LLM Skill Registry
Package AI work into
installable skills.
Reusable, versioned skill packages for LLM agents. Web design, mobile, backend, databases — discover the package contract, inspect the details, and install the skills your project needs.
Browse skill packages
Each skill defines instructions, examples, guardrails, npm package metadata, and tool recipes — ready to package into a project-level agent capability.
Full-page layout, component library, typography scales, color tokens, and responsive grid rules for modern web UIs.
Scaffold a React + Vite app with routing, state management (Zustand), ESLint, Prettier, and Tailwind — production-ready in seconds.
Hero sections, feature grids, testimonials, pricing tables, and CTA patterns — copy, adjust, ship.
Scaffold Flutter project with folder structure, state management (Riverpod/Bloc), routing (GoRouter), environment configs, and Firebase init.
Pre-built Flutter widgets: auth screens, dashboards, bottom nav, modals, form validation, and dark-mode theming with Material 3.
Express + TypeScript API with JWT auth, Zod validation, rate limiting, error handling middleware, and OpenAPI docs.
Python FastAPI scaffold with Pydantic models, async DB (SQLAlchemy 2.0), Alembic migrations, OAuth2, and pytest suite.
Firebase project initialization with Auth, Firestore, Storage, Functions, and App Check — with security rules templates.
SQL injection prevention, parameterized queries, row-level security (Postgres RLS), connection pooling, secrets management, and audit logging.
Firestore and Storage security rules patterns — user-owned data, role-based access, admin overrides, and validation rules with test coverage.
Document modeling patterns, indexing strategies, aggregation pipelines, Atlas Search setup, and data validation with Mongoose schemas.
Full auth flow with JWT refresh tokens, OAuth providers (Google, GitHub), session management, CSRF protection, and MFA scaffolding.
Create, test, publish, reuse.
Skills are living packages — they get sharper, safer, and more capable with each version.
Create a SKILL.md with purpose, triggers, constraints, examples, and tool recipes. Commit to your GitHub repo.
Tag a release on GitHub — CI publishes to the npm registry automatically. Your skill is discoverable instantly.
Install to your local project with one command. Skills auto-wire into your LLM agent's context on startup.
How users install a skill in their project.
Ship each skill as a small npm package. The package contains the skill instructions, examples, manifest, and optional helper code; the CLI copies it into the user's project.
Package anatomy
Every skill package should keep the same predictable contract so agents and tooling can read it without guessing.
- 1SKILL.md is the main instruction file: purpose, triggers, constraints, steps, examples, and quality checks.
- 2skill.json stores machine-readable metadata: name, version, category, entry file, supported agents, and files to copy.
- 3examples/ includes prompts, expected outputs, project templates, and before/after references for the agent.
- 4package.json publishes the package to npm and exposes optional JS helpers for apps that want runtime access.
Install flow
Get started in 30 seconds.
Install the CLI globally, then browse and install skills right in your terminal.
Install the CLI
GitHub-linked skills
Every skill is tied to a GitHub repo. Star, fork, or open a PR to improve any skill directly.
05 / Contribute
Build a skill,
publish your work.
Submit your skill package to the registry. Every skill gets a GitHub repo, npm package, and contributor profile.
v0.1 — Open Beta
run
ship