Official, code-first Agent Skills for AI coding agents (Claude Code, Cursor, Codex, Copilot, Gemini CLI and any other client of the open standard) working on NativeScript projects. Each skill is a folder with a SKILL.md: a one-line description the agent matches on, then concrete code that has actually run, plus the pitfalls around it.
Skills are organised by area so the set can grow across platforms, frontend flavors and tooling without becoming a flat list:
skills/
├── core/ @nativescript/core fundamentals, runtime quirks, custom views
├── ios/ Apple frameworks from TypeScript
├── android/ android.* from TypeScript, Java interop
├── frontend/ angular/ · vue/ · react/ · svelte/ · solid/
├── ui/ cross-platform UI & motion patterns
├── device/ device capabilities without plugins
├── testing/ tests, simulator/emulator automation, verification
├── tooling/ CLI, bundlers, assets, CI, release
└── plugins/ authoring and packaging NativeScript plugins
Every install door below flattens this to <skills-dir>/<name>/ — the category is for humans and for the listing pages; agents match on the skill name and description.
| You use… | Install | Update |
|---|---|---|
| Any agent (Claude Code, Cursor, Codex, Copilot, Gemini, Windsurf, …) | npx skills add NativeScript/skills (add --copy to vendor real files, -g for global) |
npx skills update |
| Claude Code plugin | /plugin marketplace add NativeScript/skills then /plugin install nativescript-skills@nativescript |
/plugin update nativescript-skills@nativescript |
| npm, vendored & pinned in your repo | npx @nativescript/skills install --agent claude,agents (writes .nativescript-skills.json; --only ios,frontend/angular filters by category or name) |
npx @nativescript/skills@latest update (check shows stale/modified files) |
| TanStack Intent | npm i -D @nativescript/skills && npx @tanstack/intent@latest install |
npm update @nativescript/skills |
| By hand | cp -R skills/<category>/<name> <project>/.claude/skills/ (or ~/.claude/skills/) |
cp again |
Details, trade-offs and the release process: docs/DISTRIBUTION.md.
0 skills in 0 categories. Generated by npm run sync from the skills/ tree and each category's README.
No skills yet — see CONTRIBUTING.md to add the first one.
npm run lint— Agent Skills spec lint plus the tree rules (depth, naming, uniqueness, category READMEs, one eval task per skill);npm run validate— TanStack Intent's validator;npm run sync:check— generated manifests match the tree. All run in CI.evals/— a skillgrade suite: one realistic task per skill, executable graders where the code can run headless, and a check that the agent actually discovered the skill among all of them. See evals/README.md; the latest local run is summarised in docs/EVALUATION.md.
npm run new <category> <ns-name> scaffolds skills/<category>/<ns-name>/SKILL.md and prints the eval task stub to add to evals/tasks.mjs. Then npm run sync and npm run check. The full guide — where a skill goes, naming, what a good SKILL.md and a good eval task look like — is in CONTRIBUTING.md.
Apache-2.0 © NativeScript