MCPInstall

rslib-modern-package

Design, modernize, or review JS/TS npm packages using an opinionated Rslib baseline for ESM, exports, types, dependencies, and release readiness.

Installs
115
Last 24h
—
Trending rank
—
Security
Not audited

Install

Works with every agent the skills CLI supports.

npx skills add https://github.com/rstackjs/agent-skills --skill rslib-modern-package
  1. 01

    Install

    Run the command in your project. The CLI asks which agents to install for.

  2. 02

    Describe the task

    Ask your agent for something this skill covers. It loads rslib-modern-package automatically when relevant.

  3. 03

    Review the result

    The agent follows the SKILL.md steps, scripts and references bundled with the skill.

SKILL.md

The instructions your agent reads when the skill is triggered.

Rslib modern package

Use this skill when creating a new Rslib library, modernizing an existing JS/TS package, or reviewing a package against an opinionated modern library standard.

This skill is opinionated: it describes a recommended modern package contract and may suggest breaking changes when they make the package simpler, safer, and easier for modern consumers.

Standard

Default recommendation for new JS/TS libraries:

  • ESM-first, preferably pure ESM.
  • Strict TypeScript and correct declaration files.
  • Explicit public API through package.json#exports.
  • Small named-export API surface.
  • Few runtime dependencies, without treating zero dependencies as a religion.
  • Small, tree-shakeable output with accurate sideEffects.
  • Clear dependency placement: runtime dependencies, peer dependencies, optional dependencies, and dev dependencies are not interchangeable.
  • Published package is tested as an artifact, not just as source files.
  • Release flow is automated, traceable, and SemVer-aware.
  • README.md explains usage for humans; AGENTS.md preserves package invariants for future agents.

Workflow

  1. Inspect the package contract first

    • Read package.json, lockfile/package manager, rslib.config.*, tsconfig*, CI/release config, README.md, AGENTS.md, and existing dist output.
    • Identify package kind: Node utility, browser library, isomorphic utility, CLI, UI/component library, framework plugin, SDK, or adapter.
    • List supported runtimes, current entry points, deep imports, runtime dependencies, peer dependencies, optional integrations, files with side effects, and published files.
    • Run npm pack --dry-run early when changing package shape so the real tarball contents guide the review.
  2. Define target environments explicitly

    • Do not say "supports modern environments" without defining them.
    • Verify the current Node.js release schedule before choosing engines.
    • As of May 9, 2026, Node.js 22 and 24 are LTS, and Node.js 20 is EOL. For new Node-facing packages, recommend engines.node >=22 unless real consumers need an older runtime.
    • Browser packages should state whether they require native ESM, a bundler, Workers support, SSR compatibility, DOM APIs, CSS processing, or specific browser baselines.
    • Compatibility drops are breaking changes: old Node/browser versions, undocumented deep imports, default/named export shape, bundled vs external dependency behavior, and import side effects.
  3. Prefer pure ESM, but explain compatibility cost

    • New packages should use "type": "module" and ESM source/output.
    • Rslib's default format is ESM; keep that default unless there is a clear reason to add another format.
    • Evaluate compatibility from real consumers and supported runtimes instead of assuming every historical module format is required.
    • Modern Node.js can load synchronous ESM from CommonJS via require(esm); do not assume CJS consumers always require a separate CJS build.
    • If you rely on require(esm) compatibility, document and test its constraints: supported Node versions, no top-level await in the loaded graph, namespace-object return shape, default export behavior, and CJS/ESM cycle limits.
    • Prefer Node built-in specifiers such as node:fs/promises.
  4. Make exports the public API

    • Treat package.json#exports as the product contract.
    • Export only paths users are meant to import.
    • Do not allow imports like pkg/dist/foo.js by exporting ./dist/*. That makes the generated output layout part of the public API and turns internal file moves into breaking changes.
    • Instead, expose only intentional public paths such as pkg and pkg/foo.js, mapped to the actual files in dist.
    • Keep subpath style consistent: either all with extensions such as ./foo.js, or all without extensions. Prefer paths with extensions when browser import maps matter.
    • Keep "types" first inside conditional exports.
    • Adding exports to an older package can be breaking because undeclared deep imports stop working.
    • Add ./package.json only when consumers legitimately need package metadata.
  5. Design a small API surface

    • Prefer named exports for multi-API packages.
    • Avoid default-export objects that gather every function into one object.
    • Public functions should be few, stable, well-named, and semver-maintained.
    • Keep internal types, caches, helper functions, adapter details, and error internals private unless they are part of the contract.
    • Avoid top-level work during import: file scans, network calls, timers, process mutation, global registration, DOM access, prototype mutation, or environment detection with side effects.
    • Async APIs that may be canceled should accept AbortSignal.
    • Prefer stable error classes, error codes, or typed error shapes over string matching.
  6. Use Rslib as the implementation path, not the whole standard

    • For detailed Rslib configuration guidance, use the rslib-best-practices skill.
    • In this skill, only check whether Rslib output, declarations, package.json#exports, files, dependencies, and docs agree with the modern package contract.
    • Keep Rslib configuration small and intentional; avoid adding build complexity that does not improve the package contract.
  7. Keep dependencies small and intentional

    • Start from platform APIs, not from dependency search.
    • Prefer built-ins when the runtime supports them: URL, URLSearchParams, Intl, fetch, AbortController, structuredClone, crypto.randomUUID, Web Streams, TextEncoder, TextDecoder, node:fs/promises, and node:crypto.
    • Small runtime dependencies are fine when they reduce maintenance risk or implementation complexity.
    • Avoid large utility packages for one or two helpers.
    • Evaluate dependencies for ESM support, exports, types, transitive dependency count, package size, license, maintenance activity, security history, install scripts, side effects, native install fragility, and granular imports.
    • Put required runtime packages in dependencies.
    • Put host-owned frameworks and toolchains in peerDependencies, such as React, Vue, Svelte, Rspack, Rsbuild, webpack, TypeScript, and framework runtimes.
    • Keep peer ranges reasonably broad; do not pin peers to a patch version unless required.
    • Put build tools, test tools, type tools, docs tools, and Rsbuild/Rspack plugins in devDependencies.
    • Use optionalDependencies or optional peers via peerDependenciesMeta for optional integrations.
  8. Make TypeScript strict and package-oriented

    • Prefer TypeScript source for TS libraries; otherwise use high-quality JSDoc plus generated declarations.
    • With TypeScript 6 or tsgo-era defaults, strict checking may already be enabled; preserve that default and do not turn it off. For older TypeScript versions or inherited configs, set strict: true explicitly.
    • In Rslib projects, consider enabling lib.dts.tsgo to speed up declaration generation when the project can use tsgo.
    • Keep module and moduleResolution aligned with how declarations are emitted and how consumers resolve the package; NodeNext and bundler-style resolution are both valid in the right toolchain.
    • Use verbatimModuleSyntax so type-only imports/exports are explicit.
    • Use isolatedDeclarations when practical so exported APIs are explicit enough for declaration-oriented tooling.
    • Emit declarations; use declaration maps when editor navigation matters.
    • Use import type and export type for type-only dependencies.
    • Do not rely on consumers setting skipLibCheck to hide broken package types.
    • Test declarations as consumers see them, especially when using subpath exports.
  9. Keep sideEffects accurate

    • Use sideEffects: false only when importing package files has no top-level side effects.
    • If CSS, polyfills, registrations, global listeners, prototype changes, or other import-time mutations exist, list the files with side effects instead.
    • Do not set sideEffects: false just to improve bundle size; incorrect values can remove required CSS or setup code.
    • Do not change globals just because a file is imported. If setup is required, expose an explicit setup() or install() function and let users call it themselves.
  10. Make package.json authoritative

    • Required modern shape: "type": "module", explicit exports, correct declarations, files allowlist, accurate sideEffects, sensible engines, and release scripts.
    • Include README.md, AGENTS.md, and LICENSE in files when they exist.
    • files should prevent tests, fixtures, private docs, build caches, local configs, and large generated artifacts from leaking into the tarball.
    • Avoid stale main/module fields unless compatibility evidence requires them; if kept, they must agree with exports.
    • Keep runtime dependency fields accurate. A package that works locally only because a runtime dependency is in devDependencies is broken.
  11. Validate the published artifact

    • Run normal lint, typecheck, tests, and rslib build.
    • Smoke test built ESM output.
    • Run type-level tests when the public API is type-heavy.
    • Run npm pack --dry-run and inspect included files.
    • In Rslib, prefer rsbuild-plugin-publint to run publint after build; use npx publint as a CLI fallback.
    • In Rslib, prefer rsbuild-plugin-arethetypeswrong to run Are The Types Wrong after build when declarations are shipped; use npx --yes @arethetypeswrong/cli --pack . as a CLI fallback.
    • Install the packed tarball into clean consumer fixtures for important packages.
    • Test ESM import, bundler import for browser/component libraries, CLI execution for bin packages, and every public subpath export.
  12. Prepare README.md and AGENTS.md before publishing

    • Always check whether both files exist before publishing or modernizing a package.
    • If either file is missing, recommend adding it; for implementation tasks, create a concise version unless the user asks not to.
    • README.md should include: package name, one-sentence purpose, install/usage, key features or API links, supported environments, docs/related links, changelog or contribution link, and license.
    • AGENTS.md should include: stack, package contract, common commands, source layout, code style, validation commands, and release checklist.
    • Keep both files synchronized with package.json#exports, supported runtimes, and actual Rslib output.
  13. Publish with supply-chain hygiene

    • Follow SemVer and document breaking changes.
    • Maintain a changelog for user-visible changes.
    • Use prerelease versions and dist-tags for beta/next channels.
    • Prefer CI publishing with npm provenance or trusted publishing.
    • Avoid long-lived publish tokens where trusted publishing is available.
    • Remember that a published package name/version pair cannot be reused safely.

Review red flags

  • exports is missing, points to files not emitted by Rslib, or allows public imports such as pkg/dist/foo.js.
  • module/main fields disagree with exports.
  • Type declarations do not match runtime entry points.
  • Runtime dependency is accidentally listed only in devDependencies.
  • React/Vue/Svelte/Rspack/Rsbuild/webpack/TypeScript is bundled or placed in dependencies when it should be a peer.
  • sideEffects: false is set while importing CSS, polyfills, global registrations, global listeners, or prototype changes.
  • Package has install scripts without a strong reason.
  • Top-level import reads user files, starts timers, touches the network, mutates globals, or assumes window/process.
  • Published tarball contains private source maps, tests/fixtures that are not useful to consumers, large generated docs, local config secrets, or build caches.

Checklist

  • Supported environments are explicit.
  • Package is ESM-first, preferably pure ESM.
  • package.json has "type": "module".
  • Public entry points are declared in exports.
  • No accidental reliance on undeclared deep imports.
  • Rslib output, declarations, exports, and files agree.
  • TypeScript strict mode is enabled.
  • Declarations are emitted and validated.
  • Runtime dependencies are justified and small.
  • Host frameworks/toolchains are peers.
  • Build/test/type/docs tools are dev dependencies.
  • sideEffects is accurate.
  • npm pack --dry-run has been inspected.
  • publint passes.
  • Are The Types Wrong check passes when declarations are shipped.
  • Built ESM smoke test passes.
  • README.md exists.
  • AGENTS.md exists.
  • Release flow uses SemVer, changelog, and provenance/trusted publishing when available.

Documentation

FAQ

What does the rslib-modern-package skill do?

Design, modernize, or review JS/TS npm packages using an opinionated Rslib baseline for ESM, exports, types, dependencies, and release readiness.

How do I install rslib-modern-package?

Run npx skills add https://github.com/rstackjs/agent-skills --skill rslib-modern-package in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/rslib-modern-package/ (personal) or .claude/skills/rslib-modern-package/ (project).

Which AI agents can use rslib-modern-package?

Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.

Is rslib-modern-package safe to install?

No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.

Who maintains rslib-modern-package?

It is published by rstackjs in rstackjs/agent-skills. It has 115 recorded installs.

More from rstackjs/agent-skills

View all →

rsbuild-best-practices

rstackjs/agent-skills

Configure, review, or troubleshoot Rsbuild applications, including build/dev commands, assets, type checking, and bundle optimization.

rstest-best-practices

rstackjs/agent-skills

Set up, write, or review Rstest tests and configuration, including environments, mocks, snapshots, coverage, and CI.

rslib-best-practices

rstackjs/agent-skills

Configure, review, or troubleshoot Rslib library builds, output formats, declarations, dependency handling, and integrations.

rspack-best-practices

rstackjs/agent-skills

Configure, review, or troubleshoot Rspack builds, including loaders, CSS, assets, type checking, bundle optimization, and profiling.

rspress-best-practices

rstackjs/agent-skills

Configure, review, or troubleshoot Rspress v2 sites, including content conventions, navigation, MDX, assets, search, and deployment.

rspack-v2-upgrade

rstackjs/agent-skills

Use when upgrading a Rspack 1.x project to v2, including dependency and configuration updates.

AI & Agents517 installs

Popular ai & agents skills

View category →

google-agents-cli-publish

google/agents-cli

This skill should be used when the user wants to "publish an agent", "publish my ADK agent", "register an agent with Gemini", "publish to Gemini" (formerly Gemini Enterprise), or needs guidance on the agents-cli publish gemini command. Also use when the user wants to "manage agents in Agent Registry", "list/update/delete registered agents", or "register an MCP server". Covers ADK vs A2A registration modes, programmatic and interactive usage, flag reference, auto-detection from deployment metadata, Agent Registry fleet management, and troubleshooting. Part of the agents-cli skills suite. Do NOT use for deployment (use google-agents-cli-deploy).

AI & Agents513.9K installs

google-agents-cli-eval

google/agents-cli

This skill should be used when the user wants to "run an evaluation", "evaluate my agent", "evaluate my ADK agent", "write an eval dataset", "analyze eval failures", "compare eval results", "optimize agent", or needs guidance on the Gemini platform eval methodology and the Quality Flywheel. Covers eval metrics, dataset schema, LLM-as-judge scoring, and common failure causes. Applies to any agents-cli project, whatever framework the agent is written in. Do NOT use for agent API code patterns (ADK: use google-agents-cli-adk-code), deployment (use google-agents-cli-deploy), or project scaffolding (use google-agents-cli-scaffold).

AI & Agents513.5K installs

skill-creator

anthropics/skills

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.

OfficialAI & Agents416.4K installs

codex-pet

prime-skills/runcomfy-agent-skills

Codex Pet generator on RunComfy. Build a Codex-compatible Codex Pet spritesheet.webp + pet.json from a single reference image, drop it into ${CODEXHOME:-$HOME/.codex}/pets/ / and Codex picks it up as a custom Codex Pet next to the 8 built-ins. This skill produces the exact Codex Pet atlas Codex expects (1536x1872 PNG/WebP, 8 cols x 9 rows, 192x208 cells, 9 animation states — idle, running-right, running-left, waving, jumping, failed, waiting, running, review). Calls OpenAI GPT Image 2 edit ONCE via the local RunComfy CLI as runcomfy run openai/gpt-image-2/edit to produce a canonical Codex Pet pose, then assembles all 9 animation rows programmatically with ImageMagick micro-transforms — no Codex Pro, no $imagegen, no OPENAIAPIKEY required, only RUNCOMFYTOKEN. Triggers on "codex pet", "create codex pet", "make codex pet", "hatch codex pet", "/hatch image", "desktop pet codex", "codex pets", "spritesheet.webp", or any explicit ask to build a custom pet for OpenAI Codex.

AI & Agents402.1K installs

full-output-enforcement

leonxlnx/taste-skill

Overrides default LLM truncation behavior. Enforces complete code generation, bans placeholder patterns, and handles token-limit splits cleanly. Apply to any task requiring exhaustive, unabridged output.

AI & Agents401K installs

executing-plans

obra/superpowers

Use when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available

AI & Agents245.7K installs