Writing Node.js .mjs Modules
SkillFiles & storageUse when creating or modifying any .mjs file in the Handsontable monorepo - scripts, utilities, or library modules. Covers .mjs conventions, native node: imports, top-level await, cross-platform compatibility, and fs/promises async patterns. Trigger on any new .mjs file creation, not just files in scripts/ directories.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Writing Node.js .mjs Modules skill
What this skill tells your AI
The instructions your AI receives, as published by handsontable/handsontable in .claude/skills/node-scripts-dev/SKILL.md and read by ahel’s review.
All Node.js-side code in the monorepo -- scripts, utilities, and library modules -- uses ESM (.mjs). Follow these conventions for any .mjs file, whether it lives in scripts/, lib/, or elsewhere.
File conventions
- Extension: Always
.mjs(never.jsor.cjsfor Node.js-side code). - Location:
scripts/for CLI-invoked scripts.lib/for shared utilities and library modules. Package-specific paths are fine (e.g.,performance-tests/lib/,wrappers/react-wrapper/scripts/). - Invocation:
node scripts/your-script.mjsfrompackage.jsonscripts, or as acmdvalue inhandsontable/scripts/tasks.json(see below). - Scope: These conventions apply to all
.mjsfiles -- standalone scripts, library modules, Playwright helpers, build tooling, etc.
Adding npm scripts to the handsontable core package
The handsontable/ package uses a unified dispatcher. Do not add raw shell commands directly to handsontable/package.json scripts. Instead:
- Add the task to
handsontable/scripts/tasks.json:
"my-task": {
"cmd": "node scripts/my-script.mjs",
"deps": ["build:styles"],
"mode": "inherit"
}
- Add a thin shim to
package.jsonthat delegates to the dispatcher:
"my-task": "node scripts/run.mjs my-task"
tasks.json schema
| Field | Required | Values | Purpose |
|---|---|---|---|
cmd | yes | shell string | Command run via spawn(..., { shell: true }) |
deps | no | task name array | Tasks that must complete first (resolved by DAG scheduler in parallel mode; resolved sequentially in direct invocation mode) |
mode | no | quiet (default) | inherit | interactive | quiet = suppress output with spinner; inherit = stream output (linters); interactive = full TTY pass-through (Jest) |
cwd | no | path relative to handsontable/ | Working directory override |
passthrough | no | boolean | Append extra CLI flags (after --) to the cmd |
note | no | string | Human annotation only, ignored at runtime |
Pipeline definitions
To group tasks into an ordered pipeline (e.g., a build or test sequence), add to the pipelines block:
"pipelines": {
"my-pipeline": {
"before": ["clean"],
"tasks": ["my-task", "other-task"],
"after": ["prepare-package-for-publish"]
}
}
before and after steps run sequentially. tasks run sequentially with --sequential or via DAG with --parallel.
Other packages
For wrapper packages (wrappers/react-wrapper/, wrappers/angular-wrapper/, wrappers/vue3/) and other monorepo packages (performance-tests/, visual-tests/), add directly to that package's package.json scripts as usual — those packages do not use the run.mjs dispatcher.
Native module imports
Always use the node: protocol prefix for built-in modules. This makes it explicit that the import is a Node.js built-in, not a third-party package.
// GOOD
import { readdir, readFile, writeFile, rename, rm } from 'node:fs/promises';
import { join, resolve, dirname } from 'node:path';
import { fileURLToPath } from 'node:url';
import { exec } from 'node:child_process';
import { promisify } from 'node:util';
// BAD - missing node: prefix
import { readFile } from 'fs/promises';
import { join } from 'path';
Top-level await
Use top-level await directly in the module scope. No need to wrap in an async function main() unless you need structured error handling.
// GOOD - direct top-level await
const content = await readFile(filePath, 'utf-8');
const entries = await readdir(dir, { withFileTypes: true });
// Also acceptable - when you need a try/catch boundary
try {
await doWork();
} catch (err) {
console.error(err);
process.exitCode = 1;
}
Prefer native modules
Do not add third-party dependencies for tasks that Node.js handles natively:
| Task | Use | Not |
|---|---|---|
| Read/write files | node:fs/promises | fs-extra |
| Delete recursively | rm({ recursive: true, force: true }) | rimraf |
| Move/rename | rename() | mv |
| Run child process | node:child_process + promisify(exec) | execa |
| Parse CLI args | node:util parseArgs() or manual process.argv | yargs, commander |
| Path manipulation | node:path | slash, normalize-path |
| Glob matching | node:fs readdir + filter | glob, fast-glob (unless complex patterns needed) |
Cross-platform compatibility
Scripts must work on Linux, macOS, and Windows. This is the monorepo's most common scripting gotcha.
- No bash constructs in
package.jsonscripts: noif [ ],mv,rm -rf,&&chaining with||. - Use
node:pathjoin()for all paths -- never hardcode/separators. - Use
node:fs/promisesasync APIs (readdir,rename,rm,access) -- not their sync counterparts. - Use
import.meta.dirname(Node 21+) ordirname(fileURLToPath(import.meta.url))for__dirnameequivalent. - Reference:
wrappers/react-wrapper/scripts/prepare-types.mjsas a well-structured example.
Existence check pattern
const exists = async (path) => access(path).then(() => true, () => false);
Script structure template
// Description of what this script does.
//
// Usage: node scripts/my-script.mjs [--dry-run]
import { readdir, readFile, writeFile, mkdir } from 'node:fs/promises';
import { join } from 'node:path';
const ROOT = import.meta.dirname
? join(import.meta.dirname, '..')
: join(dirname(fileURLToPath(import.meta.url)), '..');
const dryRun = process.argv.includes('--dry-run');
// ... script logic using top-level await ...
const files = await readdir(join(ROOT, 'src'), { withFileTypes: true });
for (const entry of files) {
if (!entry.isDirectory()) continue;
// process entry
}
console.log('Done.');
Error handling
- Set
process.exitCode = 1on failure instead ofprocess.exit(1)-- allows cleanup to finish. - Silent
catchblocks must include a comment explaining why the error is swallowed. - Log actionable error messages -- include the file path or operation that failed.
Existing scripts for reference
| Script | Purpose |
|---|---|
scripts/sync-skills-to-cursor.mjs | Sync .claude/skills to .cursor/rules |
scripts/translate-to-native-npm.mjs | Workspace command delegation |
scripts/verify-bundles.mjs | Post-build version verification |
scripts/pre-release.mjs | Pre-release version generation |
wrappers/react-wrapper/scripts/prepare-types.mjs | Cross-platform type preparation |
Signals
- GitHub stars
- 22k
- Forks
- 3k
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
node-scripts-dev- Source
- github.com/handsontable/handsontable