Engineering Skills Directory
A curated index of 70+ skills from across the ecosystem — Anthropic, OpenAI, Google, React, Next.js, TypeScript, Node.js, and more. Each links directly to the official documentation.
70 skills total
Claude Tool Use
Give Claude access to external tools and APIs. Define tools as JSON schemas, Claude decides when to call them, and your code executes them — enabling real-world actions from any Claude model.
Claude Vision
Pass images directly to Claude for analysis, OCR, chart interpretation, UI feedback, and multimodal understanding. Supports JPEG, PNG, GIF, and WebP up to 5MB.
Prompt Caching
Cache large, repeated prompt prefixes (system prompts, long documents, tool definitions) to cut latency by up to 85% and reduce input token costs by 90% on cache hits.
Extended Thinking
Let Claude reason through complex problems step-by-step before answering. Claude 3.7 Sonnet and Claude 4 Opus produce higher-accuracy outputs on math, coding, and multi-step reasoning tasks.
Claude Batch API
Process thousands of Claude requests asynchronously at 50% of the cost of real-time requests. Ideal for bulk classification, data extraction, and offline evaluation pipelines.
Claude Streaming
Stream Claude responses token-by-token using server-sent events for real-time chat UIs. Dramatically improves perceived latency for conversational applications.
Computer Use
Give Claude control of a desktop environment — it can see the screen, move the cursor, click, type, and run terminal commands. Build autonomous computer agents with the beta API.
Files API
Upload files once and reference them across multiple Claude requests by file ID. Ideal for large PDFs, codebases, and datasets where re-uploading every request is wasteful.
Citations
Ask Claude to cite specific passages from provided documents when answering questions. Produces grounded, verifiable responses with source attribution for RAG applications.
Text Embeddings (Voyage AI)
Anthropic partners with Voyage AI to offer best-in-class embeddings for semantic search, clustering, and retrieval-augmented generation. Direct drop-in for RAG pipelines.
OpenAI Function Calling
Connect GPT-4o to external tools and APIs. Define functions as JSON schemas and the model returns structured calls your application executes, enabling reliable structured output extraction.
Structured Outputs
Guarantee GPT-4o responses conform exactly to a JSON schema you define. 100% schema adherence — no post-processing, no validation errors on production traffic.
Assistants API
Build stateful AI assistants with persistent threads, built-in file retrieval, code interpreter, and function calling. Manages conversation history and tool orchestration for you.
GPT-4o Vision
Feed images into GPT-4o for analysis, description, OCR, chart reading, and visual QA. Combine with text prompts for multimodal understanding in any application.
OpenAI Embeddings
Convert text to dense vector representations for semantic search, clustering, anomaly detection, and RAG. text-embedding-3-large offers industry-leading retrieval accuracy.
Fine-Tuning
Train GPT-4o-mini on your own examples to improve response style, format consistency, domain knowledge, and reduce prompt length for production workloads.
Realtime API
Build low-latency, bidirectional speech-to-speech applications with GPT-4o. Native audio input and output with interruption handling for voice agents and conversational AI.
Whisper Speech-to-Text
Transcribe and translate audio in 57 languages with best-in-class accuracy. Supports MP3, MP4, WAV, FLAC and more. Open-source model also available for self-hosting.
DALL-E 3 Image Generation
Generate high-quality images from text prompts with DALL-E 3 via API. Supports multiple sizes, quality levels, and natural language style control with strong prompt adherence.
OpenAI Batch API
Submit thousands of requests as a single batch file and receive results within 24 hours at 50% cost savings. Perfect for offline evaluation, bulk classification, and data enrichment.
Gemini API
Access Google's Gemini family of models (Pro, Flash, Ultra) via REST or SDK. Supports text, images, video, audio, and PDFs in a single multimodal request up to 1M tokens context.
Gemini Function Calling
Define functions as tools for Gemini to call based on user intent. Enables structured data extraction, API orchestration, and agentic workflows with the Gemini Pro models.
Gemini Long Context
Process up to 2 million tokens in a single prompt with Gemini 1.5 Pro — entire codebases, books, or hours of video. The largest context window of any publicly available model.
Gemini Code Execution
Let Gemini run Python code in a sandboxed environment during generation. Model writes code, executes it, reads the output, and iterates — enabling data analysis and math computation.
Google AI Studio
Browser-based IDE for prototyping with Gemini models. Test prompts, explore model capabilities, tune system instructions, and generate API keys without writing code first.
Gemini Grounding with Google Search
Connect Gemini responses to live Google Search results for up-to-date, factual answers. Reduces hallucination on time-sensitive queries with automatic source attribution.
React Server Components
Render components entirely on the server with zero client-side JavaScript. Access databases, file systems, and secrets directly in your component tree. The foundation of Next.js App Router.
useOptimistic
Apply optimistic UI updates instantly during async transitions. The component shows the new state immediately while the real state resolves, then auto-rolls back on error.
Suspense
Declaratively specify loading states while async data fetches. Wrap any component tree in `<Suspense fallback={...}>` and React handles the loading/ready transition automatically.
use() Hook
Unwrap Promises and Context inside render — the React 19 primitive that powers `async` server components and enables reading cached async values without useEffect.
React Context
Share state across a component tree without prop drilling. Use createContext, Provider, and useContext for theming, auth state, locale, and other global values.
Transitions & useTransition
Mark expensive state updates as non-urgent so React can interrupt them for higher-priority interactions. Keeps the UI responsive during heavy re-renders or data fetches.
Error Boundaries
Catch JavaScript errors anywhere in a component subtree and display a fallback UI instead of crashing the whole application. Essential for production resilience.
memo & useMemo
Skip expensive re-renders by memoizing component output and computed values. Use memo() to wrap components and useMemo() to cache the return value of expensive calculations.
useRef & forwardRef
Access DOM nodes and persist mutable values across renders without triggering re-renders. Essential for animations, focus management, measuring elements, and integrating non-React libraries.
Next.js App Router
The modern Next.js routing system built on React Server Components. Layouts, pages, loading, error, and route groups live in a file-system hierarchy under the `app/` directory.
Server Actions
Call server-side functions directly from client components and forms. No API route needed — just mark a function `'use server'` and call it from the client with progressive enhancement built-in.
Image Optimization
Automatic WebP/AVIF conversion, responsive srcset generation, lazy loading, and blur placeholders via the `<Image>` component. Eliminates the most common Core Web Vitals regressions.
Middleware
Run code before every request on the edge — auth checks, redirects, rewrites, A/B testing, and header injection — with sub-millisecond latency and global distribution.
Parallel Routes
Render multiple pages simultaneously in the same layout using named slots (`@slot`). Powers dashboards, modals, and split-pane views where independent sections load independently.
Metadata API
Define static and dynamic metadata (title, description, OG images, robots, canonical URLs) from layout and page files. Next.js generates all `<head>` tags automatically.
Incremental Static Regeneration
Rebuild static pages on-demand or on a time interval without a full redeploy. `revalidate` and `revalidatePath` let you keep static pages fresh while serving them at CDN speed.
Route Handlers
Create REST API endpoints inside the App Router using `route.ts` files. Supports all HTTP methods, streaming responses, CORS headers, and edge runtime deployment.
TypeScript Generics
Write reusable, type-safe code that works across multiple types. Generics are the foundation of utility types, React component props, and flexible API client patterns.
Utility Types
Built-in type transformations: Partial, Required, Pick, Omit, ReturnType, Awaited, and more. Compose complex types from simpler ones without duplicating type definitions.
Discriminated Unions
Model mutually exclusive states by combining union types with a shared discriminant field. The TypeScript pattern for exhaustive type narrowing — eliminates entire classes of runtime bugs.
Template Literal Types
Build string types programmatically using template literal syntax. Powers strongly-typed CSS class name generators, event names, API path builders, and i18n key systems.
satisfies Operator
Validate that an expression matches a type without widening it. Catch mismatches at definition time while preserving the most specific inferred type for downstream use.
Zod Schema Validation
Define schemas once, infer TypeScript types automatically, and validate at runtime. The standard for type-safe form validation, API response parsing, and environment variable checking.
Node.js Streams
Process data piece-by-piece without buffering the entire payload in memory. Essential for file processing, HTTP responses, log pipelines, and transforming large datasets efficiently.
Worker Threads
Run CPU-intensive JavaScript in parallel threads without blocking the event loop. Each worker has its own V8 instance and communicates via message passing and SharedArrayBuffer.
ES Modules in Node.js
Use native ESM (`import`/`export`) in Node.js with `.mjs` files or `"type": "module"` in package.json. Enables top-level await, tree-shaking, and alignment with browser modules.
Node.js Crypto
Built-in cryptographic functions for hashing, HMAC, encryption/decryption, key generation, and secure random bytes. No third-party library needed for common security operations.
Prisma ORM
Type-safe database access for TypeScript. Define your schema in Prisma Schema Language, run migrations, and query with a fully-typed client that eliminates SQL injection and runtime shape errors.
Drizzle ORM
Lightweight TypeScript ORM with a SQL-like query API. Define schemas in TypeScript, get full type inference, and generate migrations — with zero runtime overhead and serverless-first design.
Supabase
Postgres as a service with auto-generated REST and real-time APIs, authentication, storage, and edge functions. The open-source Firebase alternative with full Postgres capabilities.
Redis Caching
Cache expensive database queries, session data, rate limit counters, and computed values in sub-millisecond in-memory storage. Use Upstash for serverless-compatible Redis via HTTP.
Vector Databases
Store and query high-dimensional embeddings for semantic search and RAG. Options include Pinecone, Weaviate, Chroma, and pgvector (Postgres extension) — each with different latency/cost tradeoffs.
Tailwind CSS v4
Utility-first CSS framework with a CSS-first configuration model in v4. Define design tokens with `@theme`, use arbitrary values, and build entire UIs without leaving HTML.
CSS Grid Layout
Two-dimensional layout system for building complex, responsive page shells. Define rows and columns with `grid-template`, place items with `grid-area`, and avoid absolutely all absolute positioning hacks.
CSS Container Queries
Apply styles based on the size of a container element, not the viewport. Components that adapt to their layout context — sidebars, card grids, and embeddable widgets — without JavaScript.
View Transitions API
Animate between page states or DOM updates with CSS-powered transitions — no animation library needed. The browser captures before/after snapshots and animates between them automatically.
Vitest
Vite-native test runner with Jest-compatible API. Near-instant HMR for tests, TypeScript and ESM support out of the box, built-in code coverage, and snapshot testing.
Playwright E2E Testing
End-to-end browser testing for Chromium, Firefox, and WebKit in the same test. Auto-waits eliminate flaky sleeps, trace viewer pinpoints failures, and API testing needs no browser.
React Testing Library
Test React components the way users interact with them — by accessible role, label, and text. Promotes accessible UIs as a side effect of good testing practices.
Mock Service Worker (MSW)
Intercept network requests at the service worker level for realistic API mocking. Same mock definitions work in tests (Node.js) and the browser — no more diverging mock/prod behavior.
Vercel Edge Functions
Deploy JavaScript/TypeScript functions to Vercel's edge network — 100+ global locations, sub-millisecond cold starts, and access to the Web Fetch API, crypto, and streaming responses.
GitHub Actions CI/CD
Automate build, test, and deployment pipelines directly from your repository. YAML-defined workflows trigger on push, PR, schedule, or manual dispatch with 10,000+ marketplace actions.
Docker Containerization
Package applications and dependencies into portable containers that run identically in development, CI, and production. Multi-stage builds minimize image size for Node.js and Next.js apps.
Cloudflare Workers
Deploy JavaScript/TypeScript to Cloudflare's 300+ edge locations. Access KV storage, D1 (SQLite), R2 (object storage), and Queues — a complete serverless platform at the edge.