is a socket a joining player's own avatar replaces, so **reuse the canonical player setup rather than hand-rolling one** — a divergent player breaks multiplayer. `scene.summary` and `scene.tree` always report the scene's player intent, so you can see a scene's type without knowing this tool exists. ## Start a scene - **create** — create a NEW scene from a typed template (`empty`, `flat`, `player`, `menu`) and load it. Omit the template to get the catalog of what each contains. `opts.startup = true` also makes it the world's entry scene. The starting point is a choice, not a scene handed to you silently. - **setStartup** — make the currently active scene the world's startup scene, so it survives world reloads. No argument, no clear. ## Understand a scene - **tree** — the entity hierarchy as an indented tree, led by a one-line player-state header. The quick "what is the current structure" view. - **find** — the single entity-discovery tool over the whole live population. Search by `name`, `component`, or `attribute` (each a list), scope by hierarchy or scene, and add projections with `include`: `transform` (`position`, `rotation`, `eulerAngles`, `scale` as arrays, read `position[1]`), `componentValues` (each component's live field values), `attributes`, `scriptComponents`. Every match carries `components` (the component names) without one. Returns `{ matches, total, truncated, searchedAxes }`. - **summary**: entity + root counts, the top-level rows (at most `limit`, 20 when omitted), the active camera, the **player** setup (intent + counts + the verb to change it), and the world's **startup** scene. `under` scopes the counts and rows to the subtrees of named entities, and `fields` keeps only the named fields, so `{ fields = { "entityCount" } }` answers the count alone. - **list** — the SCENES you can load, so you find one by name instead of guessing. `scope` filters `world` (default, this world's own scenes), `library` (builtin `@builtin::` templates / demos / canonical setups), or `all`; each entry says whether it is currently loaded / active. - **loadedLayers** — the scene LAYERS composing the live scene right now: the root scene plus any additive overlays (editor UI, effect passes), each with its backing scene, status, visibility, and additive flag. A layer whose load produced failures reads `status = "partial"`. ## See what a load did, and what a scene costs - **observe** — what every load did and what each loaded scene costs. `lastLoad` is the last load's report: what it produced, the root it replaced and the overlays that went with it, the entity counts on each side, the milliseconds each phase took, and every failure with the entity, component or lifecycle hook it is about. `layers` counts what the engine attributes to each layer, `cost` is each layer's per-frame entrypoint tick summed across the window `window` reports. - **whyPartial** — why a layer is not whole: the nearest cause from a closed set, the engine's own words for it, and the full failure list. The answer to "the scene loaded but something in it is missing". - **cost** — which loaded scene is eating the frame time, ranked by what each layer's entrypoint tick costs. A scene entrypoint's per-frame `update` / `editorUpdate` runs on the scheduler rather than as a component, so `profiler.scripts` does not list it; this does. Pass `true` to open a fresh window, let frames pass, then call again. ## Reshape the player - **player** — read the setup, flip the typed intent between `"spawns"` and `"none"`, and reshape: `body` adopts an entity or an avatar asset ref, `camera` wires an entity carrying a Camera, `behavior` sets the player camera's behavior to any `cameraBehavior` component. `"none"` removes the prototype and spawns; `"spawns"` authors the canonical setup when missing. - **spawns** — the spawn POINTS: where joining players arrive, as opposed to `player`, which decides whether they arrive at all and what they become. No argument reports every PlayerSpawn with its transform, the prototype it instantiates (and whether that reference still resolves), and its team / role / maxPlayers / spawnPolicy / placement. `"add"` places one, `"set"` reconfigures one (omitted fields keep their value), `"remove"` deletes one; `opts.spawn` picks WHICH when the scene has several. A spawn's transform is where a body lands and its forward axis is the heading it faces — the edit-mode overlay draws that ring, arrow and standing capsule. ## Change the content - **clean** — clear the scene down to a protected keep-set. By default only the player setup survives (never orphaned); `keep` protects a `scene.find` selection (e.g. `{ component = { "Light" } }`); `keepPlayer = false` wipes the player too; `dryRun` previews. Only the active scene layer is touched. - **remove** — delete entities by name, id, array, or a `scene.find` query, subtree-deep. Removing the default floor is `scene.remove("ground")`. - **save** — save a loaded scene back to its own `scene.json`; a no-arg call reuses the last scene name. - **load** — load a scene snapshot into the live world. - **reload** — re-run a loaded scene's build after editing its entrypoint. `load` of an already-loaded, unchanged scene is an idempotent no-op; this rebuilds it, in edit mode. Omit the name to reload the active scene. A rebuild saves the scene, as a write to `build.luau` does, so `scene.json` holds what the reload built and the next play loads it. A library scene (under `/libs/`) is rebuilt in the live scene alone. - **replace** — replace a loaded scene's content with another scene's content, keeping the target's identity. Supports `dryRun`. ## Maintenance - **migrate_v6_to_v7** — convert a v6 scene's `scene.json` to the v7 format in place. User-initiated only. World mode (play/edit) and additive layers are `wld`; accepting or rejecting play-mode changes is `sceneAuthoring`.

# assetType (asset type) The `assetType` is the **type of types** — the meta-type that every `<typename>.assetType/` folder is an instance of, including itself. A `<typename>.assetType/` folder declares a new asset type called `<typename>`: its on-disk shape, its identity rule, and the structural contract instances of the type must satisfy. This type is self-hosting. `assetType.assetType/` conforms to its own schema — a `type.yaml` plus a `README.md` — so the type system is defined in terms of itself rather than a hidden engine special-case (the same way a C compiler is written in C). Every asset in the engine, including each `.assetType` definition, now resolves to a registered assetType: a `.material` resolves to `material.assetType`, and `material.assetType` resolves to `assetType.assetType`, which resolves to itself. ## Why it exists Assets used to be mapped to their type *implicitly*, by matching the folder suffix against a name-keyed registry. That match carried no stored link: nothing on a `Foo.material` recorded *which* `material.assetType` it was written against, so two worlds shipping a same-named type could silently disagree. Making `assetType` a real, resolvable asset closes that gap — the link from an asset to its type is now an explicit reference pinned in the asset's `.refs` sidecar (`via: "asset_type"`), exactly like every other dependency. ## Where it lives - Source: `/zero/source/.../<typename>.assetType/` - Identity: `<typename>` (the `.assetType` suffix strips from the identity; the folder retains the suffix on disk). - Folder shape: - `type.yaml` — the structural spec the validator reads. **Required.** - `README.md` — type-level documentation. **Required.** - `about.yaml` — three short plain-language texts saying what the type is for (see "Say what your type is for" below). Every built-in type ships one. - `behavior.luau` — optional behavior/scaffolding module. - `template/` — optional canonical placeholder body the templater copies into a new asset of the type. - `<name>.module/` — optional shared code the type ships to **every instance** of it, conventionally `shared.module/` (see below). ## Shared code its instances reach (`asset.containing` + `.modules`) A type ships code every instance of it uses by declaring the modules in its own `behavior.luau`, as a map of tracked `require`s: ```luau -- inside <typename>.assetType/behavior.luau M.modules = { shared = require(".shared") } -- the sibling shared.module/ ``` An instance reaches it with one location-independent line: ```luau -- inside any foo.<typename>/init.luau local api = asset.containing(__FILE__).modules.shared ``` `__FILE__` is the chunk's own VFS path; `asset.containing` resolves the calling asset, and `.modules.<name>` follows the instance's pinned `typeRef` (the type's **guid**, recorded in the instance's `.refs` as `via = "asset_type"`) to the module its type declares. Because resolution follows that link rather than a category name, the same line in an instance of a different type resolves to that type's module, and identically-named modules in two types never collide. `asset.typeRef(target)` reads that guid for any target; `asset.resolve(asset.typeRef(target))` gives the full `AssetRef<assetType>`. See `assetTypes/README.md` § "Shared code: type-owned modules" for the full mechanism. ## How to create one Authoring a new asset type is just creating a `<typename>.assetType/` folder under `/zero/source/`. The engine registers `<typename>` as a category **the moment the folder is written** — no engine restart, no Rust change, and it works for built-in (`@builtin/assetTypes/`) and user-defined types identically. Once `<typename>` is registered, `<name>.<typename>/` folders are recognised as assets of that type and resolve their `asset_type` link back to this definition. Creating a `<name>.<typename>/` folder *before* its `<typename>.assetType/` exists fails fast: the suffix has no registered type, so the folder can't be a valid asset. Author the type first. ## Say what your type is for (`about.yaml`) Every type carries a small `about.yaml` beside its `type.yaml`, and the README you write for the type documents how to author an asset of it. Both are required, and both are meant to be written as you make the type, while you know what it is for. `about.yaml` answers three questions in plain words, one key each: ```yaml intent: "Sets the typeface of the text in your game: UI, labels, signs and 3D lettering." function: "A font, from a TTF, OTF or WOFF file, made available to all of Zero's text, including 3D text." usage: "You choose it by name for your UI or your text, and the text appears in that typeface." ``` - **`intent`** is what the type is for in someone's game. Name the whole range of things it covers, in the words a game maker would use for them. - **`function`** is how it works today: what an asset of the type is made of and what the engine does with it. - **`usage`** is what a person or an agent does with one: how they pick it, place it, call it, or hand it to something else. A few habits keep it useful. Write about the type as a whole, so the text holds for every asset of it, and leave out any one asset's name. Say what the type does right now. Stay with words a game maker would say, and mention a file name or a call only where a game maker would know it. The limits are in bytes: `intent` 800, `function` 1500, `usage` 1200. Each is generous, and a few sentences usually say it. Why bother: agents working in a world read these three texts to learn what your type is for, so they reach for it when it fits and use it the way you meant. People searching find the type by what they want to make, and the three texts let them tell at a glance whether it is the one they need. `asset.create("assetType", name)` scaffolds the file with all three keys, and a README whose first paragraph is a `[[describe]]` note. Each value starts with `[[describe]]` followed by a note on what to write; replace the whole value with your own words. A value or a README line that still starts with `[[describe]]` counts as not written yet, which the push's check of each asset it carries against its type reports, naming the file. A type you leave as it was is not asked, so older types stand. Many types at once, or one you would rather not write by hand: `services describe` lists the types that lack these texts and what filling them costs, and `{ mode = "all" }` writes them into the same files, where you edit them like anything else you wrote. Where the texts show up: `asset.inspect("<typename>", "assetType")` returns them as `detail.about`, `assetTypeRef:getAbout()` returns them as a table with the keys `intent`, `function` and `usage`, the type's `types/<typename>` guide page ends with an "About this type" section holding them, and the Inspector's Type section lists them at the top. The guide page is this README plus that section, so a README that explains the type well and an `about.yaml` that says what it is for read well together. ## Ask your assets to say what they are A type can ask each of its assets to carry a description: a file, written by the author or generated, that says what that asset is and what it is for. The people and agents who find an asset read it to decide whether it fits, and the author knows what it is for at the moment of making it. **1. Declare the file with a `described` check.** This is the whole requirement: ```yaml files: required: - path: "README.md" # any file name works kind: "file" content: - { check: "described", message: "Say what this widget is for and where it goes." } ``` With that in place: - An asset whose file is missing, empty, or still has a line starting with `[[describe]]` does not meet the type's requirements. `asset.validate` names it, `zm status` and `world.publishBlockers()` list it, and a push that carries it is refused with `message`. - `services describe` finds it (`mode = "find"` lists it and the price) and can write the description for it (`mode = "all"`). It writes only this file. For a YAML file that holds several texts, name the keys and their limits in bytes, the way this type does for `about.yaml`: ```yaml content: - check: "described" keys: - { key: "intent", max_bytes: 800 } - { key: "usage", max_bytes: 1200 } message: "Write intent and usage." ``` `headings` names sections a Markdown file must have. `assetTypes/README.md` has the full check. **2. Optionally, steer what gets generated with a `describe:` block:** ```yaml describe: guidance: "Say what the widget shows, what input it reacts to, and where it fits in a UI." examples: - "<identity of an asset of this type whose description reads well>" ``` `guidance` says what a description of your type covers, and `examples` names assets whose descriptions show the shape you want. Without the block, generation still works from the general describing guide and the check's `message`. An example is an asset of your type in the library whose description is a true, written one, never a scaffold's placeholder, and never the asset being described. When the library has none, leave `examples` empty rather than name one that is not. A type whose assets the model cannot perceive (a sound it cannot hear, a motion it cannot see) says so in `guidance`: the description comes only from what the asset's own files state, such as its name and the `.metadata` where a generation service records the prompt that made it, and when they state nothing the model answers `cannot` with that reason. `services describe` then fails that one asset with the reason and leaves its file as it was. **3. Start the file unwritten in your template.** Put a heading and one line that starts with `[[describe]]` and a short note on what to write, in `template/<file>`, so a new asset starts unwritten rather than with text that only looks written. Do not put example content in it: every new asset carries the template verbatim, and the same text on many assets makes them impossible to tell apart in search. Types whose assets say what they are through their own content (a mesh, a texture, a sound) leave this out: declare nothing and nothing is asked. ## The `behavior.luau` contract `behavior.luau` is the type's code. It returns ONE table, and the keys below are the ones the framework reads off it for EVERY type — the surface you get by declaring them. Anything else on the table is your own: reachable by name through `require("@builtin::assetTypes.assetType.shared.ref").loadTypeModule("<typename>")` for a system that knows your type (`computeShader` publishes `dispatchKey` that way), and never called by the framework. ```luau local M = {} M.ref = { ... } -- methods on every AssetRef of this type M.modules = { ... } -- shared code the instances reach M.global = {} -- reserved key; ship the empty table function M.onCreate(name, opts) ... end return M ``` ### The tables | Key | Shape | What the engine does with it | |-----|-------|------------------------------| | `M.ref` | `{ [name] = function(self, ...) }` | Every function becomes a method on every `AssetRef` of this type: `ref:name(...)`, with `self` the ref. This is how a type gives its instances an API. Two names in here are contracted — see below. | | `M.modules` | `{ [name] = require(".sibling") }` | Shared code the type's instances reach through `asset.containing(__FILE__).modules.<name>`, resolved by the instance's pinned `typeRef` guid. See the section above. | | `M.refShapes` | `{ [method] = "TypeName" }` | Declares the result type of a `M.ref` method so the LSP can check what a call site does with it. See the `M.refShapes` section below. | | `M.events` | an events schema | The events an ASSET of this type fires, declared the way a component declares its own. Readers subscribe through `ref.events.<name>:connect(...)`; firing authority stays with the type. Keyed by the asset's guid, so every resolver of the same asset shares the signals and a re-resolved ref re-attaches to subscriptions already there. | | `M.namePattern` | a Lua pattern string | The name shape `asset.create` enforces for instances of this type, replacing the default `^[A-Za-z][A-Za-z0-9_]*$`. | | `M.global` | `{}` | Reserved. Ship the empty table. | ### The lifecycle hooks Each is optional; a type that omits one costs nothing. The engine calls them: | Hook | Signature | When it fires | |------|-----------|---------------| | `M.onCreate` | `(name, opts) -> { [filename] = contents }` | On `asset.create("<type>", name, opts)`. Returns the files that scaffold the new instance, as a map of relative path to contents; the framework writes them. The `opts` type annotation is the schema `asset.create` validates the caller's arguments against, so annotate it. | | `M.onReplace` | `(dest, files) -> { [filename] = contents }` | On an `asset.create` that writes over an asset already at `dest`, after `onCreate`. Receives the files `onCreate` produced and returns the files to write, so a type that re-derives its content (a rig re-imported from its source file) can carry what an author added to the asset it replaces. Optional. | | `M.onRegister` | `(self)` | Once per instance, the first time it registers: on `engine.onWorldLoaded` for instances already in the world, immediately on a live `asset.create`, and when an instance arrives after that sweep (installed, pulled, or synced from a peer). Guarded by the instance's path, so a double trigger never double-registers. An instance that leaves the runtime releases the guard, so one arriving at the same path again registers. This is what makes "write an asset into the world and it takes effect live" work for content that registers into a runtime registry. The `@builtin` library is excluded from the sweep. | | `M.onChange` | `(ref, change)` | On every VFS write to a file INSIDE one of this type's instances. `change` is `{ path, asset, type, kind, origin }`: `kind` is `"edited"` or `"seeded"`, `origin` is `"local"` or `"remote"` (an importer runs on the originator only). **Filter on `change.path`**: the hook fires for any file under the folder, so act on the one you care about. It runs as its own task and may wait on what it reads: a `vfs.read` of bytes only the blob store holds (a binary file on the web build) fetches them. The writes to one asset reach the hook in the order they were made, each once the hook before it has finished. A re-entrancy guard suppresses writes back into the same asset routed inline from the hook, and a write the hook makes is routed on a later frame like any other, so the hook must be convergent on its own: diff the meaningful state and short-circuit while your own regeneration is in flight. | | `M.onDelete` | `(ref)` | When an instance of this type leaves the runtime: it is deleted, or the engine swaps away from the world binding that held it. Its files are gone from `/zero/source` by then. A removal routed while the instance's `onChange` is waiting reaches `onDelete` once that hook, and every write queued behind it, has finished. On a swap the instance still exists in the world it belongs to, and a VFS write the hook makes reaches the world now bound. | | `M.validate` | `(assetRef) -> { { code, message, severity? } }` | On `asset.validate`, after the structural `type.yaml` check, for the type's own SEMANTIC validation. `severity` defaults to `"error"`; error-severity problems flip `ok` to false, warnings do not. A hook that raises or returns a non-table is itself reported as a `validate.hook_failed` error. `world.push` runs this per user asset, so declaring it enforces your type's rules at publish time with no further wiring. | ### The two contracted `M.ref` names Most `M.ref` methods are your type's own surface: name them what you like, return what you like. Two are read by the engine and mean the same thing for every type, so their shape is fixed: - **`instantiate(self, target?, opts?) -> (root, idMap)`** — makes an instance part of the scene. Defining it is the whole opt-in: `ref:canInstantiate()` is true exactly when it exists. Its **return is a contract** — see the next section. - **`inspect(self) -> detail`** — the type-specific half of `asset.inspect`. See its section below. ### Names a type cannot shadow Some keys resolve on every `AssetRef` before per-type dispatch, so an `M.ref` entry of the same name is never reached: `canInstantiate`, `getSource` / `getBytes` / `getText`, `exists`, `deps`, `meta`, `runtime`, `events`, `modules`, `typeRef`, and the residency flags `has_backing_asset`, `has_runtime_changes`, `cpu_resident`, `gpu_resident`. These are the behaviours every asset must expose identically, which is why they win. ### Reloading A `behavior.luau` edit ripples to existing refs with no restart: dispatch reads `mod.ref` fresh each time and `require`'s cache is re-run in place. A type whose folder appears at runtime is picked up on the next dispatch — negative results are not cached either. ## The names your type's assets hold (`references:`) An asset that names another asset by a string depends on it: the named asset travels with it into a pack or a pull, and a name nothing answers to is an unresolved dependency `worldValidation` reports and the push gate refuses. The engine reads `require(...)` and the asset lookups on its own; a name your type's format or API carries anywhere else is one you declare, in `type.yaml`: ```yaml references: # A top-level key of this type's own files that holds a name (or a table of them). keys: - { key: "shader", type: "shader" } # A call's argument, counting from 1, wherever the call is written. args: - { call: "g:node", position: 2, type: "procNode" } # A field of the options table a call takes. fields: - { call: "fx.beam", field: "material", type: "material" } ``` `type` is the category the name resolves in, one of `asset.categories()`. An argument declared once per category, at the same `call` and `position`, takes the name from whichever category answers. A key or a field holds one category: declared again with another `type`, it takes the later one. The block is declared as the type registers, when the library or the world holding it is laid into the VFS, so every reference table reads it whatever script has or has not run. `asset.declareReferenceKey`, `asset.declareReferenceArg` and `asset.declareReferenceField` declare the same things from a script, for a surface that is no type's. ## What search does with your type (`indexing:`) A type declares what search embeds for its instances. **This is not optional and there is no useful default**: ZeroMind runs your declaration and adds nothing of its own, so anything you do not point at is absent from the index — silently, and for every instance of the type forever. Twenty types once shipped with no `indexing:` block and every instance of them was unfindable. A type whose instances are never searchable (a run record, a log) says so with `indexing: none`; every other type declares a recipe. A type without an `about.yaml` beside its `type.yaml` refuses every push that carries it, whether or not the push edited the type. Texts in it that are not written yet follow the rule every `described` file follows: they refuse the push that edits, stages or commits the type. Two words carry the whole thing, and they are the same two the search tool exposes: - **`identity`** — what an instance IS. For anything you can look at, that is the picture (`content:` + `modality: image`, matched directly by a text query). For everything else it is the authored text that says what it is. - **`capability`** — what it DOES and how it is made. Code, settings, the model-written summary of them. A type with no behaviour has no `capability` entry. A type with nothing to look at has no image. Leaving a slot empty is a statement; filling it with whatever happens to be lying around is not. The one you have to decide when authoring a type is: *for an instance of this, what is the thing a person would recognise it by, and what is the thing it does?* A texture answers "the image" and "its compression settings". A module answers "its README" and "its code". Write those two answers into `indexing:` and the rest follows. Three specifics worth knowing before you write one: - **`derive` is instructed by this type's own `README.md`.** A model reads an instance's source and writes a description; what it embeds is the description, not the source. It follows your README to know what it is looking at, so a type whose README says what its instances are gets good derivations for free. There is no prompt to name. - **`derive` loses detail.** It keeps the main technique and drops secondary ones. If the source is code you want searchable by its own vocabulary, declare the same files a second time as `capability` with `extractor: verbatim`. Two entries, same role, different failure modes. - **Never `source: { field: name }`.** A filename is one or two words with no usable embedding, and a corpus indexed that way answers "language runtime" with `say_runtime_2.soundClip`. `facets:` is the second stage: a `file:` embedded whole (`.metadata`, so keys nobody declared still get indexed) plus computed keys that state a value for every instance including `false` — the raw file can imply "not rigged" and can never say it. The full schema, every extractor and the facet sources are in [`assetTypes/README.md`](../README.md); the scaffolded block with inline guidance is in [`template/type.yaml`](template/type.yaml). ## A container whose contents work together (`system: true`) A type that is a container can say its contents are one working whole by declaring `system: true` at the top level of its `type.yaml`. `package` declares it: the components, modules, scenes and tools a package ships are parts of one system. The declaration is plain data, read from the decoded `type.yaml` (`require("@builtin::assetTypes.assetType.shared.declaration").of("package").system`), and a type that leaves it out is an ordinary type. ## What your type looks like (`preview:`) A type declares how its instances are pictured, and the engine makes the picture when something needs it: staging an instance whose `preview.png` is missing or pictures content it no longer holds queues a render, a view that shows the instance asks for one, and `ref:preview()` renders one on demand. An edit renders nothing on its own. Each `preview.png` records the content it pictured, which is how a stale one is told from a current one. The preview studio owns the light, the backdrop, the framing and the camera, and nothing in the world reaches it, so an instance pictures the same in every world. Declare one of four recipes: | `kind` | The picture is | Your type provides | |---|---|---| | `instantiate` | the asset stood in the studio through its own `instantiate` | `ref.instantiate` | | `image` | a file the instance holds, fitted to the frame, nothing rendered | `source.files`, the candidate files in order | | `surface` | the asset painted on a builtin mesh | `mesh`, and `ref.previewMaterial(self) -> { material }` | | `custom` | whatever `ref.preview(self, opts)` returns | `ref.preview` | ```yaml preview: kind: surface mesh: sphere ``` A scene-instantiable type that declares nothing still answers `ref:preview()` through its `instantiate`, but persists no picture: the declaration is what makes the engine keep one. List `preview.png` among the type's files so an instance may hold it. The picture is findable only when `indexing.content` names it. The two are separate statements, and neither is derived from the other: ```yaml indexing: content: - role: identity modality: image source: { file: "preview.png" } ``` `@builtin::assetTypes.assetType.shared.previewRecipe` reads the declaration; the scaffolded block with inline guidance is in [`template/type.yaml`](template/type.yaml). ## Typing a result from the instance (`M.refShapes`) Every method on `M.ref` is checked against the type its `--!return` names, and every asset of the type gets the same one. That is right for most methods and wrong for the ones whose result is shaped by the ASSET: a method answering one entry per child folder, per row of a config, per binding an input map declares. The widest true annotation for those is `{ [string]: Thing }`, and an indexer accepts every key — so a caller's typo reads as valid and nothing reports. `M.refShapes` lets the type state the result per instance. Each entry is `function(self) -> (typeExpression, source?)`: a Luau type expression for THIS asset, and the module whose type vocabulary the expression uses. Read the asset; never run it — stating a type must not take effect. ```lua -- inputMap.assetType/behavior.luau — the shipped example. -- `activate()` hands back one handle per control the map declares. M.refShapes = { activate = function(self): (string, string) local names = {} for _, control in ipairs(M.ref.controls(self)) do table.insert(names, control.name .. ": Handle") end if #names == 0 then return "", "" end return "{ " .. table.concat(names, ", ") .. " }", "@builtin::modules.zinput.scheme" end, } ``` With it, `map.jump:onPressed(fn)` resolves and `map.noexisting` is reported with the map's real controls. Without it, both pass silently. ### What the call site has to say A per-instance type is matched by the asset's IDENTITY, so it applies only where the reference the method is called on says WHICH asset. Two spellings do: ```lua -- a component field, whose declared default names the asset public = { map = Field.assetRef("inputMap", "@builtin::inputMaps.default", Sync) } local controls = public.map:activate() -- typed for THAT map -- an annotation, naming category and identity local m: AssetRef<"inputMap", "@builtin::inputMaps.default"> ``` `asset.resolve("<identity>", "<category>")` carries the CATEGORY, so the methods the category defines are checked on the result — but not which asset it found, so a per-instance result keeps the category's declared return. A reference that names no asset (`asset.resolve(someVariable)`, a value passed in as a parameter) carries neither, and calls on it are unchecked. That is the same rule everywhere: the checker states what the code states, and a name computed at runtime states nothing. ### The result is a value, not a binding The type belongs to the expression, so it travels the way any inferred type travels. Indexing the call's result is checked; stashing it in a module local and indexing it from another function is not, because the local's declared type is what carries across — and `any` (or no annotation) carries nothing: ```lua -- checked: `d.greting` is reported function awake() local d = public.dialogue:lines() d.greeting:say() end -- NOT checked: `lines` is a module local typed `any` local lines: any = nil function awake() lines = public.dialogue:lines() end function start() lines.greting:say() end ``` Bind what you need at the call site and the names stay checked in both places — the asset-derived type at the boundary, ordinary scoping after it: ```lua local lineGreeting = nil function awake() local d = public.dialogue:lines() lineGreeting = d.greeting -- `d.greting` is reported here end function start() lineGreeting:say() end ``` ### When nothing is reported and you expected something From a call site, a type that is correct and a type that was never published look the same: no diagnostic either way. `require("@builtin::assetTypes.assetType.shared.refShapes").published()` answers which it is. It returns two maps — the category surfaces, and the per-asset results keyed by identity and then by method: ```lua local shapes = require("@builtin::assetTypes.assetType.shared.refShapes").published() shapes.categories.dialogue --> "{ lines: () -> any, lineNames: () -> { string } }" shapes.returns.Shopkeeper.lines --> "{ browsing: Line, farewell: Line, greeting: Line, wares: Line }" ``` An identity absent from `returns` was never published for, which is a different thing from published-and-correct — and the reason to look here rather than at the call site. The answers are recomputed when an instance is written, so a control added to a map reaches the checker with no restart. Only types that declare `refShapes` are read per instance, and at most 64 assets of one type are — past that the type keeps its declared return and the engine log names what was dropped. ## The instantiate hook (`M.ref.instantiate`) — a contracted return A type opts into becoming part of the scene by defining `instantiate` on its `ref` table. That is the whole opt-in: `ref:canInstantiate()` is true exactly when the hook exists, so consumers offer a scene path — an `Asset.source` field, a viewport drop, a tool argument — by CAPABILITY rather than by a list of type names. Unlike `inspect`, this hook's **return is a contract**. A caller writes one piece of code against every instantiable type, so what comes back cannot vary by type: ```lua -- behavior.luau local Instantiable = require("@builtin::assetTypes.assetType.shared.instantiable") local M = {} M.ref = { instantiate = function(self, target, opts) local root = Instantiable.root(self, target, opts) -- IN: the base opts -- ... compose whatever this type is, under `root`, NOW ... return Instantiable.result(self, root, idMap) -- OUT: the contract end, } return M ``` **IN.** `target` is an owning entity ref: the instance lands under (or, for a hierarchy type, onto) that owner. With no target the type spawns a fresh root. `position`, `rotation`, `scale`, `name` and `temporary` mean the same for every type — `Instantiable.root` applies them to a root you mint, `Instantiable.place` to one you adopt, so you never re-read the spec. Honour more opts of your own if your type needs them, and document them in the hook's `--!arg` docs. **OUT.** Return `(root, idMap)` through `Instantiable.result`: - **`root`** — the composed root, as an `EntityRef`. Composition is **synchronous**: everything your type builds is live when you return, so the caller can parent to it and read its components in the same statement. Do not defer the work to a component's `awake` and return an empty shell — a root that is not live is refused. - **`idMap`** — the `originalId -> runtimeId` map naming what you spawned, or nil for a type with no addressable children (`result` normalises it to `{}`). **A caller never receives nil.** A component that re-composes your asset on every load keeps this map and passes it back in, which is how a cross-entity reference into the composition survives a reload. `AssetRef` runs every `instantiate` through `result` on the way out whether or not your type called it, so returning something else fails at your own call rather than handing a caller a nil root. Calling it yourself is still how you say what you return, and running it twice changes nothing. `@builtin::assetTypes.assetType.shared.instantiable`'s README is the full reference for both halves. ## The inspect hook (`M.ref.inspect`) `asset.inspect(ref)` resolves `ref`, builds a common envelope shared by every asset (`identity`, `name`, `guid`, `source`, `typeName`, `typeDefinitionPath`, `scope`, `origin`, `description`, `tags`), then calls the resolved type's own inspect hook — `M.ref.inspect(self)` on `behavior.luau` — for the type-specific `detail`. The hook returns *only* `detail`; the framework fills in everything else: ```lua -- behavior.luau local M = {} M.ref = { inspect = function(self) -- self.path is the asset's own folder. Parse the asset's OWN -- files — source text, a native payload header, its `.metadata` -- — never runtime state (a live component instance, a GPU -- upload, a compiled schema): inspect must work on an asset that -- hasn't been compiled, uploaded, or registered yet. return { -- type-specific fields, whatever this type wants to surface } end, } return M ``` A type that omits `M.ref.inspect` yields the generic record: the envelope alone, `detail = nil`. That's a valid, working state — not an error — but it means agents calling `asset.inspect` / `tools.use("assets","describe", ...)` on an instance of the type see only identity and scope, none of what makes an instance of it meaningful. Declaring the hook is what makes a type's instances inspectable for what they actually are. A hook that raises leaves `detail = nil` and sets `record.warning` to the error text — `asset.inspect` itself never raises for a resolved ref. ## Discovery - `asset.list("assetType")` — every registered asset type. - `asset.inspect("<typename>")` — the type's identity, source path, and this README, plus its `about.yaml` texts as `detail.about`. - `asset.resolve("<typename>", "assetType")` — the `<typename>.assetType` asset itself (used by the validator to find the type's `type.yaml`). ## Related - `assetTypes/README.md` — the full `type.yaml` schema and validator rules. - `@builtin::assetTypes.assetType.shared.instantiable` — the instantiation contract's shared implementation, both halves. - Every other `*.assetType/` here — the concrete types built on this meta-type. ## The Inspector a type ships Every type names the file its Inspector sections come from with `inspector: { type: "inspector.luau" }` in its type.yaml. `asset.create("assetType", name)` scaffolds the block and an `inspector.luau` whose Contents section lists an asset's files, built from `ctx.parts`. `asset.validate` on a type that names none reports `assetType.inspector_required` (an error for a `@builtin` type the engine's library authors; a warning for a world's own, and for one the library carries by reference, which the world that published it answers for), and one whose file does not load `assetType.inspector_invalid`. ## In the Inspector Type: what the type.yaml declares (suffix, container, primary files, the settings and inspector blocks), whether the type ships a behavior, how many instances it has, and **Open** for its files. Schema: the settings every instance carries, as read-only typed controls at their defaults (a type whose schema is per instance says where it is declared). Instances: a link to each of the first 20 assets of the type, and how many there are when there are more.

# profiler.compare Diff two recordings made with `record` — the "did my optimization help, and where" view. Reports the change in avg / p90 frametime between a baseline (A) and a comparison (B), plus the systems that moved the most between them. A positive frametime delta means B is slower than A. Record a baseline, make a change, record again, then compare — this closes the autonomous optimize loop.

# profiler.flamegraph Call-stack depth, for when attributing cost to a component isn't enough. The sampling profiler walks the Luau stack at kHz and shows the hottest call stacks — across update loops and coroutines alike — so "which line inside this update is slow" gets an answer. Reach for it after `frame` / `scripts` or `task_scheduler` shows a component hot and you need to know exactly what it's doing.

# profiler.frame The current frame's time as a self-accounting tree — the deep view. Every schedule, system, and sub-timing shows its SELF time (own cost, excluding children), TOTAL, call count, and % of the frame; the frame-limiter / vsync wait shows as an explicit `present / idle` bucket instead of vanishing; and each heavy component's `update(dt)` is attributed by component + entity under `lua_update.vm_call`. Read the live frame with source `now`, or a stopped recording by its label (its worst or typical frame). Read from the top: the row with the largest SELF time is where the frame time actually goes.

# profiler.gpu GPU pass timings of the most recent resolved frame, measured with GPU timestamp queries. Rows are label-aggregated — every render/compute pass recorded under one label sums into a row with its call count: `compute.<shader>` per compute dispatch, `scene.*` for the scene passes (shadow, sky, gbuffer, main, transparent, deferred_light, oit), `post.<effect>` per post-process effect, `feature.*` for render-feature geometry/draw passes, plus `skinning`, `culling`, `compose`, `raytrace`, and `ui.egui`. The header shows the frame's GPU span (first pass begin to last pass end) next to the summed pass time — the sum can exceed the span when the GPU overlaps passes. The readback is asynchronous, so the report lags the live frame by a few frames. Use it when `frame` shows the time under `present / idle` (GPU-bound) and you need to know which passes the GPU spends it on. Takes `n` (rows, default 20). Reports `supported = false` cleanly on devices without timestamp queries. Steady per-frame workloads read reliably; a lone one-shot dispatch submitted in isolation can under-report on some drivers.

# profiler.hits Drain and review the frames a record-mode `watch` caught. The report is a DISTRIBUTION — every caught frame grouped by its dominant hotspot — so a scene that hitched twenty times over twenty seconds reads as "these three causes", not twenty separate anecdotes. All times are EFFECTIVE (the agent's own `execute` cost is excluded). Draining clears the buffer, so a subsequent `hits` returns only what was caught since. The snapshot is stored under a label, so you can drill into any cluster afterward with `frame <label>` or `hotspots <label>`. Arm the watchdog first with `watch <ceilingMs>`.

# profiler.hotspots The fastest triage — a flat, ranked list of the frame's most expensive nodes by SELF time (own cost, not because of children). One glance answers "what is eating the frame": a system's own work, or a specific component's `update(dt)`. Follow a hit down with `frame`. Reads the live frame (`now`) or a stopped recording's worst / typical frame by its label. Takes `n` — how many rows to return (default 12).

# profiler.memory The Luau VM's memory — total heap + GC state, then the components retaining the most bytes per instance. Frame time isn't the only budget: a script that grows its memory every frame is a leak that eventually stalls GC or crashes. Take it, play / test, take it again, and watch which component's retained bytes climb. Complements the frame-time tools — memory pressure shows up as GC cost in `frame` (the `gc` node) and as crashes under load, not as one slow system. Takes `n` — how many components to list (default 20).

# profiler.record Start / stop / check a background profiling recording that spans a play session — the path for "it hitches while I play", which a single live frame can never catch. `record("start", label)` begins capturing every frame; play or test across as many turns as you like; `record("stop", label)` ends it and returns the windowed breakdown — avg / p50 / p90 / p99 / max frametime, spike count, and the worst frame's attributed tree. `record("status", label)` reports an in-progress recording. Every read tool (`frame`, `hotspots`, `scripts`, `tasks`) accepts a recording's label as its `source`.

# profiler.retro Read the ring buffer's recent history RETROACTIVELY — the latency-immune profiler. Because the ring is always recording (once enabled with `ring on`), it does not matter that this call lands seconds after the spike: the data is historical. The report is a DISTRIBUTION — every spike frame grouped by its dominant hotspot — so one call shows the whole picture, not a single anecdotal worst frame that might be a random hiccup. All times are EFFECTIVE (the agent's own `execute` cost is excluded, so your profiling calls never masquerade as game hitches). The snapshot is stored under a label, so you can drill into any cluster afterward with `frame <label>` or `hotspots <label>`.

# profiler.ring Control the retroactive ring buffer — an always-recording, bounded history of the last N seconds of per-frame data. It exists because an agent's tool calls land seconds after anything happens: with the ring on, you read the frames AROUND a spike AFTER the fact instead of having to start a recording before it (which you can't, because you don't know when the spike will come). `ring("on", seconds)` starts it (default 20s), `ring("off")` stops and clears it, `ring("status")` reports how many frames and seconds it holds. Then read the history with `retro`. Editor profile only, and off until enabled — so it costs nothing until you ask for it.

# profiler.scripts Per-component `update(dt)` attribution — the "which of MY content is slow" view. Where `frame` / `hotspots` show engine systems, this drills the one system that runs authored code (component update loops) down to the exact component instance and entity. Rolled up per component type by default (many instances collapse to `Type xN` with summed cost; a lone instance keeps its `@ entity`); pass a `type` to drill into that type's individual instances. Use it after `hotspots` / `frame` point at `lua_update.vm_call`. Takes `n` (rows, default 20) and `minMs` (drop rows below this many ms per update, default 0.05).

# profiler.summary The engine's own one-line performance summary, exactly as it rides a tool response. The headline is the published smoothed frame time; when the frame is slow, or CPU moved since the last edit, the line expands with the severity marker, the `▲ from` delta and the systems the time went to. Takes no arguments. Returns nil before the engine has published its first frame. Read this when you want the figure other callers are reading. It is not the same measurement as the rest of this toolbox: `frame`, `hotspots` and `flamegraph` report the profiler's *effective* frame time, wall clock minus the cost of the call doing the asking, while this reports the frame the engine publishes to the agent surface, smoothed across frames. Composing your own line from the other tools produces a number that disagrees with what every other caller is told about the same engine. It exists for callers that compose their own context. A tool response normally carries this line, but it is deliberately omitted for a harness assembling its own prompt, since for that caller it is per-call overhead on every answer. That left those callers unable to read the frame at all, so a host appending its own engine-state block either went without or measured the frame a second way. Reading it does not move the baseline the delta is measured against. That baseline brackets exactly one agent edit; a reader that moved it would make the next edit's delta come out as zero, which drops the delta and the hotspot breakdown from the line together, so a real regression would stop announcing itself.

# profiler.tasks Per-component coroutine-resume attribution — the content view of the task scheduler. Where `scripts` shows where component `update(dt)` time goes, this shows where the scheduler's time goes, attributed to the component whose tasks (`task.spawn` / `task.wait` / awaited promises) it resumed. Rolled up per component type (many instances collapse to `Type xN`; a lone instance keeps its `@ entity`). Use it when `frame` / `hotspots` show `task_scheduler` hot and you need to know whose task work is behind it. Takes `n` (rows, default 20) and `minMs` (drop owners below this many ms of resume time, default 0.05).

# profiler.watch A frame-time watchdog: set a ceiling, run the scene, and the engine catches the bad frames for you. `watch(50)` records every frame whose EFFECTIVE time (the agent's own `execute` cost excluded) is at or above 50ms; `watch(50, "pause")` instead pauses gameplay the instant one crosses the ceiling, freezing the bad state for inspection (then read it with `retro`). `watch("off")` disarms, `watch("status")` reports state and hit count, and the `hits` tool reviews what record mode caught. It runs inside the engine's per-frame hook, so it is frame-accurate and immune to the very stalls it watches for — unlike an agent poll, which always lands seconds late. Editor profile only: a shipped runtime can't pause, so it is a no-op there.

# Toolbox (asset type) A toolbox is a namespace folder that groups related `.tool/` children plus shared helpers. Toolboxes are the addressing layer for tools — every tool's identity is `<toolbox>.<name>`, so the toolbox name appears in every callsite. Toolboxes also own a `shared.module/` helper module that the toolbox's tools `require(".shared")` for common utilities (result envelopes, validators, formatters). ## When to use one - You have two or more related tools that belong together (an `anim` toolbox with `play` / `stop` / `list`; an `assets` toolbox with `compose` / `validate`). - You need a shared helper module exclusively used by a tool set. - You want a clean dot-namespaced identity for your tools (`my_tools.foo`, `my_tools.bar`). If you only have one tool and no shared helpers, you still need a toolbox — it's the required parent. ## Where it lives - Source: `/zero/source/.../<Name>.toolbox/` - Identity: `<Name>` (the `.toolbox` suffix strips). - Folder shape (container-style — open child set): - `README.md` — instance documentation. **Required.** - `.metadata` — agent-editable tags + free-form fields. **Required.** - `shared.module/` — optional shared helper module (`init.luau` + `README.md`). - `<x>.tool/` — any number of `.tool/` children. The validator permits unlisted children (`allow_unlisted: true`) so tool subfolders don't need to be declared per-name. ## How to create one ```luau tools.createToolbox({ name = "<Name>", description = "..." }) -- or the generic scaffold: asset.create("toolbox", "<Name>") -- Creates: /zero/source/<Name>.toolbox/ -- README.md (instance README template) -- shared.module/ (ok / fail result-envelope constructors) -- `folder` places it in a subfolder of /source; `into` nests it in a -- resolved `.package`: asset.create("toolbox", "<Name>", { folder = "tools" }) ``` Then add tools under it: ```luau local box = asset.resolve("<Name>.toolbox") asset.create("tool", "<toolName>", { into = box }) ``` ## How it operates 1. **Registration.** Writing into a `.toolbox/` folder indexes the toolbox. It becomes the parent namespace for any `.tool/` children. 2. **Tool identity composition.** Each child `.tool/`'s identity is `<toolbox>.<tool>`. Listing the toolbox's tools enumerates the children. 3. **Shared helpers.** Tools inside the toolbox pull `shared.module/` in via `require(".shared")` (relative require from the tool's `init.luau`). 4. **Hot reload.** Editing `shared.module/init.luau` reloads every tool in the toolbox (they all `require` it). Editing the toolbox folder structure (adding / removing tools) reflects in `tools.list` immediately. ## Discovery - `asset.list("toolbox")` — every registered toolbox. - `asset.inspect("<name>")` — child tools, source, this type README. - `ls /zero/source/<Name>.toolbox` lists the files the folder holds. - `tools.list()` — every tool across every toolbox. ## Authoring conventions - **Per-tool implementation lives in each `<x>.tool/init.luau`.** The tool's full function body — argument validation, the call into engine APIs, the return value — goes in `init.luau` with `--!desc` / `--!arg` / `--!return` / `--!example` annotations above the function. **Every exported function is declared `typed function`** — the keyword compiles to the runtime argument check, so a plain `function` accepts any input and reports success on garbage. Do NOT make `init.luau` a one-line forwarder onto a function defined in `shared.module/`. - Keep `shared.module/` thin: result envelope constructors (`shared.ok(value, formatted)`, `shared.fail(message, exitCode)`), argument validators, common formatters — anything genuinely used by more than one tool. Per-tool logic does NOT belong here. - Group by domain, not by surface — `assets` toolbox holds asset operations, `entity` toolbox holds entity operations. Tools that feel unrelated probably want different toolboxes. - Document the toolbox's surface in `README.md` — what problem space it covers, who calls into it, any shared invariants. ## Common pitfalls - **Missing parent toolbox.** Tools cannot exist outside a toolbox. Always scaffold the toolbox first. - **Cross-toolbox shared.** Each toolbox has its own `shared.module/`. Don't try to share helpers across toolboxes; if you need cross-toolbox helpers, factor them into a standalone `.module`. - **Toolbox identity.** Strip the `.toolbox` suffix in identity references (`my_tools`, not `my_tools.toolbox`). ## Related types - `.tool` — the required child(ren) of a toolbox. - `.module` — for cross-toolbox utilities. ## In the Inspector Tools: how many tools the box holds, whether it ships a shared module, what each tool does, and a link to each tool.

# Module (asset type) A module is a reusable, stateless (or globally-stateful) Luau library that other code pulls in with `require()`. Modules are the unit of factored-out logic in a world: helpers, math libraries, format parsers, registries, shared algorithms. ## When to use one - You have logic you want to share across multiple components, services, scenes, or tools. - You have a registry, lookup table, or shared cache that callers all read from / write to. - You want a stable API surface (`M.fn`) other code can depend on. If the logic is bound to one entity, use a `.component`. If you want a named UI of fields rather than functions, use a `.preset`. ## Where it lives - Source: `/zero/source/.../<name>.module/` - Identity: the source path with `/` written as `.` and each segment's type suffix dropped — `/zero/source/demo/beats.module` is `demo.beats`, and `/zero/source/kit.package/beats.module` is `kit.beats`, the container's name kept and its `.package` suffix gone. `asset.create` returns this string, `asset.resolve(...).identity` reads it back, and the `require` forms below write it as `<identity>`. - Folder shape: - `init.luau` (or `init.lua`) — module body. **Required.** Returns a table (conventionally named `M`). - `README.md` — instance-level documentation. **Required.** - `.metadata` — agent-editable tags + free-form fields. **Required.** ## How to create one ```luau asset.create("module", "<name>") -- Creates: /zero/source/<name>.module/ -- init.luau (canonical `local M = {} … return M`) -- README.md (instance README template) -- `folder` places it in a subfolder of /source, and that subfolder -- leads the identity: `{ folder = "lib" }` produces `lib.<name>`. -- `into` authors it inside a resolved container (a package, a toolbox), -- whose own identity leads the same way: asset.create("module", "<name>", { folder = "lib" }) asset.create("module", "<name>", { into = asset.resolve("myPkg.package") }) ``` ## How it operates 1. **Registration.** Writing `init.luau` into a `.module/` folder indexes the asset and registers the module path with the resolver. 2. **Loading.** The first `require("<identity>")` runs the file's body and caches the returned table. Subsequent requires from anywhere return the **same table** — this is the shared-state lever for "module-local" registries. 3. **Hot reload.** Editing the file re-runs the module body and folds what it returns into the table already loaded, so that table keeps its identity: a caller that captured it with `require` reads the new source's functions and values through the reference it already holds, and a later `require` returns that same table. A key the new source drops goes away; a key something set at runtime stays. The body's own `local`s start again from their initial values, so state that must outlive an edit lives in `modules.state()` (`guides { path: "core/scripting-and-tasks" }`). 4. **Identity resolution.** A module is reached by its `<identity>`, and where the caller lives decides which spelling of it resolves. For a module in a **world**, from a caller in that same world — another module, a component, a scene entrypoint, or an `execute` chunk: - `require("<identity>")` — the identity on its own, resolved against the caller's own root. It carries every folder segment, so a module created with `{ folder = "demo" }` is `require("demo.beats")`. - `require("@root::<identity>")` — the same root-relative resolution, written out. - `require(".sibling")` — a module in the same folder, by its bare name; `..name` steps up one folder, `...name` two. Logs, stack traces and hot-reload notices name that same module `@local.source.<identity>` — its key under the world source root. The `<identity>` tail of that key is the form above, so a trace reading `@local.source.demo.beats` is `require("demo.beats")`. For a module in a **library** — content under `/zero/source/libs/@<lib>/` — `@<lib>::` addresses that library's root from anywhere: `require("@builtin::modules.transform")`. Mounting a world as a library moves its files under such a root and flips the caller root to `@<lib>::`, which is what keeps the root-relative forms above resolving across the move. ## Discovery - `asset.list("module")` — every registered module. - `asset.inspect("<name>")` — public functions, source path, this type README. - `ls /zero/source/<name>.module` lists the files the folder holds. ## Authoring conventions - Return a single `M` table from `init.luau`. Top-level statements with side effects run on first require — useful for one-time initialization, dangerous if they touch the engine before it's ready. - Write exported functions as `typed function`, not `function`: the argument types are enforced at the call and the checker knows the signature, so a caller's mistake is reported where they made it. - Annotate exported functions with `--!desc` / `--!arg` / `--!return` / `--!example` so the LSP, `tools.list`, and `asset.inspect` surface them. A function with no `--!desc` is reachable only by someone already reading this file. - Declare a VALUE with `Field.<kind>(default, mode, description)`, and a table whose members are reached through a metatable with `--!members <TypeName>` — the latter gives the surface its accepted member set, its type, and its documentation from one table. See `guides { path: "core/authoring" }` for all four rules and `modules/api/engine/entity.module/members.module` for a worked member table. - Keep module state in module-local upvalues or in `modules.state()`. Globals leak across reloads. Module-locals start again from their initial values with each hot reload; the table `modules.state()` returns is the same table on both sides of every reload. - Prefer focused modules over kitchen-sink modules. If two halves of a module have no shared state, split them. ## Common pitfalls - **Cyclic requires.** A `require` chain that loops will return the partial table (the half built before the cycle was detected). Design for one-way dependencies. - **Engine-time side effects.** Don't `entity.spawn` at module top-level — the engine may not be ready. Expose an `M.init` and call it from a scene entrypoint. - **Hot-reload + module-local cache.** A module that caches expensive computation in upvalues loses that cache on reload — fine for dev, but be aware in performance work. - **`init.lua` vs `init.luau`.** Either works, `.luau` is canonical. - **Writing one while play runs.** A `/source` write made while play is running lands on the play shadow, and a guarded play-exit discards it unless it is kept. One `.module` write shadows `init.luau`, the `README.md` and the `.metadata` together. `vfs.write(path, bytes, { durable = true })` skips the shadow, and `vfs.promotePlayShadow` keeps what is already on it — the `core/vfs` guide has the model and what each route costs. ## Related types - `.component` — for entity-bound state + lifecycle hooks. - `.service` — generates content you don't have yet (a mesh, sound, texture, …) via a metered provider. - `.tool` — for a single agent-callable function with a YAML schema. - `.package` — to group several modules + components + scenes into one shippable folder. ## In the Inspector Exports: what the module exports (name, kind, signature), parsed from its source without running it, with **Check** (the LSP's diagnostics for its entry file) and **Open**.

# action Perform a gameplay action by the name of the control the game reads it from — tap `jump`, walk with `move`, hold `sprint` — with no device in between.

# bindings Every control that is live right now, and which map contributed it.

# click A full mouse click, optionally moving the cursor first.

# device Read the active device class, or force one so a desktop session can be checked as a phone.

# drag Drag a finger across the look zone.

# fired Which controls fired since the last call — the assertion half of a simulated input.

# held What the session is holding right now — keys, mouse buttons, fingers, the on-screen stick, and connected pads. `release` lets go of exactly this set.

# key Tap a key — down, optional hold, then up.

# keyDown Press a key and leave it held until `keyUp`.

# keyUp Release a key held by `keyDown`.

# layout What the on-screen touch controls are drawing right now.

# sim.listMacros List every registered input-macro asset in the project, returning an array of asset handles ready to hand straight to `sim.macro` for playback. Use it to discover which recorded or authored macros are available.

# lockPointer Lock or release the pointer.

# inputSim.macro Play back a macro of timed input events, driving the keyboard, the mouse, a gamepad and the glass in one sequence. Accepts an inline event list, an input-macro asset handle, or the identity or path of a saved input-macro asset; playback runs as a background task that waits between events according to each event's timestamp. Each event is `{ t, op, arg }`, where `t` is seconds since the macro started (events play in order of `t`, and events sharing one `t` play in the order the list writes them), `op` names one of this toolbox's own verbs that drives input, and `arg` carries that verb's arguments: the value itself when the verb takes one, a table keyed by the verb's parameter names when it takes several. So the keyboard and mouse verbs (`keyDown`, `keyUp`, `key`, `mouseDown`, `mouseUp`, `mouseMove`, `mouseMoveBy`, `click`, `scroll`, `lockPointer`), the pad verbs (`pad`, `padOff`, `padDown`, `padUp`, `padPress`, `padStick`, `padTrigger`) and the touch verbs (`touch`, `touchDown`, `touchMove`, `touchUp`, `stick`, `tapButton`, `drag`, `pinch`) all play, and a point written for a verb is the point the op of that name takes. A `touchDown` event's `id` is the label its contact goes under for the run of the macro, and the `touchMove` and `touchUp` events carrying that label drive it. The same primitive drives a player, a car, or a UI test; author macros by hand or record them with `inputSim.recordMacro` and persist them with `inputSim.saveMacro`.

# mouseDown Press a mouse button and leave it held.

# mouseMove Move the cursor to an absolute screen position.

# mouseMoveBy Move the pointer by a relative motion.

# mouseUp Release a mouse button held by `mouseDown`.

# pad Connect a simulated gamepad and return its slot.

# padDown Hold a canonical pad button down.

# padOff Disconnect a simulated gamepad.

# padPress Press a canonical pad button, named by position.

# padStick Push a pad thumbstick to a deflection.

# padTrigger Squeeze a pad trigger to an analog travel.

# padUp Release a held pad button.

# pinch Pinch or spread two fingers.

# inputSim.recordMacro Record live input over a window of time on all three device classes, emitting a timed event whenever something changes. It reads what every device is holding each frame (keys, mouse buttons, pad buttons, the pad itself, fingers on the glass) and what each one is measuring every `sampleInterval` seconds (mouse motion, the wheel, thumbsticks, triggers, a sliding contact). It returns the captured events array plus the task handle so the caller can cancel early, and stops on its own after `duration` seconds elapse or when the optional `until_` predicate answers true. It takes those arguments the way every verb in this toolbox does, positionally, how long first, so `recordMacro(3.0)` records for three seconds and `recordMacro(3.0, 0.02)` samples that window more finely. The result works for any input-driven scenario, a player walk, a car drive, a pad rhythm or a UI flow, and can be replayed immediately with `inputSim.macro` or persisted with `inputSim.saveMacro`.

# release Put everything down between takes: fingers, the stick, held keys and mouse buttons, and connected pads.

# sim.saveMacro Persist a recorded or hand-authored event list as a named input-macro asset in the world's source, returning the asset handle of the saved macro for immediate playback. The name becomes the asset's folder name and must be non-empty. Use it to turn a `sim.recordMacro` capture into a reusable, replayable macro.

# scroll Turn the mouse wheel.

# inputSim.shared The point conversion the toolbox's absolute-point pointer verbs share. `click` and `mouseMove` take a screen point in physical pixels, in the space the input's receiver reads positions in. The input primitives beneath them place a pointer on the window. `windowPoint(x, y, target)` carries a point into that window space. For the game (`target` `"game"` or nil) the point is in the space `capture` reports, the scene's own pixels measured from the top-left of the scene's rect, and the conversion adds that rect's origin; where the scene is drawn across the whole window (a headless engine, a fullscreen play session) the rect starts at the window's top-left and the conversion is the identity. For the UI (`target` `"ui"`) the point is already on the window, where a widget sits at its `gui.bounds` rect times `ui.pixelRatio()`, and passes through unchanged.

# stick Push the on-screen movement stick to a deflection.

# tapButton Tap an on-screen touch button by the label a player sees.

# touch A whole finger tap: contact, optional hold, lift.

# touchDown Begin a finger contact and leave it down.

# touchMove Move a finger that is already down.

# touchUp Lift a finger.

# diagnostics.errors Report component and entity errors in the current scene. With no argument it returns errors for every entity; given an entity name or id it returns errors for just that entity. Read-only — it inspects state without changing anything.

# shared Module Cross-tool helpers for the `diagnostics` toolbox. Every tool inside this toolbox can pull these in with `require(".shared")`. ## Purpose This module holds the small set of helpers that more than one tool in the toolbox needs — typically the `ok(value, formatted?)` / `fail(message, exitCode?)` constructors that produce the canonical `diagnosticsToolResult` shape, plus any toolbox-wide parsers, validators, or shared state. Keep it focused: per-tool implementation stays inside each `<x>.tool/init.luau` (with its own `--!desc` / `--!arg` / `--!return` / `--!example` annotations). ## Usage ```luau local shared = require(".shared") function M.someTool(...) if not ok then return shared.fail("reason", 1) end return shared.ok(value, "") end ``` ## Exports See `init.luau` for the full export surface. Typical exports include `shared.ok(value, formatted?)` and `shared.fail(message, exitCode?)` for building the toolbox's standard result envelope.

# diagnostics.summary Summarize the current scene: entity count, a per-type breakdown of components, the number of entities carrying errors, camera count, and the active culling mode with its stats when available. A quick one-call health snapshot of what the scene contains. Read-only — it inspects state without changing anything.

# diagnostics.validate Validate scene integrity by surfacing broken references — materials that don't exist, missing model assets, and entities carrying errors. It walks every entity's error messages and extracts the named materials and meshes that failed to resolve, then reports whether the scene is clean. Read-only — it inspects state without changing anything.

# capture.cleanup Clears every file from the capture output directory, removing the PNGs left behind by earlier captures. Safe to call when the directory is missing or already empty.

# collage A grid of frames in one image. One picture that answers a question no single frame can: what does this look like from every side, what does this frame look like under each diagnostic pass, what moved over the last three seconds, which of these twenty shots is the shot. ## The axis is always stated A collage varies one thing across its cells, and the argument you pass is what that thing is: | Argument | One cell per | |----------|--------------| | `setups` | whole camera set-up — a shot list, each entry aimed the way a single capture is | | `viewpoints` | named view of one subject — a list, `"front top left"`, or `"sides"` for all six | | `passes` | render pass — `final`, `albedo`, `normal`, `depth`, `motion_vectors`, or any content-registered capture view | | `duration` | sample, spread across that many seconds of gameplay | `setups` and `viewpoints` are one axis — both say where the camera stands — so a request names one of them. Naming a station axis **and** `passes` lays out a 2D grid: a row per station, a column per pass. Reading across a row is one set-up under every pass; reading down a column is one pass from every set-up. A collage with no axis is refused, naming all of them. It does not fall back to time — a silent time default is exactly what taught callers that a collage could only ever be time, so nobody looked for a grid of views. `duration` cannot combine with `setups`, `viewpoints` or `passes`. A cell that differs both in what it looks at and in when it was taken answers neither question; take one collage per axis. ## The shot list Each entry of `setups` is written the way a single capture is aimed — `source`, `camera`, `entity`/`entities`, `position`, `lookAt`, `rotation`, `distance`, `angle`, `margin`, `viewpoint`, `basis`, `projection`, `orthoHeight`, `fov`, `near`, `far` — plus `label` to name the shot on the result. A field an entry leaves out comes from the collage's own options, so a shot list writes down only what makes each shot differ: ```lua zero capture collage --setups '[ { "position": [-14.6, 1.35, 4.5], "lookAt": [-16.3, 1.1, 1.6], "fov": 28, "label": "doorway" }, { "position": [8, 3, 12], "lookAt": [0, 1, 0], "fov": 50, "label": "wide" }, { "entity": "hero", "viewpoint": "front", "basis": "local", "label": "closeup" } ]' ``` Three stations, three lenses, three subjects, one instant, one picture — which is the comparison a framing pass is made of. A key outside that vocabulary is refused where it was written, so a misspelled `fov` names itself instead of quietly framing a shot nobody asked for. ## Reading the result The grid shape is worked out from the number of cells — pass `grid` (`"3x2"`) to state it, and it is an error rather than a silent crop if it cannot hold them. Every cell comes back labelled in `cells`, with its `label`, the `viewpoint` and/or `pass` that made it differ, the `eye` it was rendered from and the lens it was rendered through. A grid is meant to be readable from the response, not guessable from the pictures. Only a time collage reports the window it covered — a station or pass grid samples one instant, and reporting a span for it would invite reading the difference between cells as motion. ## Aiming it Exactly as a single capture is aimed: `entity` (with `viewpoint` and `basis`, or `angle` and `distance`), `position` with `lookAt`, `camera`, or nothing at all for the main scene camera. `projection`, `orthoHeight`, `isolate`, `renderLayers`, `fov`, `width` and `height` all mean what they mean for one frame, and apply to every cell a set-up of its own does not override. The shot most of this exists for is one call: ```lua zero capture collage --entity crate --viewpoints sides --basis local \ --projection orthographic --isolate ``` Six sides of one object, each an orthographic elevation, nothing else in frame.

# feed The newest frame a capture produced on this engine, and who is watching for it. Every capture core publishes here. `latest()` is that frame: a `src` an image widget draws (the render target's guid while the engine still holds it, the written path when the capture produced a file), with the size, the source and the pass beside it. `watch(fn)` calls `fn` as each new frame arrives, and `unwatch(token)` drops the subscription. `shown()` says a viewer is drawing frames from the feed. From then on the feed holds the newest capture's render target so the picture stays up after the capture that made it is finished with it, and lets it go when a later frame replaces it; `shown(false)` stops. With nobody drawing, a capture's target is released the moment its capture is done. The release path asks `retains(rtHandle)` before destroying a target, which is how a held frame survives its own capture. `capture.shared.releaseFrame` is the one place that asks.

# capture.fromCamera Render the scene from a specific named camera's authored pose and settings, returning the picture as an image value (and writing a PNG as well when a `path` is given). By default it uses the camera's exact world-space pose plus its whole lens (fov/near/far, and the projection it builds with the orthoHeight an orthographic camera covers), its diagnostic debugChannel, its post-process setting, its lighting environment (an isolated camera's background and ambient with it) and the scene layers it draws: the camera's own `renderLayers` with the editor's chrome (`EditorUI`) and the authoring overlays (`debug`) taken off, since those are drawn for whoever is editing rather than by that camera into the world. So the frame is precisely what that camera draws of the scene, without changing what is on screen. Pass overrides (`renderLayers`, `debugChannel`, `fov`, `near`, `far`, `projection`, `orthoHeight`, `width`, `height`) to change any of those for the one capture without touching the camera; `renderLayers = "all"` is how the overlays are captured on purpose.

# capture.fromEntity Renders a screenshot orbiting a target entity. The camera position is computed from a distance and a yaw/pitch angle relative to the entity's world position, with an optional offset applied before framing and a configurable field of view and output size. Returns the picture as an image value, and writes a PNG as well when a `path` is given; fails when the entity is missing or its position can't be read. Runs as a background task that waits for the GPU readback, so it works in play mode.

# capture.fromPosition Renders a screenshot from an explicit world-space camera position and returns the picture as an image value. Aim it with a look-at target (defaulting to the origin) or a quaternion rotation that takes precedence, and set field of view, output size, and near/far clip planes. A PNG is written as well when a `path` is given. It spawns an ephemeral camera, triggers one render, reads the frame back, and despawns the camera automatically — running as a background task that waits for the readback, so it works in play mode.

# isolate Show the subject without the other geometry drawn, then put everything back. A prop inside a built scene is behind whatever stands in front of it. The render-layer system already expresses "draw only these" — an entity is a member of layers, a camera renders a filter over them — but it does not remember what it displaced, so every caller wanting one clean shot of one thing hand-rolls the same record, swap and restore. This is that, once, with the restore guaranteed. ## What it changes Exactly one thing: the other geometry is not drawn. Sky, post-process and UI are left exactly as the capture's own render-layer spec stated them, so a final frame stays a final frame. A caller who wants the geometry read flat — no lighting, no atmosphere — asks for a **pass** (`pass = "albedo"`, `pass = "normal"`), which is what passes are for. ## What it does not change Excluded geometry still **casts shadows** onto the subject and still bounces light into it. Lights are not filtered by render layer at all, and the layer mask drives object culling and the geometry passes rather than the shadow passes. This is the render-layer system's deliberate behaviour — a wall stays out of the shot while it still exists for lighting, shadows and physics — so a shadow with no visible caster in an isolated capture is the system working, not a bug. ## The layer One reserved layer name, `captureIsolate`, reused by every isolated capture. Membership is a 32-bit mask with six reserved bits. A layer minted per call would exhaust the namespace in twenty-six captures and leave a trail of empty layers that `renderLayer.list` would show forever. ## Restoring `begin` records each affected entity's membership before moving anything, so the token describes the whole subtree even if the move fails part-way. `restore` puts every entity back to the layers it carried — and an entity that carried **no** explicit membership ends with none, rather than an explicit `default` that looks identical in a picture and different in the data. `restore` is safe to call twice and safe on a token whose subjects have since despawned, which is what lets a caller restore unconditionally on every exit path: success, failure, and once around a whole collage grid rather than between its cells.

# capture.oneshot The low-level one-shot capture primitive. Spawns an ephemeral offscreen camera configured from a source — the viewport, an entity (or several), a world-space position, or a UI window — triggers a single render, and returns the camera and render-target handles synchronously so the caller drives the readback and cleanup itself. This synchronous shape is the only capture path that works from edit mode, where the background-task scheduler the convenience wrappers rely on does not tick. When framing an entity, omitting the distance auto-fits the camera to the union world-space bounds of the target(s), so the subject is guaranteed fully in frame at any scale — a small prop and a large building both fill the shot. Because the data drives the camera, an empty result is a real signal that the geometry or material is wrong rather than a framing mistake; a distance can be passed to force a fixed orbit radius, and an angle sets only the viewing direction. Further options select the debug visualization channel, toggle which layers (scene, sky, post-process, UI) render, and set field of view, clip planes, and output size. The returned camera is left disabled so it can be re-triggered to produce additional frames on the same target, as when stitching a collage.

# shared Module Cross-tool helpers for the `capture` toolbox. Every tool inside this toolbox can pull these in with `require(".shared")`. ## Purpose This module holds the small set of helpers that more than one tool in the toolbox needs — typically the `ok(value, formatted?)` / `fail(message, exitCode?)` constructors that produce the canonical `captureToolResult` shape, plus any toolbox-wide parsers, validators, or shared state. Keep it focused: per-tool implementation stays inside each `<x>.tool/init.luau` (with its own `--!desc` / `--!arg` / `--!return` / `--!example` annotations). ## Usage ```luau local shared = require(".shared") function M.someTool(...) if not ok then return shared.fail("reason", 1) end return shared.ok(value, "") end ``` ## Exports See `init.luau` for the full export surface. Typical exports include `shared.ok(value, formatted?)` and `shared.fail(message, exitCode?)` for building the toolbox's standard result envelope.

# subject What a capture of an entity is a picture of. Every path that frames or isolates an entity (`fromEntity`, `oneshot`, a `setups` or `viewpoints` collage, `isolate`) resolves the name, id or proxy it was given here, so they agree on which entity a name means. ## A player is drawn as its body A player identity (an entity carrying `UserIdentity`) is the registry anchor a connected user is known by. The player is drawn as the body its `avatar` names, a standalone entity at the scene root, so a capture of an identity frames and isolates that body. In play, a player prototype's spawn turns the prototype's copy into the joining user's identity under the prototype's own name and stands the body it names at the spawn point. A capture of the name an author gave their player prototype therefore frames the player's body in play, and the prototype's subtree in edit. ## What a refusal names `describe(target, via)` is the phrase a framing refusal uses for each subject it looked at: the entity's name and id with the number of descendants it holds, and the player identity it is the body of when it was reached through one, or, for a player identity with no live body bound, that the body is missing. | Function | Returns | |----------|---------| | `resolve(t)` | the entity a capture pictures for a name, id or proxy (nil when none matches), and the player identity it was reached through | | `bodyOf(target)` | the live body a player identity is drawn as, or nil | | `isPlayerIdentity(target)` | whether the entity carries `UserIdentity` | | `describe(target, via)` | the refusal phrase for one resolved subject |

# viewpoints The named camera stations a capture can be taken from, and the rule that decides what a name like `front` means. A **viewpoint** is a direction with an image-up: where the camera stands relative to what it is looking at, and which way is up in the resulting picture. Naming one replaces reverse-engineering a `{yaw, pitch}` pair, and names the seven stations that answer "is this built correctly" — the six sides and the corner. | Viewpoint | Sees | |-----------|------| | `front` | the side the subject faces | | `back` | the opposite side | | `right` | the subject's right | | `left` | the subject's left | | `top` | plan view, the subject's front toward the top of the image | | `bottom` | the underside | | `iso` | the corner view where all three axes project equally | An entity faces its local -Z, so the camera that sees its front sits on its +Z. `top` and `bottom` take their image-up from the subject's forward axis, because world up is undefined when you are looking straight down it — that is what puts the subject's front at the top of a plan view instead of leaving the roll to whatever the yaw happened to be. ## Basis `basis` says which axes the name is measured against, and it has **no default**: - **`"local"`** — the subject's own axes. `front` is the side the subject faces, whichever way it is turned, and the frame is fitted to the subject's own extents (`entity:orientedBounds()`). - **`"world"`** — the world axes. `front` is the side facing world +Z, whatever the subject is doing. For a crate turned 40 degrees these are different pictures, so the caller states which one it wants. A default would mean one call site's `front` is the crate's front and another's is the world's, and the difference only shows up in the image. With no subject to take axes from — a capture framed on a world position rather than an entity — both values mean the world axes, and both are accepted. ## Orthographic size `orthoHeightFor` answers the question an orthographic frame asks that a perspective one does not: how big is the frame? A perspective camera backs off until the subject fits; an orthographic camera does not converge, so its size is decided by the subject alone — project the bounds' eight corners onto the camera's up and right axes and take whichever needs more room. Its `half` extents and `axes` must be stated in the same frame. A resolved viewpoint carries both: `frame` axes to use with local-frame extents, and its world axes to use with world extents.

# capture.viewport Renders a screenshot mirroring the active viewport. It mirrors the on-screen camera through the shared capture core, so the frame carries that camera's pose, its lens (`fov`, clip planes and the projection it builds), the lighting environment it states (an `isolated` camera's background and ambient) and the layers it draws the screen with, the editor's own overlays included, so the image matches what's on screen. The raster is the viewport's own pixel size, the size `getViewportSize()` reports. Returns the picture as an image value, and writes a PNG as well when a `path` is given; fails when no active viewport camera exists. Its one argument is a path to also write the PNG to, or a table overriding any of that for one capture: `width` / `height` for a raster of some other size, `renderLayers` for the frame under a different layer spec — `"all !EditorUI !debug"` is the frame without the editor's chrome over it — and `postProcessing` to take it with the grade off. ```lua tools.use("capture", "viewport", "/source/tmp/capture/shot.png") tools.use("capture", "viewport", { path = "/source/tmp/capture/clean.png", renderLayers = "all !EditorUI !debug" }) tools.use("capture", "viewport", { width = 960, height = 540 }) ```

# debug.inspect Inspect entity component state — script components, native ECS components, or both — for every entity matching the target. ```lua debug.inspect("player") debug.inspect("lamp", { include = "ecs" }) debug.inspect("enemy_*", { types = { "Health" }, limit = 5 }) ``` The target is an entity id, a name, an entity proxy, or an array of them. A name resolves to **every** matching entity (names are not unique; globs like `enemy_*` work), an id to itself. Options: ```lua { include = "all", -- "script" | "ecs" | "all" (default) types = nil, -- restrict to these component types; full identity -- or leaf name ("Camera" matches -- "@builtin::components.Camera") limit = 10, -- cap on matches } ``` Returns: ```lua { query = { target, count }, matches = { { id = "ent_...", name = string?, script = { [componentType] = data }, -- serialized public data ecs = { [componentName] = fields }, -- native component fields }, ... }, } ``` `script` data is the serialized form (entity refs as id strings, asset refs as `{ __ref, name, type }` envelopes) produced by the `component_snapshot` module; `ecs` is `ecs.snapshot` per entity. For live gameplay reads and writes, use `entity(id).component.get` / `getAll` — the data those return is the running component's proxy, not a snapshot.

# debug.playSession Show or hide the debug overlay during play mode. ```lua debug.playSession(enabled?) ``` - `enabled` — `true` (default) to show the overlay in play, `false` to hide it. Returns `{ playSession = boolean }`. Edit mode always shows the overlay; play mode shows it only while this session is on — the way to watch bones, colliders, or bounds on a running, animating scene (bones only move once the scene is unpaused). The flag lives in memory only and is never saved, so debug visibility can never be permanently toggled on by accident; it resets on engine restart.

# debug.problems Surface the recent errors and warnings the caller has not read yet. ```lua debug.problems() ``` Non-crashing failures — a contract violation, a component `awake`/`update` throw, a component auto-disabled after erroring, an unresolved require, a bridged renderer or asset error — land in the engine's log ring but do not surface in `execute` results unless you happened to ask for them at the right level in the call where they occurred. `debug.problems` reads them back out. **When something you built does not behave, call this first.** The engine has usually already logged the reason. Each call advances a read-cursor and returns the problems logged since the previous call, **oldest first** — so the root-cause error (the `awake()` throw that disabled a component) reads ahead of the cascade it triggered. Repeated calls report only what is new. ```lua debug.problems({ level = "warn", -- minimum severity: "warn" (default) or "error" limit = 50, -- cap on entries returned (default 50) all = false, -- true: recent problems regardless of the cursor, cursor untouched peek = false, -- true: read the next batch without advancing the cursor }) ``` Returns: ```lua { problems = { -- oldest first { seq, level, type, message, timestamp }, ... }, shown = number, -- #problems truncated = boolean, -- more unread problems remain past `limit` cursor = number, -- read-cursor after this call counts = { errors, warnings }, -- lifetime totals (survive ring eviction) } ``` `all = true` gives a one-shot overview of the most recent problems without disturbing the drain cursor. `peek = true` previews the next batch and leaves it unread for the next call. The full log ring — every level, searchable, countable, filterable by the script or entity that logged a line — is the `logs` toolbox, whose `search`, `tail`, `errors` and `summary` tools read it. Reach for those when you need more than the unread problems: a specific message, a specific entity, a specific minute, or the lines that led up to a failure. The same ring is also readable directly from Luau through the `logs` global, and as plain files at `/zero/runtime/logs/engine`, `/zero/runtime/logs/errors` and `/zero/runtime/logs/script` when a grep for a message you can name is shorter than a query. This tool is the read-cursor view over the warn/error slice of it.

# debug.scope Focus the debug overlay on specific entities, or clear the focus. ```lua debug.scope(target?) ``` - `target` — a name or `ent_` id, an array of them, or the `{ id = ... }` records `scene.findAll` returns. Names resolve to every matching entity. Omit (or pass an empty list) to clear the focus and draw for every entity again. Returns `{ scoped = { ids }, count = number }`, the scope after the change. Every enabled category honours the scope, so the overlay focuses on the chosen objects — hierarchy bounds, for instance, draw one box per scoped part instead of one union box per top-level object. The scope lives in memory only and is never saved.

# debug.set Show or hide a debug-visualization overlay category. ```lua debug.set(category, enabled?) ``` - `category` — a category name (`bounds`, `frustum`, `bones`, `colliders`, `icons`, `wireframe`), or `"all"` for every category. - `enabled` — `true` (default) to show the category, `false` to hide it. Returns `{ [category] = boolean }`, the category state after the change (every category when `"all"` is passed). Errors on an unknown category name, listing the known ones. The overlay is ambient editor tooling drawn for the whole scene; this only flips which categories draw. `bounds`, `bones`, `colliders`, and `wireframe` are off by default (heavier per-mesh overlays); `frustum` and `icons` are on.

# shared Module Cross-tool helpers for the `debug` toolbox. Every tool inside this toolbox can pull these in with `require(".shared")`. ## Purpose This module holds the small set of helpers that more than one tool in the toolbox needs — typically the `ok(value, formatted?)` / `fail(message, exitCode?)` constructors that produce the canonical `debugToolResult` shape, plus any toolbox-wide parsers, validators, or shared state. Keep it focused: per-tool implementation stays inside each `<x>.tool/init.luau` (with its own `--!desc` / `--!arg` / `--!return` / `--!example` annotations). ## Usage ```luau local shared = require(".shared") function M.someTool(...) if not ok then return shared.fail("reason", 1) end return shared.ok(value, "") end ``` ## Exports See `init.luau` for the full export surface. Typical exports include `shared.ok(value, formatted?)` and `shared.fail(message, exitCode?)` for building the toolbox's standard result envelope.

# debug.state Report the current debug-visualization state. ```lua debug.state() ``` Returns: ```lua { categories = { [name] = boolean }, -- which categories draw scope = { ids }, -- focused entity ids (empty = all) count = number, -- #scope playSession = boolean, -- play-mode overlay session on? mode = string, -- engine mode ("edit" / "play") } ``` A read-only snapshot of what `debug.set`, `debug.scope`, and `debug.playSession` have configured — the state the ambient overlay draws from this frame.

# worldValidation.check Validates a world and every library imported into it in one pass — the cargo-check equivalent for authored content, returning every structural problem in world code and library code together. It is read-only: it walks the source once, checks each asset folder against its type schema and each script against the textual checks, and reports the intentional placeholder checks flagged so they can be filtered out without hiding real problems. Filter options narrow the report by severity, category, code, source, or path; each pass returns the full report alongside a one-line summary.

# worldValidation.checkLibraries Validates every imported library, skipping the world's own content, so a world's dependencies can be audited as a group without world noise. Each library gets its own entry in the report. Takes the same filter options as the full check and returns the report with a one-line summary.

# worldValidation.checkLibrary Validates one named library in isolation, by its directory name, so a single dependency's health can be inspected independent of the rest. If the named library doesn't exist the report carries a single missing-library error. Takes the same filter options as the full check and returns the report with a one-line summary.

# worldValidation.checkWorld Validates only the world's own authored content, excluding every imported library, so you can tell whether your code is clean independent of any library's health. It behaves like the full check with the scope narrowed to the world. Takes the same filter options and returns the report with a one-line summary.

# worldValidation.report Generates a validation report with structured filtering and formatting behind a single options table, so one call can scope the pass, filter the problems, choose a render format, and optionally write the result to a file. Scope selects the world, the libraries, all of it, or one named library; severity, category, code, source, and path narrow which problems are kept, and a limit caps the count and marks the result truncated. Format renders the report as human text, markdown, JSON, or just the summary, and a save path writes the rendered report to disk.

# worldValidation.summary Runs a full validation pass and returns only the one-line health summary and a pass/fail flag — a quick health gate for scripts or PR templates. Takes the same filter options as the full check, most usefully the switch that ignores the placeholder-check noise.

# worldValidation Cargo-check-style validation for a Zero world. Walks every script and every asset under `/source/` and reports problems in one structured report. The world and its imported libraries are validated as cleanly separated buckets so authoring problems in user content can be spotted without being drowned out by issues in `@builtin` or any other installed library — and vice versa. The system is pure Luau on top of `vfs.*` and `asset.*`. It does not spawn entities, allocate GPU resources, or mutate world state — the report is read-only. ## Two-bucket separation ``` /source/ ← user-authored content (this world) foo.component/ main.scene/ libs/ @builtin/ ← imported library: @builtin ... @mylib/ ← imported library: @mylib ... ``` Everything under `/source/libs/<name>/` is treated as belonging to the library `<name>`. Everything else under `/source/` is world content. Reports always carry the bucket split so a quick `report.world.ok` check tells you whether YOUR code is clean independent of library health. ## Contents - `validator.module/` — orchestrator, public API entry point (`check`, `validateWorld`, `validateLibraries`, `validateLibrary`, `filter`, `format`, `summary`, `saveReport`, `placeholders`). - `scriptValidator.module/` — per-file Luau script validation. Includes a placeholder syntax-check hook for the future when a runtime parser is available. - `assetValidator.module/` — asset-folder validation. Delegates to `asset.validate`, the engine primitive the `world.push` gate runs, and maps its problems into the report — structure, content constraints, semantic hooks, and the private-subasset exemption all behave exactly as they do at publish time. - `vfsScanner.module/` — recursive VFS scanner that enumerates scripts and asset roots under any path. - `reportFormatter.module/` — filtering + formatting helpers (`human`, `markdown`, `json`, `summary`). - `tools/worldValidation.toolbox/` — agent-callable surface: - `worldValidation.check` — full validation pass. - `worldValidation.checkWorld` — world bucket only. - `worldValidation.checkLibraries` — every imported library. - `worldValidation.checkLibrary` — one library by name. - `worldValidation.report` — generate a report with filtering / formatting options; can write to a file. - `worldValidation.summary` — one-line health summary. ## Portability The package's library modules use the `~` package-relative require form — drop the package under another library, rename the parent directory, or install it into a world under a different identity and every internal lookup keeps resolving. The package is also self-contained: no file outside `worldValidation.package/` is needed for the validator to run. The toolbox at `tools/worldValidation.toolbox/` is the one exception: code-mode tools are mounted under the synthetic `@<lib>::tools.<ns>` namespace, which has no `.package` ancestor in its chunkname, so `~` can't walk back to a package alias from a tool source. Tools therefore use the absolute `@builtin::systems.worldValidation.validator` form to reach the validator module, matching the pattern other package-shipped toolboxes (`characterController`, `volumeProbe`) already follow. Asset validation goes through `asset.validate`, which resolves each asset's registered type (built-in or world-authored) and its `type.yaml` from the live registry — the validator hardcodes no VFS paths. ## Quick start ```luau local V = require("@builtin::~worldValidation.validator") -- Full validation local report = V.check() if report.ok then print("clean") else print(V.format(report, "human")) end -- World only, errors only, saved as markdown local r = V.validateWorld({ severity = "error" }) V.saveReport(r, "/source/.validation/report.md", { format = "markdown" }) ``` From an agent / tools: ```luau tools.call("worldValidation.check", {}) tools.call("worldValidation.report", { scope = "world", severity = "error", format = "markdown", savePath = "/source/.validation/report.md", }) tools.call("worldValidation.summary", {}) ``` ## Problem schema Every issue surfaced by the validator follows the same shape: ```luau { severity = "error" | "warning" | "info", category = "structure" | "syntax" | "semantic" | "reference" | "schema" | "placeholder", code = "<short.code>", -- e.g. "structure.missing_required_file" path = "<vfs path>", -- where the problem lives message = "<details>", -- one-line human reason source = "world" | "library:<name>", placeholder = boolean, -- true if produced by a placeholder check } ``` ## What's actually validated | Layer | Implementation | Covers | |---|---|---| | Luau code | `lsp.check(path)` per script — the engine's embedded LSP | parser errors, type errors, every registered lint, including diagnostics inside asset folders (component / module / scene init.luau, etc.). | | Asset folders | `asset.validate` per asset — the same primitive the `world.push` gate runs | missing required files / one_of_group alternatives (`error`), content constraints like an empty README or tag-less `.metadata` (`error`), unexpected entries when `allow_unlisted: false` (`warning` — the publish gate doesn't block on them), the type's own semantic `validate` hook, unknown/unresolved types (`warning`). Private subassets (a bundle's interior meshes / materials) are exempt from the describe-this-asset requirements, exactly as at publish time. | ## Notes - The validator is read-only. It never writes to `/source/` except when `saveReport` is invoked with an explicit destination path. - Validation is single-frame, synchronous — no `await` boundaries. - Hot reload: editing this package's modules reloads them on next require. Running validation again picks up the new logic.

# guide A guide is a single-file documentation asset: a `<name>.guide/` folder whose body is `guide.md`, surfaced through the `guides` MCP tool, the `man` shell builtin and the Docs panel. Drop one anywhere under `/zero/source/`, a `.package` included, and it shows beside the built-in guides. ## In the Inspector Guide: the guide's word count, its headings, and **Open in Docs**, which opens it in the Docs panel.

# wld.edit Return the engine to edit mode. Restores the edit state that was auto-snapshotted on the previous transition into play, and stops script and physics execution so the scene sits still for authoring. This is the counterpart to entering play — the round trip lets you run a scene, then rewind to exactly the state it had before it started.

# wld.hideLayer Hide a loaded scene layer by name. The layer's entities stay in the world and keep their state — only their rendering stops. Showing the layer again makes them visible without a reload, while unloading despawns them entirely. Useful for toggling optional overlays such as a UI or debug layer on and off cheaply.

# wld.info Read the current world's high-level state in one call: the bound world name, whether it is in edit or play mode, the active scene layer, and the list of saved scene snapshots in the live world. A quick orientation snapshot for deciding what to do next without querying each piece separately.

# wld.listWorlds List the worlds persisted on disk under the local worlds directory. Returns the name of each stored world, giving a quick inventory of what has been saved locally.

# wld.loadLayer Load a scene as an additive layer alongside the current scene, composing its entities on top of what is already live rather than replacing it. An optional origin offset positions the loaded content in world space, and the layer can be marked persistent or given an explicit name so it can be hidden, shown, or unloaded later.

# wld.mode Report the engine's current mode. Returns `"edit"`, where scripts and physics are paused for authoring, or `"play"`, where they run. Use it to branch on the current state before flipping modes.

# wld.play Enter play mode. First auto-snapshots the current edit state so returning to edit can restore it, then starts scripts and physics so the scene runs live. This is the counterpart to returning to edit — play the scene to see it in motion, then rewind to the pre-play snapshot.

# wld.promoteAndSwitch Promote the source-file edits made during play (the play-shadow set — what accepting the "edits" group promotes), then switch the engine to the requested mode, in one call. Replaces the changes() / acceptChanges("edits") / wld.edit()-or-wld.play() sequence for the edit-and-retest loop. Entity changes stay in the sceneAuthoring review flow: a flip to edit with unaccepted entity changes stops at the leave-play safeguard.

# shared Module Cross-tool helpers for the `wld` toolbox. Every tool inside this toolbox can pull these in with `require(".shared")`. ## Purpose This module holds the small set of helpers that more than one tool in the toolbox needs — typically the `ok(value, formatted?)` / `fail(message, exitCode?)` constructors that produce the canonical `wldToolResult` shape, plus any toolbox-wide parsers, validators, or shared state. Keep it focused: per-tool implementation stays inside each `<x>.tool/init.luau` (with its own `--!desc` / `--!arg` / `--!return` / `--!example` annotations). ## Usage ```luau local shared = require(".shared") function M.someTool(...) if not ok then return shared.fail("reason", 1) end return shared.ok(value, "") end ``` ## Exports See `init.luau` for the full export surface. Typical exports include `shared.ok(value, formatted?)` and `shared.fail(message, exitCode?)` for building the toolbox's standard result envelope.

# wld.showLayer Show a previously hidden scene layer, restoring the rendering of its entities. The counterpart to hiding a layer — since hiding leaves the entities in the world, showing brings them back into view instantly with no reload.

# wld.unloadLayer Unload an additive scene layer by name, despawning its entities entirely and removing its contribution from the composed scene. This fully releases the layer's content; to keep the entities alive but stop drawing them, hide the layer instead.

# tests.compare Diff two JSON test reports and surface new failures, fixes, and added or removed tests — the "did my change break anything?" query. Save a baseline before a change, run the suite again afterwards, and compare the two: a non-empty new-failures list means the change introduced a regression. The baseline is required; the later report defaults to the most recent run when omitted.

# tests.list List every registered test suite. Discovery reads the asset registry, so the result is exactly the set the runner and CI execute. Pass a suite name to narrow to one, or ask for details to load each matching suite and include its test count.

# tests.loadHttp Fetch the engine's test content over HTTP and materialize it into the `@builtin` library so `tests.list` and `tests.run` can discover and run the suites. The engine's test content (its `.testSuite` assets, and the fixtures and packages only they use) is embedded in no build and reaches no end user. A target with a filesystem reads it off disk (the native runner boots `--dev`); a filesystem-less target (the WASM engine in a browser) uses this tool at test time. The browser bundle serves `src/lua/*` as static files under `/lua/`, and a generated `suites-manifest.json` lists each suite's files (`suites`) and the files the suites share (`support`, such as `tests/fixtures/` and `pkgtest/`). Every load materializes the shared files along with the suites it picks. A suite that reads other suites (their sources, or runs them) names the library folder that holds them on a `--!needs <folder>` line of its `init.luau`, such as `--!needs systems/procgen.package`; a load that picks it picks every suite under that folder as well, so the suites it reads are in the engine wherever it runs. Every file is written under `/zero/source/libs/@builtin/<rel>`, which registers it as a `@builtin::` asset (a suite as `@builtin::tests.suites.<name>`) with the same identity, origin, and internal-`require` resolution as a `--dev` boot. The HTTP fetch is asynchronous, so by default the work runs as a scheduler task and the call returns `{ taskId }` to watch; the browser harness polls it to completion before discovering suites. Pass `{ async = false }` to block and return `{ suites, support, files, base }`.

# tests.run Run the engine test suite and return structured results — every registered suite by default, a single named suite, or a chosen subset. Each run writes a Markdown and a JSON report to the temporary source area and can save an extra copy in a format you pick. Every suite runs through its own cooperative coroutine so multi-frame tests yield across engine frames correctly; an async run schedules the whole sweep as one task and returns its id immediately. Testing is an edit-mode authoring concern — the operation is refused in play mode, since it spawns and despawns entities, loads scenes, and flips modes. `onWorker = true` runs each suite inside a worker job (`jobs.dispatch`) rather than on the engine thread, and records `onWorker = true` on the result. The result keeps the engine-thread shape; a suite the worker cannot start counts as one failed result whose error is the text the worker gave. From the shell, `ON_WORKER=1 ./scripts/run_luau_tests.sh <suite>` adds the option.

# tests.shared (module) Internal helpers shared by the `tests` toolbox tools. Not bound to a global — required by sibling tools via `require(".shared")`. ## Surface - `M.DEFAULT_JSON` / `M.DEFAULT_MD` — canonical report paths the engine test runner writes (`/source/tmp/test_results.json`, `/source/tmp/test_results.md`). - `M.readReport(src)` — read a JSON report from a VFS path, or pass a decoded table through unchanged. - `M.diff(before, after)` — classify every test in `after` relative to the `before` baseline: `{ newFailures, fixed, stillFailing, newTests, removedTests }`. ## Example ```lua local shared = require(".shared") local before = shared.readReport("/source/before.json") local after = shared.readReport(shared.DEFAULT_JSON) local d = shared.diff(before, after) print(#d.newFailures .. " regressions") ```

# workerSuite The entry a worker job runs for `tests.run { onWorker = true }`. `run({ suite = "<name>" })` finds the `.testSuite` asset by the name `tests.run` reports it under, runs it, and returns its report to the engine-thread caller as a copy. A suite that cannot start on the worker raises, and the raise reaches the caller as the job's rejection.

# scene.clean Clear the active scene down to a protected keep-set. ```lua scene.clean() -- keep only the player setup scene.clean({ keep = { component = { "Light" } } }) -- also keep lighting scene.clean({ keepPlayer = false }) -- full empty scene.clean({ dryRun = true }) -- preview, no changes ``` By default only the player setup survives (every `PlayerPrototype` subtree and every `PlayerSpawn`), so a clean never orphans the player. `keep` protects a CATEGORY of entities — a `scene.find` query on `component` / `attribute` / `tag` / `under` — not a single named entity. A clean is a LIVE mutation only: it does not write the saved `scene.json`, so `scene.revert` undoes it and `scene.save` commits it. If the scene has unsaved edits, clean refuses (to avoid folding its removals into work no revert could restore) — save or revert those edits first. There is no confirm/force flag. The guard does not apply while play mode runs unpaused. Returns `{ removed, kept, intent }`, plus a `warning` when the clean left the scene with no lights.

# scene.cost Which loaded scene is costing the frame time. Ranks every loaded layer by what its entrypoint's `update` / `editorUpdate` costs, timed where the tick runs and attributed to the layer that owns it — the answer to "the scene is slow and the profiler's script rows do not account for it", because a scene entrypoint's per-frame tick is not a component and does not appear among them. `totalMs` is a sum across the window since the last reset (or engine start) and `avgMs` is the per-tick cost, which is the per-frame cost for a tick that runs every frame. Answers in edit mode as well as play.

# scene.create Create a NEW scene from a template and load it. ```lua scene.create("sandbox") -- returns the template catalog scene.create("level_1", "flat") -- ground + lighting, no player scene.create("arena", "player") -- ground + lighting + default player scene.create("title", "menu", { startup = true }) -- a menu scene the world opens into ``` `template` is typed: `"empty"` (blank canvas), `"flat"` (ground + lighting, no player), `"player"` (ground + lighting + the canonical reusable player a joining user's avatar replaces — prefer this over hand-rolling a player), `"menu"` (lighting only, no ground, no player). Omit the template to get the catalog of templates and what each contains; an unknown value errors with the valid choices. Pass `opts.startup = true` to also make the new scene the world's startup scene. The scene is minted as a fresh asset and loaded. Returns `{ scene, template, contains, startup? }`, or `{ templates }` when the template is omitted.

# scene.find The single entity-discovery tool. It reads the flat live population, so it reaches **runtime clones, temporary entities, and detached roots** — the entities a hierarchy walk from a scene root silently misses. Pick a search axis (each a **list**, so many candidates resolve in one call): ```lua scene.find({ name = { "muzzle", "flash" } }) -- name terms (substring by default) scene.find({ component = { "Light", "Camera" } }) -- entities carrying any of these scene.find({ attribute = { name = "team", value = "red" } }) ``` Multiple axes = union; each hit reports **which** criterion matched, so you learn the right wording. Modifiers: - `substring` (default `true`) — case-insensitive name-part matching, so `"muzzle"` reaches `muzzle_light_ent_620d…`. Set `false` for exact. - `includeInternal` (default `false`) — entities marked internal are excluded unless you opt in. - `roots` — top-level entities only. `maxDepth` — cap the hierarchy depth. - `under` — only descendants of these entities (by name or id). `ancestorsOf` — only ancestors of these. - `additive` (default `false`) — search only the active scene, or set `true` to also search loaded overlay (additive) scenes. - `limit` — cap the returned matches; `total` still reports the full count. ## Projections — `include` By default a hit says *which* entity matched. `include` makes the query answer the question outright, so you inspect a set in **one call** instead of querying and then interrogating each entity: ```lua scene.find({ under = { "room" }, include = { "transform", "componentValues" } }) ``` | Projection | Adds to every match | |---|---| | `transform` | `position = { x, y, z }` (world), `rotation = { x, y, z, w }` (world quaternion), `eulerAngles = { x, y, z }` (degrees), `scale = { x, y, z }` (local), each an **array** | | `componentValues` | `componentValues`: each component's **live** field values, keyed by component name | | `attributes` | `attributes`: the entity's attributes as `{ name = value }` | | `scriptComponents` | `scriptComponents`: every script-component instance the live entity carries, as `{ type, identity, instance }` | The `transform` vectors are arrays, the vector shape the scene and entityOps tools take and report, so a projected `position` passes straight to `entityOps.transform`. Read them by index: ```lua local hit = scene.find({ name = { "crate" }, include = { "transform" } }).matches[1] local x, y, z = hit.position[1], hit.position[2], hit.position[3] -- entity(hit.id).position holds the same value keyed x, y, z ``` Every match already carries `components`, the **names** of its components, without `include`; their live values are the `componentValues` projection. Nothing else is added unless `include` names it. An unrecognised projection is an error listing the accepted ones and naming the one to write instead (`include = { "components" }` points at `componentValues` and `scriptComponents`, `"position"` at `transform`). Projections read the **live** entity, so a component added in the same frame reports its values here even though the entity census has not caught up. Returns `{ matches, total, truncated, searchedAxes }`, each match a `{ id, name, parent, depth, components, layer, matched }` with the FULL canonical id and name, plus whatever `include` asked for. ## Writing the query under its parameter name `scene.find`'s one parameter is named `query`, so the search can also be written under that name. Fields written beside the `query` table join it, and both calls below search the same: ```lua scene.find({ query = { name = { "ground" } }, include = { "transform" } }) scene.find({ name = { "ground" }, include = { "transform" } }) ``` A field written both inside `query` and beside it is refused, naming the field. This replaces the old `find`, `findAll`, and `findByComponent`.

# scene.list List the scene snapshots available in the current world. ```lua scene.list() ``` Returns the array of scene-snapshot names registered in the live world.

# scene.load Load a scene snapshot into the live world. ```lua scene.load("main") ``` Loads via `layers.load(name)` and records `name` as the toolbox's last scene name, so a subsequent no-arg `scene.save()` persists back to the same snapshot. Returns `{ loaded = name }`.

# scene.loadedLayers The scene LAYERS loaded into the live world right now. ```lua scene.loadedLayers() ``` Reports the root scene plus any additive overlays (editor UI, effect passes) currently loaded — one record per layer with its `name`, backing `scene`, `status`, `visible`, whether it is `additive`, and whether it is the `root`. This lists loaded LAYERS (the runtime state), not the scenes available to load — for that use `scene.list`. Returns `{ LoadedLayer }`.

# scene.migrate_v6_to_v7 Converts a v6 scene's `scene.json` to the v7 format, in place and user-initiated only — the migration never runs automatically. v6 scenes carry top-level `player{}` and `camera{}` object blocks; v7 replaces both with a single top-level string `player` intent, either `"spawns"` or `"none"`, decided from the v6 `player.required_in_play` flag. On `"spawns"` it injects the final-shape `PlayerSetups`/`Spawns` structure that mirrors the `scene.assetType` template — a `DefaultSocialPlayer` prototype with a child body carrying the avatar asset and an owner-scoped camera rig, and a `DefaultSpawn` pointing back at it — with injected ids derived from the scene folder so no two converted scenes collide. Every existing entity, the lighting, and the `format` are preserved untouched. It also scans the sibling `entrypoint.luau` for references to the removed player/camera flow and returns the flagged call sites for review without ever rewriting the Luau. It is idempotent and safe: a scene already at version 7 or higher is skipped, and a scene with a missing, too-new, or invalid version is reported as an error without being mutated. Reach for it to bring an old scene onto the current player system.

# scene.observe What every scene load did, and what each loaded scene costs. One read carrying the last load's report — what it produced, the root it replaced and the overlays that went with it, the entity counts on each side, how long each phase took, and every failure it produced with the entity, component or lifecycle hook it is about — plus a per-layer inventory of what the engine attributes to each layer and the per-frame cost of each layer's entrypoint tick. Recorded at each load's own completion point and read lazily, so nothing is measured per frame beyond the tick timings themselves. Answers in edit mode as well as play mode.

# scene.player The scene's player setup, in one tool. ```lua scene.player() -- report: intent + prototypes (body/camera wiring) + spawns scene.player("none") -- playerless scene; removes every PlayerPrototype + PlayerSpawn scene.player("spawns") -- player per joining user; authors the default setup if missing scene.player("spawns", { body = "@builtin::avatars.minimal_player" }) -- swap the avatar scene.player("spawns", { behavior = "@builtin::controller.first_person" }) -- first-person scene.player("spawns", { camera = "my_cam" }) -- wire an existing camera entity ``` The intent is typed: `"spawns"` or `"none"`. `"spawns"` on a scene with no PlayerPrototype authors the canonical setup — humanoid body, third-person camera (OwnerOnly), a spawn at the origin — the same shape a fresh scene ships and the `player` template produces. `"none"` strips the player machinery entirely. A `roles` table (second argument) reshapes the player and implies `"spawns"`: - `body` — an entity name/id to adopt, **or** an avatar asset ref (`@builtin::avatars.minimal_player`) that repoints the body's avatar. A joining user's own avatar replaces this body socket, so reuse the canonical setup rather than hand-rolling one. - `camera` — an entity carrying a Camera component; moved under the prototype and scoped OwnerOnly so each joining user gets their own view. - `behavior` — any `cameraBehavior` component (built-in or authored) set on the player camera's `Camera.behavior`. - `prototype` — picks one when the scene has several. Setting anything saves the scene. The report names the root scene layer's `intent` and `scene` path, and lists every PlayerPrototype and PlayerSpawn the loaded layers hold. In play the scene holds each authored prototype as the captured template a joining player is cloned from; the report lists it through the spawn that names it, with `template = true`. With no scene loaded as the root layer the report carries `unloaded` in place of `intent` and `scene`, saying so and how to load one, and the setters refuse until a scene is loaded. The behavior is attached to the player camera's entity as a component of its own type, so it is tuned there: `component.get("<its type>")` on the camera entity and set its fields. A `component.add` of that type writes its values onto the attached instance instead of adding a second one.

# scene.reload Re-run a loaded scene's build. ```lua scene.reload() -- the active root scene scene.reload("main") -- a named scene ``` `scene.load` of a scene that is already loaded and unchanged is an idempotent no-op, so editing an entrypoint and loading it again rebuilds nothing. This unloads and re-loads the layer in place, running the entrypoint's build again — in edit mode, so iterating on a build script does not mean flipping to play and back through the play shadow. A rebuild saves the scene, the way a write to `build.luau` does: `scene.json` holds what the reload built, so the next play loads it, and unsaved edits in the scene are saved with it. A library scene (under `/libs/`) is rebuilt in the live scene alone, and its files stay as the library states them. Returns `{ reloaded = <name>, built = true, content = n, editorOnly = n, saved = bool }` once the layer has been rebuilt (`saved` is false for a library scene), `{ reloaded = <name>, built = false, reason, message }` when the build did not run, or `{ error = "..." }` when no matching layer is loaded.

# scene.remove Remove entities from the active scene. ```lua scene.remove("ground") -- delete the default floor scene.remove({ "probe_a", "probe_b" }) -- several by name/id scene.remove({ component = { "PointLight" } }) -- every entity with that component ``` Targets are entity NAMES or ids, arrays of them, `scene.find` records, or a `scene.find` query — so you delete a selection without looking it up first. Each target is removed with its whole subtree. A remove is a LIVE edit of the scene: `scene.revert` undoes it, and `scene.save` writes it into the scene's `scene.json`, which is the file `zm add` stages and `zm commit` records. To clear a scene while keeping the player setup, use `scene.clean` instead. Returns `removed`, one `{ id, name }` record per removed target, and `playerHint` when a target was part of the player setup.

# scene.replace Replace a loaded scene's **content** with another scene's content. The target scene keeps its identity (guid, name, place in the world); its files — `scene.json`, `entrypoint.luau`, everything else in the scene folder — become copies of the source's, and the scene reloads so the world reflects the swap immediately. ```lua scene.replace("@builtin::scenes.test_arena") scene.replace("my_template", { dryRun = true }) scene.replace("my_template", { scene = "hud_overlay" }) ``` Targets the active scene unless `opts.scene` names another loaded scene. Target content files the source doesn't have are removed; identity sidecars are untouched. `dryRun = true` reports exactly which files would be written and removed, plus the entities the source declares, without touching anything.

# scene.revert Discard a scene's unsaved edits and restore its saved `scene.json`. ```lua scene.revert() -- undo unsaved edits on the active scene scene.revert("main") -- revert the loaded scene named 'main' ``` The undo for `scene.clean` / `scene.remove` and any edit-mode change, and the counterpart to `scene.save`. Every edit since the last save lives in the scene's dirty overlay until you commit it; this throws that overlay away and respawns the scene from its saved state. Because the dirty overlay belongs to the scene ASSET (not the runtime layer), a revert works whether or not the scene is currently loaded — pass a `name` to revert a specific scene, or omit it for the active one. This reverts only the edit-mode dirty overlay; it is not the play-mode baseline restore. Returns `{ scene, discarded }` — `discarded` is true when unsaved edits were thrown away, false when the scene was already clean.

# scene.save Save the loaded scene named `name` back to its own `scene.json`. ```lua scene.save() -- the last-used scene name (initially "main") scene.save("main") ``` Resolves the scene by name via `layers.find` and saves that scene — saving scene A never writes scene B. Errors when no scene by that name is loaded. Returns `{ scene = name }`.

# scene.setStartup Make the currently active scene the world's startup scene. ```lua scene.setStartup() -- pin the active scene as the world's entry point ``` Takes no arguments: it pins whatever scene is live now as the scene the world auto-loads every time it loads, so after building or loading a scene you call this to make it stick across world reloads. There is no clear/unset — a world with no startup scene has no scene context and every script fails, so that is not a valid choice. The current startup scene shows in `scene.summary`. Returns `{ scene }` — the name of the scene now set as the world startup.

# scene.toolbox shared module Cross-tool sharing for the `scene.*` toolbox: the hierarchy snapshot core (`snapshot`, `findAll`, `leaf`) that backs tree / findAll / findByComponent / summary, and the last-scene-name state that save / load share. Tools call these functions; the functions call engine APIs.

# scene.spawns The scene's spawn points. `scene.player` decides whether a scene spawns players and what they become; this decides **where they arrive** — the PlayerSpawn entities themselves. ```lua tools.use("scene", "spawns") -- report every spawn tools.use("scene", "spawns", "add", { position = { 12, 0, -4 } }) tools.use("scene", "spawns", "set", { spawn = "DefaultSpawn", rotation = { 0, 90, 0 } }) tools.use("scene", "spawns", "remove", { spawn = "RedSpawn" }) ``` ## What a spawn decides - **Its transform** — where a joining player's body lands while `placement` is `at_spawn_transform`, and its forward axis is the direction they face. The edit-mode debug overlay draws that as a ring, a heading arrow and a standing capsule, so a spawn is visible where it stands. - **Its prototype** — which PlayerPrototype subtree gets instantiated for the joining user. A spawn whose prototype is unset or no longer resolves cannot spawn anyone; the report says so (`prototypeResolves`) and the overlay draws it red. In play the scene holds the prototype as the captured template each joining player is cloned from, and the spawn's prototype resolves to that template. - **team / role / maxPlayers / spawnPolicy** — the PlayerSpawn component's own fields, carried onto the players it spawns. ## Picking a spawn `opts.spawn` names the spawn to act on, by name or id. With one spawn in the scene it is implied; with several it is required, and an unmatched name is an error listing what the scene has. On `"add"` the same field names the new entity instead. New spawns are parented alongside the spawns already in the scene (else under the scene's `Spawns` container), so spawn points stay collected in one place. Mutating saves the scene.

# scene.summary High-level "what's in the scene" report, and the call that counts the entities in the scene or in one group of it. ```lua scene.summary() scene.summary({ limit = 0 }) scene.summary({ under = "level_1", limit = 5 }) scene.summary({ under = "level_1", fields = { "entityCount" } }) ``` Returns `{ entityCount, rootCount, roots, truncated, more?, under?, activeCamera, player, startupScene }`: the visible entity count, the top-level rows (with their hierarchy fields), the active camera id, the player setup with the verb that changes it, and the world's startup scene. `roots` carries at most `limit` rows, 20 when the call names none. `limit = 0` answers the counts, camera, player and startup scene alone, the read to make when the rows are not the question. `rootCount` is always the full count, and `truncated` with `more` says when rows were left out. An answer to a call naming neither `limit` nor `fields` carries `narrowWith`, the options that shorten it. `under` names entities (an id, a name, a proxy, or an array of them) and scopes the report to what is beneath them: `entityCount` counts their descendants and `roots` holds their children. A reference that names no entity is refused by name. `includeInternal = true` counts entities marked internal too. `fields` keeps only the named fields of the answer: `{ fields = { "entityCount" } }` answers the count alone, the cheapest read to repeat while a scene changes. A name no answer carries is refused, and the refusal lists the fields. For the full structure, `scene.tree` renders the hierarchy itself, and `scene.find { roots = true }` lists every root.

# scene.tree Render the entity hierarchy as an indented tree — the quick "what is the current structure" view. ```lua scene.tree() -- the whole active world, 4 levels scene.tree("player") -- just the player subtree(s) scene.tree(nil, { depth = 2 }) -- roots + their direct children ``` Each line shows the entity's name and its component types; children nest under their parent. Anything deeper than `depth` (or beyond the node budget) is summarized as an explicit `… +N below` count instead of being silently dropped. Options: ```lua { depth = 4, -- levels to expand (1 = just the roots) components = true, -- annotate each entity with its component types includeInternal = false, -- include entities marked internal (and their subtrees) limit = 500, -- node budget } ``` Returns: ```lua { text = string, -- the rendered tree (print this) roots = { { id, name, components?, childCount, children?, omitted? } }, total = number, -- entities in scope (full subtrees) shown = number, -- nodes rendered truncated = boolean, -- depth or budget cut something } ``` The target is an entity id, a name (every matching entity), a proxy, or an array of them; omit it for the whole world from its roots. Children sort by name, so repeated calls give stable output.

# scene.whyPartial Answer why a loaded layer is not whole. Returns the nearest cause from a closed set — `loaderRaised`, `entrypointCompileFailed`, `entrypointBodyRaised`, `entrypointRaised`, `buildRaised`, `entityFailed`, `parentMissing`, `parentRefused`, `parentAbandoned`, `componentUnresolved`, `componentRefused`, `subscriberRaised`, `updateRaised` — with the engine's own words for it and the full list of failures, each naming the entity, component identity or lifecycle hook it is about. A layer that produced everything its scene declared reports `ok = true` and no reason.

# yaml.module YAML decode + encode for authored engine content. Pure Luau, no engine dependencies — a general parsing library, not data-system plumbing. ```luau local Yaml = require("@builtin::modules.yaml") local doc = Yaml.decode(vfs.read(path)) vfs.write(path, Yaml.encode(doc)) ``` ## Supported subset - Block mappings (`key: value`) and block sequences (`- item`), nested to any depth. - Compact sequence-of-mapping entries (`- path: x` / ` kind: y`). - Flow collections — `{ a: 1, b: 2 }` and `[1, 2, 3]` — single line. - Plain scalars with type inference: `~` / `null` → nil, `true` / `false` → boolean, numeric literals → number, everything else → string. - Single- and double-quoted strings (`''` escapes a literal quote inside a single-quoted string; `\n` `\t` `\r` `\\` `\"` inside a double-quoted one). A quote opens a quoted scalar where a scalar begins — after `key:`, after a `- ` entry marker, or inside a flow collection — so a plain scalar carries an apostrophe as data (`note: the world's rules`). Quoted keys are read in flow mappings. - `#` comments — honored outside quotes, so a `#` inside a quoted string is kept as data. - Literal (`|`) and folded (`>`) block scalars — the bare indicators only. A block scalar's content is literal text: quotes (balanced or not), `#`, `key: value` and `---` inside it are data. - An optional leading `---` document marker. ## Loud-error contract Constructs outside the subset above raise instead of misparsing: anchors/aliases (`&`, `*`), tags (`!`), directives (`%`), multi-document streams (a second `---`), block-scalar chomping/indentation indicators (`|-`, `|+`, `>-`, `|2`, ...), tab indentation, and inconsistent indentation (both block structure and block-scalar bodies). Every error carries the offending line number (`yaml: line N: ...`), so a bad config points straight at the line to fix. Tab indentation is rejected everywhere — including block-scalar body lines whose leading whitespace contains a tab. A tab after the first non-space character of a line is content and passes through. ## API - `Yaml.decode(text: string): any` — decode a YAML document into a Luau value (table, scalar, or nil for an empty document). Raises on malformed input or an unsupported construct. - `Yaml.decodeWithComments(text: string): (any, boolean)` — decode as `decode` does, and report whether the text carries a comment by the decoder's own rule: a `#` at a line's start or after whitespace, outside a quoted scalar and outside a block scalar. Re-encoding the value drops such comments, so an editor that writes a document back reads this first. - `Yaml.encode(value: table): string` — encode a Luau table as YAML (block style, two-space indent, alphabetically sorted keys). Raises on values YAML can't represent (functions, userdata, non-string mapping keys). A top-level empty table encodes to an empty document, which decodes back to `nil` (the YAML empty-document ambiguity).

# interfaces The screen-building skill: author a UI as a styled widget tree the way you write HTML and CSS, register it, and then see it and drive it through the `gui` toolbox rather than reasoning about the tree. Covers the layout and paint model, entrance and idle animation, callbacks that fire in play mode, and the build → capture → play → iterate loop that gets a screen from blank to built.

# debugging Finding out **when** something went wrong in a running world. The job this skill covers is not "what is the current value" — the current value is downstream of the moment that matters. It is: attach an instrument to running code, keep working, and be told the instant the thing happens. It exists because the reflex it replaces does not work. Running something, waiting, and taking a capture aims at a moment you cannot see: seconds pass between the call and the frame, and adding a wait only moves the guess. Polling is the same guess with more steps. Covers: probing a function call, a source line, or a value; using a condition to name the moment rather than every step toward it; reading the state a value was taken in before trusting the value; and cancelling what you armed. The trap it spends the most words on is the expensive one — a right value read in the wrong world, where the match already ended or the scene already swapped, nothing announced it, and every number afterwards means something else.

# assetType runtime The machinery that reads a `<typename>.assetType/behavior.luau` and turns its declarations into live behaviour. A `behavior.luau` is a table of declarations — `ref` methods, `modules` shared code, an `events` schema, a `namePattern`, `refShapes`, and the `onCreate` / `onRegister` / `onChange` / `onDelete` / `validate` / `settingsDefaults` / `settingsCorrections` hooks. None of it runs itself. This is the code that runs it, and it lives here because reading a type's declarations is the **assetType type's own behaviour** — the same way `material.assetType` owns what a `.material` does. ## Sub-modules | Module | Reads | Responsibility | |---|---|---| | `ref.module` | `ref`, `modules` | Builds every `AssetRef` envelope, attaches the shared metatable, and dispatches per-type methods and shared modules onto it. Also owns edit-mode write-behind persistence. | | `create.module` | `onCreate`, `namePattern`, `validate` | Backs `asset.create`: resolves the type, checks the name against its pattern, runs `onCreate` (or clones `template/`), and writes the instance. `validate.module` beside it checks a creation `opts` table against the type's `createSchema`. Before running `onCreate`, resolves the type's settings chain (below) and hands the result to the hook as `opts.settings`. | | `settings.module` | `settings`, `settingsDefaults`, `settingsCorrections` | A type's declared settings schema and where its values live (`type.yaml`'s `settings:` block). `Settings.resolveForCreate` layers the schema's own defaults, the type's own `settingsDefaults(name, opts, asked)` hook (a DEFAULT the schema can't express — a name-derived convention, a fact read off a payload the caller handed it, a value following a field `asked` shows a higher layer naming; any later layer overrides it), the Preset Manager, an importer's settings, the caller's `settings` table, then the type's own `settingsCorrections(values, name, opts)` hook (a CROSS-FIELD CONSTRAINT the schema can't express, applied last over the fully merged values). A correction over a default/preset/importer value lands silently; a correction that would override what the caller's own `settings` table explicitly asked for refuses the creation instead, naming the field and why. Neither hook is ever handed a table to mutate; each returns what it wants changed (`settingsCorrections` may also return a `{ [field]: reason }` table for that refusal message). `Settings.cloneValues` deep-copies a values table — `create.module` hands `onCreate` a copy, never the table it stamps. `Settings.valuesOf` / `writeValues` read and write an asset's own values file. | | `typeYaml.module` | `type.yaml` | The folder a type lives in and its parsed `type.yaml`, decoded once per content version of the file. `settings.module` and `inspector.module` read their blocks through it. `about` / `aboutAt` read a type's `about.yaml` (`intent`, `function`, `usage`), leaving out any key that is empty or still starts with `[[describe]]`; `aboutRows` and `aboutMarkdown` lay the texts out for the Inspector and the `types/<name>` guide. | | `inspector.module` | `type.yaml`'s `inspector:` block | Which inspector a type declares for its assets (`inspector: { type: ... }` or `{ instance: ... }`), and the declaration that file returns — the sections it adds, the generic ones it replaces, and their order — or the problem that kept it from loading. | | `onRegister.module` | `onRegister` | Fires an instance's once-only registration hook — on arrival, on world load, and for `@builtin` library instances. | | `changeDispatch.module` | `onChange`, `onDelete`, `refShapes` | Routes a VFS source write or a folder delete to the enclosing typed asset's type hooks. | | `refShapes.module` | `refShapes`, `ref` | Publishes what every `AssetRef<category>` answers to, so a member read on an asset-typed value is checked against that category's real surface. | | `events.module` | `events` | The per-asset event runtime behind `ref.events`. | | `instantiable.module` | `ref.instantiate` | The instantiation contract an asset type opts into: `root` stands the root entity, `place` applies the base placement opts to one the type adopted, and `result` shapes the `(root, idMap)` every `instantiate` returns. Also registers the `sceneInstantiable` field-constraint validator. | | `declaration.module` | `type.yaml` | The type's decoded declaration, read once and held until a write inside an `.assetType` folder; what every question about a type as a whole reads. | | `previewRecipe.module` | `type.yaml` `preview:` | How an instance's `preview.png` is produced: which of the four recipes (`instantiate`, `image`, `surface`, `custom`) the type declares, and what that recipe needs. | ## Reaching it ```luau local rt = require("@builtin::assetTypes.assetType.shared") rt.ref.loadTypeBehavior("material") ``` Fields resolve lazily. The prelude requires the sub-modules directly, in dependency order (`ref`, then `changeDispatch`, `onRegister.install()`, `refShapes.install()`) — a table that required all six on load would take that ordering away from it.