Moonbit Js Binding logo

Moonbit Js Binding

Community
mizchi
moonbit-js-binding

Write MoonBit bindings to JavaScript with `extern "js"`. Use for FFI declarations against browser/Node/Deno APIs or npm packages, wrapping JS objects in opaque types, bridging Promises (`async fn` / `Promise::wait()`), `moon.pkg` exports (esm/cjs/iife), and null/undefined at the JS boundary.

Overview

Publishermizchi
Repositoryskills
Skill namemoonbit-js-binding
Stars
333
Forks
4
Bundled files
16
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 16 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by mizchi on GitHub. Read the source before you install it.

Installation

Install the Moonbit Js Binding AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/mizchi/skills.git /tmp/skills
mkdir -p .claude/skills
cp -r /tmp/skills/moonbit-js-binding .claude/skills/moonbit-js-binding
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Moonbit Js Binding in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Moonbit Js Binding on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Moonbit Js Binding is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

MoonBit JS Binding Guide

Step-by-step workflow for binding JavaScript APIs (browser DOM, Node built-ins, npm packages) to MoonBit through the js backend.

Verified against MoonBit v0.10.12 and moonbitlang/async 0.21.0.

When To Use

Use this skill when:

  • Adding extern "js" fn declarations for a JS API
  • Embedding inline JS (#| (x) => ...) or importing by module+name (= "Math" "max")
  • Configuring moon.pkg with link.js exports and format: esm/cjs/iife
  • Wrapping JS objects with #external pub type and %identity casts
  • Bridging Promises with async fn + %async.suspend + %async.run
  • Mapping undefined / null distinctly into MoonBit Option
  • Consuming npm packages via require(...) or ESM imports
  • Generating .d.ts for TypeScript consumers

Plan Mode Instructions

If you are in plan mode and creating a plan for moonbit-js-binding work:

  1. In your final plan file, include an explicit instruction at the beginning that tells the executing agent to read the moonbit-js-binding skill before starting implementation.
  2. The instruction should look like this:
    **Before starting implementation:** Use the Skill tool to load the moonbit-js-binding skill, which provides comprehensive guidance on `extern "js"` declarations, JS value wrapping, Promise bridging, and moon.pkg export configuration.

Type Mapping

Map JS types to MoonBit types before writing any declarations.

JavaScriptMoonBitNotes
booleanBoolDirect
number (integer)Int, UIntJS is IEEE-754; large Int64 will lose precision — use BigInt instead
number (float)Float, DoubleDirect
bigintBigIntDirect
stringStringDirect (both UTF-16)
Uint8ArrayBytes, FixedArray[Byte]Direct — no copy
Array<T>Array[T], FixedArray[T]Direct
FunctionFuncRef[(...) -> R] or closureDirect
any (opaque)#external pub type JsValueWrap in a nominal type
T | undefinedT? via is_undefined checkSee Null/Undefined section
T | nullNullable[T] wrappernull is NOT undefined
Promise<T>#external pub type Promise[T]Bridge with async fn/wait()
constant enumnumberpub(all) enum E { A; B }

Warning — Number precision: JavaScript has no integer type. Values crossing the FFI boundary round to IEEE-754 doubles. Int is fine for \|x\| < 2^53; beyond that use BigInt. UInt64 / Int64 currently map through JS number and lose bits above 2^53.

Warning — Trait objects & MoonBit enums: Passing Map[K, V], Result[_], Json, or non-#external trait objects across FFI exposes MoonBit's internal runtime shape. Convert to plain structs, JsValue, or primitives first.

Workflow

Follow these 4 phases in order.

Phase 1: Project Setup

Configure the module and package for JS output.

Module (moon.mod) — set preferred_target so moon check, moon build, moon test default to js. Since v0.10.9 the compiler's own default target is wasm, so without this every command needs --target js:

moonbit
name = "user/pkg"

version = "0.1.0"

source = "src"

preferred_target = "js"

import {
  "moonbitlang/async@0.21.0",
}

The JSON form moon.mod.json (with the hyphenated preferred-target, source and deps keys) is deprecated and slated for removal. moon fmt migrates an existing project to the DSL.

Package (src/moon.pkg) — gate .mbt files to the js backend and configure link output.

File format — moon.pkg (DSL) vs moon.pkg.json (JSON): MoonBit still reads either filename, but JSON support is deprecated and scheduled for removal. This skill uses the DSL form moon.pkg (no extension) throughout. Do NOT write moon.pkg.json; it is a different syntax (pure JSON, no options(...) wrapper, no trailing commas). Mixing them mid-project causes "Unable to read moon.pkg" parse errors. If you cloned an older template using .pkg.json, run moon fmt to convert it.

targets: — only needed for backend-specific files: The targets: block gates individual .mbt files to a subset of backends. Files that contain extern "js", extern "c", or backend-specific #cfg blocks must be listed. Pure MoonBit files (no FFI) do NOT need a targets: entry — if every file in the package is pure MoonBit, omit the targets: block entirely. The example above lists ffi.mbt / async.mbt because they contain extern "js"; a simple library with no FFI only needs link: { ... }.

import {
  "moonbitlang/async",
} for "test"

options(
  targets: {
    "ffi.mbt": ["js"],
    "async.mbt": ["js"],
    "async_test.mbt": ["js"],
  },
  link: {
    "js": {
      "exports": ["add", "greet"],
      "format": "esm",
    },
  },
)

Key fields:

FieldPurpose
targetsGate .mbt files to ["js"]. JS-only FFI files MUST be gated or they break other backends.
link.js.exportsList public functions to re-export from the output module. Use "name:alias" to rename.
link.js.format"esm" (ES modules, default for modern bundlers), "cjs" (Node require), or "iife" (browser script tag).
import ... for "test"moonbitlang/async is only needed in tests (async test) — keep it out of production build.

Warning — supported-targets vs targets: Do NOT use supported-targets: ["js"] at the package level. It blocks downstream consumers. Gate individual files with targets instead.

Warning — preferred_target = "js": This is a default, not a lock. moon test --target wasm-gc still works on files that support it — useful for cross-backend libraries.

Two ways to gate code by backend — targets: (file-level) vs #cfg() (declaration-level):

ApproachGranularityWhen to use
targets: { "foo.mbt": ["js"] } in moon.pkgWhole fileFile is wholly JS-specific (FFI-only, backend-exclusive logic). Cleanest when the JS code doesn't need to share state with other backends.
#cfg(target="js") on a declarationSingle fn, const, or blockA mostly-shared module needs a backend-specific implementation of one symbol. Keeps related code in one file.

Conditions you will see inside #cfg(...):

mbt
#cfg(target="js")                          // JS only
#cfg(target="native")                      // native only
#cfg(not(target="js"))                     // everything except JS
#cfg(any(target="wasm", target="wasm-gc")) // either wasm backend
#cfg(any(target="native", target="llvm"))  // either non-GC backend
#cfg(false)                                // always exclude (quick disable)

moonbitlang/core uses this pattern extensively — e.g. Int16::from_int64 has a %i64_to_i16 intrinsic on native/wasm but a fallback implementation on JS where 64-bit integers don't have a native representation:

mbt
///|
#cfg(not(target="js"))
pub fn Int16::from_int64(self : Int64) -> Int16 = "%i64_to_i16"

///|
#cfg(target="js")
pub fn Int16::from_int64(self : Int64) -> Int16 {
  Int16::from_int(self.to_int())
}

The same file-level rule applies to #cfg: every target your function is callable from must have exactly one matching definition. Overlapping #cfg guards are a compile error. #cfg(target="js") + #cfg(not(target="js")) cover everything; #cfg(target="js") alone on a non-intrinsic declaration means that function does not exist on other backends and any code that calls it must itself be #cfg-gated (or live in a file gated via targets:).

Choosing between them: If the whole file is extern "js" wrappers, gate the file via targets: — simpler, and moon check --target wasm-gc skips the file entirely. If you are writing a cross-backend library where 90% of the logic is shared and only a few functions differ, reach for #cfg.

Phase 2: FFI Layer

Write extern "js" declarations. Keep them private (or prefixed ffi_); expose safe wrappers in Phase 3.

Inline JS (#| literal):

mbt
///|
extern "js" fn ffi_console_log(msg : String) -> Unit =
  #| (msg) => console.log(msg)

///|
extern "js" fn ffi_sqrt(x : Double) -> Double =
  #| (x) => Math.sqrt(x)

The #| literal is the function body as raw JS source. The MoonBit compiler inlines it at the call site — no closure allocation at runtime.

Module-and-name form (imports an existing JS function directly):

mbt
///|
extern "js" fn math_max(a : Double, b : Double) -> Double = "Math" "max"

Use this when you want to reference a globally accessible function without wrapping it in a closure. Slightly faster and smaller.

#module("pkg") — compile-time ESM / CJS import:

For named exports from JS modules (Node built-ins, npm packages, local .js files), annotate the extern "js" with #module("module-id"). The right-hand side is the export name inside that module (not an inline JS body):

mbt
///|
#module("node:path")
extern "js" fn path_basename(p : String) -> String = "basename"

///|
#module("node:path")
extern "js" fn path_extname(p : String) -> String = "extname"

What the compiler emits:

link.js.formatGenerated code
"esm"import { basename as basename$7 } from "node:path";
"cjs"const { basename: basename$1548 } = require("node:path");
"iife"same CJS-style destructure (wrapped in the IIFE)

Why #module() beats a hand-rolled require("...") wrapper:

  • Statically analyzable. Vite / esbuild / Rollup can trace the dependency, tree-shake, and resolve paths correctly.
  • Format-agnostic source. The same .mbt file works under ESM (where require isn't defined) and CJS without a second code path.
  • Smaller generated code. No per-call closure; the destructured local is reused.
  • Bundler-compatible paths. Works for npm package specifiers ("lodash"), Node built-ins ("node:path"), relative files ("./helpers.js"), and URL-style specifiers when supported by the runtime.

Constraints:

  1. Named exports only. The RHS must be a key exported from the module. For default exports or modules whose top-level value is the function itself, you need a bridge re-export on the JS side, or fall back to inline import/require in an FFI body.
  2. One export per declaration. You cannot share a destructure across multiple #module FFIs — each annotation generates its own import line. The compiler deduplicates the module specifier, so you pay one require per unique module, not per function.
  3. Module-name literal must be compile-time constant. Dynamic module paths (#module(some_variable)) are not supported. For dynamically chosen modules, fall back to require(name) in inline JS.
  4. No inline body allowed. #module("pkg") + extern "js" fn f() = "name" is the only shape. Mixing with #| (...) => ... is a compile error.
  5. Browser globals don't need it. For window, document, fetch, Math, etc., use inline JS or the module+name form (= "Math" "max"); they are not imported, just accessed.
  6. Side-effectful imports are still side-effectful. The import/require fires when the compiled module loads. If pkg has top-level side effects (CLI banners, global patching), they run during import even if you never call the FFI function.
  7. Test vs release format. moon test uses CJS harness regardless of your link.js.format. If a module is ESM-only (no CJS entry point in package.json exports), its tests may fail even though moon build --release works. Gate those tests or use a dual-format package.

When to use which form:

FormUse when
#|(x) => ... inlineSmall expression, complex logic, need to compose multiple JS calls in one wrapper
= "Math" "max" module+nameCalling a pre-defined global function (no module import needed)
#module("pkg") + = "name"Calling a named export from an importable module (Node built-in, npm package, local file)
require("pkg").get("name") helperDynamic module name, or module-name only known at runtime

Multi-line inline JS:

mbt
///|
extern "js" fn ffi_do_work(x : Int) -> Int =
  #| (x) => {
  #|   const y = x * 2;
  #|   return y + 1;
  #| }

Opaque JS value pattern:

For anything without a MoonBit equivalent (DOM nodes, npm objects, undefined, null), wrap it in #external pub type:

mbt
///|
#external
pub type JsValue

///|
/// Zero-cost bitcast — compiles to nothing.
pub fn[A, B] identity(a : A) -> B = "%identity"

%identity is a compiler intrinsic: it moves a value across types at the type-checker level and emits no JS code. Use it to lift primitives into JsValue and back. It is unchecked at runtime — wrong use is a silent type error.

Generic FFI operations on JsValue:

mbt
///|
pub extern "js" fn JsValue::get(self : JsValue, key : String) -> JsValue =
  #| (obj, key) => obj[key]

///|
pub extern "js" fn JsValue::set(
  self : JsValue,
  key : String,
  value : JsValue,
) -> Unit =
  #| (obj, key, value) => { obj[key] = value }

///|
pub extern "js" fn JsValue::call_method(
  self : JsValue,
  name : String,
  args : Array[JsValue],
) -> JsValue =
  #| (obj, name, args) => obj[name](...args)

These three primitives ( get / set / call_method ) can express virtually any JS object interaction.

Null / undefined distinction:

MoonBit T? cannot represent T | null | undefined safely — you'd get Some(null). Check explicitly with is_undefined / is_null:

mbt
///|
pub extern "js" fn is_undefined(v : JsValue) -> Bool =
  #| (v) => v === undefined

///|
pub fn[T] js_get_opt(obj : JsValue, key : String) -> T? {
  let v = obj.get(key)
  if is_undefined(v) {
    None
  } else {
    Some(identity(v))
  }
}

For libraries that distinguish null from undefined (e.g. FileReader.error), introduce Nullable[T] and Nullish[T] wrapper types and narrow with explicit checks.

Phase 3: MoonBit API

Build safe, typed, pub wrappers over the raw ffi_* externs.

mbt
///|
pub fn console_log(msg : String) -> Unit {
  ffi_console_log(msg)
}

///|
pub fn sqrt(x : Double) -> Double {
  if x < 0 { panic() }
  ffi_sqrt(x)
}

Exports to JS consumers:

Anything listed in moon.pkg link.js.exports must be pub. The compiler emits a named ESM export (or CJS property, or IIFE global) for each. A .d.ts is generated automatically from the public signature.

#export_name (v0.10.4) pins the emitted symbol name on a single pub, non-generic function and is now preferred over backend-specific exports link configuration for new exports. It does not work on generic functions or functions with optional arguments, and the name must be a valid C identifier that is unique in the package.

mbt
#export_name("add_ints")
pub fn add(a : Int, b : Int) -> Int { a + b }

Export configuration is scoped to the package that produces the artifact — an attribute inside a dependency does not add a symbol to your output.

Phase 4: Promise / Async Bridging

MoonBit async uses two compiler intrinsics that translate to JS Promise internals:

IntrinsicRole
%async.suspendConvert a resume/reject callback pair into an await point
%async.runStart an async computation from a sync context

Canonical setup (see references/promise-bridging.md for full detail). Declare in this order — each item depends on the previous:

mbt
///|
#external
pub type Promise[T]

///|
/// Cast Promise[T] to JsValue so JsValue methods (get/call_method) apply.
/// Required by `Promise::wait` below.
pub fn[T] Promise::to_any(self : Promise[T]) -> JsValue = "%identity"

///|
pub async fn[T, E : Error] suspend(
  f : ((T) -> Unit, (E) -> Unit) -> Unit,
) -> T raise E = "%async.suspend"

///|
pub fn run_async(f : async () -> Unit noraise) -> Unit = "%async.run"

///|
/// The bridge. `await promise` in JS = `promise.wait()` in MoonBit.
pub async fn[T] Promise::wait(self : Promise[T]) -> T {
  suspend(fn(ok, err) {
    self.to_any().call_method("then", [identity(fn(v : JsValue) { ok(identity(v)) })])
        .call_method("catch", [identity(fn(e : JsValue) { err(identity(e)) })])
    |> ignore
  })
}

When do you need %async.run vs just %async.suspend?

SituationNeed run_async?Why
async test "..." { ... } blockNoThe test harness already runs the block in an async context
pub async fn calling .wait()NoThe caller is already async; await chains naturally
Exported MoonBit sync function that kicks off a Promise and returns it to JSYesYou are crossing sync → async; run_async schedules the body
Constructing a Promise[T] from an async fn executorYesThe executor body must run even though new Promise(...) is sync

In short: async test users never write run_async. It is needed only when %identity-casting async computations into JS Promise land from a sync context.

Standard pattern — FFI returning Promise + async wrapper:

mbt
///|
extern "js" fn ffi_fetch_text(url : String) -> Promise[String] =
  #| (url) => fetch(url).then(r => r.text())

///|
pub async fn fetch_text(url : String) -> String {
  ffi_fetch_text(url).wait()
}

Callers write let body = fetch_text(url) inside an async test or async fn — no .wait() needed at the call site. Tests:

mbt
async test "fetch_text resolves" {
  let body = fetch_text("https://example.com")
  inspect(body.length() > 0, content="true")
}

Note — moonbitlang/async: async test blocks require import { "moonbitlang/async" } for "test" in moon.pkg. The dependency is only needed at test time. Add with moon add moonbitlang/asyncversion 0.17.0 or later is required (older 0.1.x releases don't support the current %async.suspend ABI). Verified with 0.21.0.

Breaking changes to watch when upgrading past 0.20: @http.Headers is now a case-insensitive map rather than Map[String, String]; @fs dropped the create / truncate parameters and @fs.write_file now defaults to CreateOrTruncate; @stdio.{Input,Output}::fd raises instead of returning; @async.protect_from_cancel defaults to resume_on_cancel=true. The runtime also aborts on detected deadlocks (@async.set_deadlock_handler to configure).

Phase 5: Testing

bash
moon check                           # type-check, fast
moon test                            # run all tests (js is default)
moon test src --filter "greet"       # filter by name
moon build                           # emit .js + .d.ts to _build/js/release/build/

Build output paths — debug vs release:

CommandOutput directory
moon build (default)_build/js/debug/build/<pkg>.js + .d.ts
moon build --release_build/js/release/build/<pkg>.js + .d.ts
moon test (auto)_build/js/debug/test/<pkg>.{internal,blackbox}_test.js

.d.ts is generated in both modes. For publishing to npm, use --release. For local experimentation, the default debug output is sufficient and links faster.

Running the generated module from Node:

bash
# debug build
node --input-type=module -e 'import("./_build/js/debug/build/<pkg>.js").then(m => console.log(m.add(2,3)))'

# release build (after `moon build --release`)
node --input-type=module -e 'import("./_build/js/release/build/<pkg>.js").then(m => console.log(m.add(2,3)))'

A runnable end-to-end example lives under assets/js_binding_proj/ in this skill — see the directory for a minimal project with 8 passing tests (sync + async) that you can copy, adapt, and run with moon test.

Decision Table

SituationPatternKey Action
Call a JS global (one-liner)Inline #| with arrow functionextern "js" fn ... = #| (x) => ...
Reference existing global (no wrap)Module+name form= "Math" "max"
Named export from npm / Node built-in#module("pkg") + = "exportName"Compiler emits import { exportName } or require, bundler-friendly
Opaque JS object#external pub type TWrap + cast with %identity
JS any containerSingle JsValue type with get/set/call_methodBuild object-protocol helpers once, reuse
T | undefinedCheck is_undefined(v), return T?Never assume Some(null) is meaningful
T | null (distinct from undefined)Nullable[T] wrapper typeExplicit is_null branch
JS Promise#external type Promise[T] + wait()Bridge via %async.suspend
Export to JS consumerslink.js.exports in moon.pkgList pub functions; .d.ts auto-generated
Gate a whole file to one backendtargets: { "f.mbt": ["js"] } in moon.pkgCleanest when the whole file is backend-specific
Gate a single function to one backend#cfg(target="js") above the fnUse when 90% of the file is shared; pair with #cfg(not(target="js")) for the other backends
Large integer (|x| > 2^53)BigInt, not Int64JS number loses bits above 2^53
Call an npm packageWrap require("pkg")JsValue.get("fn")See references/interop-patterns.md
Catch a JS exceptionWrap call in try { ... } catch { ... } on JS sideSee references/error-handling.md

Common Pitfalls

  1. Forgetting to gate files with targets. A file containing extern "js" that isn't gated to ["js"] breaks moon check --target wasm-gc and native. Always gate FFI files.

  2. Using T? to model T | null | undefined. MoonBit's None is not JS null, and Some(null) is a nonsense value. Split with is_undefined and is_null before lifting.

  3. Relying on Int64 across FFI. JS number is IEEE-754 double. Any 64-bit integer > 2^53 silently rounds. Use BigInt or split into two Ints.

  4. Exposing MoonBit internals by passing Map/Result/trait objects directly. These have a MoonBit-specific runtime layout. Convert to plain structs, Array[(K, V)], or JsValue at the boundary.

  5. Missing moonbitlang/async for async test. The test-only import is easy to forget. Add import { "moonbitlang/async" } for "test" to moon.pkg before writing the first async test.

  6. Forgetting %async.run when calling async from sync JS. Exported sync functions can't await. If you need to fire an async operation from a sync export, wrap it with run_async(async fn() noraise { ... }) and return a Promise explicitly.

  7. Calling .wait() without being in an async context. .wait() is async — it only compiles inside an async fn or async test. Non-async callers need the Promise directly.

  8. Using %identity where a real conversion is needed. %identity is a type-checker escape hatch with no runtime effect. Using it to cast IntString will produce a runtime error with no warning. Only use when the JS runtime representation genuinely matches.

  9. Exporting a pub function that isn't in link.js.exports. It is callable from other MoonBit packages but not re-exported from the compiled JS module. Add the name to exports.

Example Project

A minimal, runnable project demonstrating every pattern above:

assets/js_binding_proj/
├── moon.mod                # preferred_target = "js", moonbitlang/async dep
└── src/
    ├── moon.pkg            # targets + link.js.exports
    ├── ffi.mbt             # extern "js" + JsValue + inline JS
    ├── lib.mbt             # Safe public wrappers (add, greet, js_get_opt)
    ├── lib_test.mbt        # Sync tests
    ├── async.mbt           # Promise[T] + suspend + wait
    ├── async_test.mbt      # async test blocks
    ├── modules.mbt         # #module("node:path") ESM/CJS import demo
    ├── modules_test.mbt
    ├── cross_target.mbt    # #cfg(target="...") per-declaration gating demo
    └── cross_target_test.mbt

Run:

bash
cd assets/js_binding_proj
moon test              # 12 tests pass (sync + async + #cfg + #module)
moon build             # emits ESM + .d.ts
node --input-type=module -e 'import("./_build/js/debug/build/js_binding_proj.js").then(m => console.log(m.add(2,3), m.greet("world"), m.basename("/a/b/c.html")))'
# → 5 Hello, world! c.html

References

@references/promise-bridging.md @references/interop-patterns.md @references/error-handling.md @references/typescript-integration.md @references/cfg-and-target-gating.md

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Moonbit Js Binding AI skill do?

Write MoonBit bindings to JavaScript with `extern "js"`. Use for FFI declarations against browser/Node/Deno APIs or npm packages, wrapping JS objects in opaque types, bridging Promises (`async fn` / `Promise::wait()`), `moon.pkg` exports (esm/cjs/iife), and null/undefined at the JS boundary.

Why use Moonbit Js Binding on TypingMind?

Because you install it once and use it with any model. Moonbit Js Binding is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Moonbit Js Binding in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/mizchi/skills/tree/main/moonbit-js-binding. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Moonbit Js Binding?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Moonbit Js Binding?

As many as you like. As long as a model supports skills, you can use Moonbit Js Binding with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Moonbit Js Binding AI skill free?

It is published on GitHub by mizchi. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇