Skills
98,873 packaged capabilities and playbooks for agents. Install any of them with one command.
Development
Provides guidance for PyTorch-native agentic RL using torchforge, Meta's library separating infra from algorithms. Use when you want clean RL abstractions, easy algorithm experimentation, or scalable training with Monarch and TorchTitan.
Provides guidance for PyTorch-native agentic RL using torchforge, Meta's library separating infra from algorithms. Use when you want clean RL abstractions, easy algorithm experimentation, or scalable training with Monarch and TorchTitan.
Trains large language models (2B-462B parameters) using NVIDIA Megatron-Core with advanced parallelism strategies. Use when training models >1B parameters, need maximum GPU efficiency (47% MFU on H100), or require tensor/pipeline/sequence/context/expert parallelism. Production-ready framework used for Nemotron, LLaMA, DeepSeek.
Emulated Twilio REST APIs for local development and testing. Use when the user needs to test Twilio Messaging, Verify, Voice, phone numbers, webhooks, status callbacks, inbound SMS simulation, or Twilio SDK integrations without hitting the real Twilio service.
Master TypeScript's advanced type system including generics, conditional types, mapped types, template literals, and utility types for building type-safe applications. Use when implementing complex type logic, creating reusable type utilities, or ensuring compile-time type safety in TypeScript projects.
Use when creating, editing, or reviewing TYPO3 extension documentation (Documentation/*.rst, guides.xml, README.md, XLF translations), rendering docs with Docker, using TYPO3 RST directives, adding screenshots, deploying to docs.typo3.org, improve docs, fix documentation, or XLIFF 2-space indentation (TYPO3 v14+).
Use when planning, building, designing, code review, refactor, fixing, improving, optimizing, enhancing, or checking UI/UX work. Covers product patterns, landing pages, chart choices, visual styles, palettes, typography, UX rules, and 9 stacks (React, Next.js, Vue, Svelte, SwiftUI, React Native, Flutter, Tailwind, shadcn/ui). Projects: website, landing page, dashboard, admin panel, e-commerce, SaaS, portfolio, blog, mobile app, .html, .tsx, .vue, .svelte. Elements: button, modal, navbar, sidebar, card, table, form, chart. Topics: color palette, accessibility, animation, layout, typography, font pairing, spacing, hover, shadow, gradient.
Integrates uni-ad monetization into uni-app applications including banner, interstitial, rewarded video, and native ad types with event handling and placement configuration. Use when the user needs to add ad monetization, configure ad types, handle ad load/show/close/error events, or optimize ad revenue in a uni-app project.
Unity level design-to-code translation. Trigger & event architecture, encounter design contracts, checkpoint & respawn, cinematic & scripted sequences, environmental storytelling hooks, level streaming & loading seams. DESIGN INTENT format: INTENT/WRONG/RIGHT/SCAFFOLD/DESIGN HOOK. Based on Unity 6.3 LTS.
Unity 6 multiplayer and networking guide. Use when building multiplayer games, working with Netcode for GameObjects, NetworkManager, NetworkObject, NetworkBehaviour, RPCs (ServerRpc, ClientRpc), NetworkVariables, or Unity multiplayer services (Relay, Lobby). Covers client-server architecture, state synchronization, and scene management. Based on Unity 6.3 LTS documentation.
Unity physics query correctness patterns. Catches common mistakes with Raycast, SphereCast, OverlapSphere, NonAlloc allocation, LayerMask construction, trigger interaction, hit ordering, and query type selection. PATTERN format: WHEN/WRONG/RIGHT/GOTCHA. Based on Unity 6.3 LTS.
Jujutsu (jj) reference for anything past describe/new/commit/push -- the non-interactive split form, which revision a bookmark must target before pushing, absorb, evolog, oplog recovery, immutable heads, revsets. Load whenever a git habit could produce a wrong or editor-blocking jj command.
Master the uv package manager for fast Python dependency management, virtual environments, and modern Python project workflows. Use when setting up Python projects, managing dependencies, or optimizing Python development workflows with uv.
Visual Ralph orchestration for frontend UI from generated references, static references, or live URL targets, using $ultragoal with built-in visual verdict and pixel-diff evidence until the implementation matches and leaves a reproducible design system.
Configure @weapp-tailwindcss/lynx 0.3.2 and Tailwind CSS 4 for ReactLynx + Rspeedy, covering the Rspeedy plugin, Lynx native CSS, static className, static theme variables, arbitrary values, encoder warnings, and iOS/Android runtime verification. Use for ReactLynx, Rspeedy, Lynx native CSS, arbitrary value compatibility, or Lynx bundle styling; not for React Native, uni-app App, Web, or regular mini program builds.
Durable, type-safe queue workers with @falcondev-oss/workflow on Redis. Use when working with WorkflowNamespace / createWorkflow / defineWorkflow, handle() and returns<T>(), step.do / step.wait / step.waitUntil, workflow.run / runIn / runAt / work(), job.wait(), upsertSchedule cron schedules, groupId ordering and numeric priority, or the createRedis connection setup.
Design durable workflows with Temporal for distributed systems. Covers workflow vs. activity separation, saga patterns, state management, and determinism constraints. Use when building long-running processes, distributed transactions, or microservice orchestration
Audit a frontend against WCAG and fix what fails: semantic HTML, keyboard navigation, focus management, screen-reader support (ARIA), color contrast, forms, and media alternatives. Use when the user asks 'is this accessible', 'make this WCAG/AA compliant', 'add keyboard navigation', 'fix screen reader support', 'a11y audit', 'we got an accessibility complaint/legal notice', or whenever building/reviewing UI components with interactive behavior — even if accessibility isn't mentioned. Produces a prioritized accessibility audit (issue × WCAG criterion × severity × fix) plus the implemented fixes in code.
Use this skill to gate Phoenix LiveView realtime UIs with a pass/fail adversarial review — a single blind reviewer subagent judges a diff or file set against 32 decidable rules covering state ownership and lifecycle (patch vs remount, callback load placement, connected-mount guards, form recovery), realtime data flow (broadcast placement, scoped topics, presence mechanisms), async responsiveness (blocking external calls, socket-copying closures, lifecycle-owned tasks, rendered failure states), render and wire efficiency (the constructs that silently disable HEEx change tracking), streams (growing collections in assigns, the stream DOM contract, bounded infinite scroll), component and context boundaries, client trust (per-event authorization, scoped lookups, live_session boundaries, revocation disconnects), and mechanism-presence interaction feedback (JS commands, in-flight feedback, debounce, hook contracts, overlay focus). Verdicts only, never fixes.
Use this skill to gate TanStack Start plus TypeScript web-app changes with a pass/fail adversarial review — a single blind reviewer subagent judges a diff or file set against 22 decidable rules covering client/server boundary leaks, server-function and server-route usage, auth and security, SSR data loading, boundary type safety, and compiler config. Trigger it before merging TanStack Start work, when asked to gate, adversarially review, or pass/fail a Start or TanStack codebase, or as a final check on agent-authored Start features. It renders verdicts only and never fixes the work; for teaching-style review feedback use a distillation skill instead.
Build AI agents on Cloudflare Workers using the Agents SDK. Load when creating stateful agents, durable workflows, real-time WebSocket apps, scheduled tasks, MCP servers, or chat applications. Covers Agent class, state management, callable RPC, Workflows integration, and React hooks.
Generate working TypeScript/JavaScript boilerplate for SAP GenAI Hub using @sap-ai-sdk/orchestration, @sap-ai-sdk/foundation-models, @sap-ai-sdk/openai (native OpenAI SDK), @sap-ai-sdk/langchain (LangChain), @sap-ai-sdk/document-grounding (RAG), @sap-ai-sdk/prompt-registry, @sap-ai-sdk/llm-batch (batch jobs), @sap-ai-sdk/rpt (tabular prediction), or @sap-ai-sdk/context-registry (data destinations, tabular artifacts, scenario configurations). WHEN: user asks for TypeScript, JavaScript, Node.js, or JS/TS code that calls a GenAI Hub model, wants a TypeScript starter project, asks how to use @sap-ai-sdk in their app, asks how to use LangChain (JS/TS) with SAP AI Core, or wants JS/TS for RAG/document-grounding, prompt templates, batch LLM jobs, tabular (RPT) prediction, or managing data destinations / tabular artifacts / scenario configurations via the Context Registry. DO NOT USE FOR: Java code — use `aicore-sdk-java`.
Creates performant CSS-first web animations and chooses the right feedback for an interaction: transitions, View Transitions, scroll-driven motion, Motion for React, reduced motion, and perceived performance. Triggers on “animate this,” “hover effect,” “feel faster,” “skeleton loader,” and “/animations.”
AVFoundation / AVKit / PhotoKit reference. AVPlayer, AVPlayerViewController, VideoPlayer, AVPlayerItem, AVMutableComposition, AVCaptureSession, AVCaptureDevice, AVCapturePhotoOutput, AVCaptureDeviceDirectionCoordinator, CameraCaptureAccessory (iPhone Duo), AVAudioEngine, AVAudioSession, AVAudioPlayer, AVSpeechSynthesizer, PHAsset, PHImageManager, PhotosPicker, PHPhotoLibrary.
Autonomously improve a real artifact (code, training recipe, agent harness, data pipeline, prompt) against an objective and an evaluator, using Hypothesis Tree Refinement (HTR) from the Arbor paper. Use this whenever someone wants to iteratively optimize something over many experiments without overfitting — e.g. "get my model's eval score up", "improve this agent/harness", "tune this pipeline", "beat the baseline on this benchmark", "run a search over approaches and keep the best", "do an MLE-bench / Kaggle-style optimization", or any long-horizon "make this artifact better and don't just memorize the dev set" task. Trigger it even when the user doesn't say "Arbor" or "hypothesis tree" but describes repeated experiment-and-evaluate loops, branching exploration of competing ideas, or worries about a dev/test gap. Runs Claude itself as the coordinator with subagent executors in isolated git worktrees; for the standalone `arbor` CLI tool see references/arbor-upstream.md.
Extracts key architectural decisions from technical-requirements.yaml and formalizes them as individual ADR markdown files in an adrs/ directory. Each ADR captures context, the decision made, alternatives considered, and consequences — creating a durable record of why the system is built the way it is.
Use auto-run-agent to automatically analyze, evaluate, design, and optimize a target project. The full workflow includes: explore the project's current state → identify problems → design a task list → create the runner environment → start autonomous execution. Suitable for any project that needs batch optimization.
[Extends reviewer] Senior Backend Code Reviewer with 12+ years Java/Kotlin experience. Use when reviewing Java/Spring/Kotlin code, checking code quality and style, identifying code smells and anti-patterns, verifying security practices, ensuring test coverage, or running static analysis tools (Checkstyle, SpotBugs, SonarQube). Invoke alongside /rev for backend reviews.
Author a Bounded backend: policy.json rules and invariants (rollingSum/windowSum/flowBound/conserve/tenantTag/tenantEdge/bound), functions (ctx.user/ctx.bounded/ctx.ai/ ctx.services/ctx.secrets), the actor and identity model (@user, runAs/actAs, @origin, service keys, reserved identity sets), data and queries, realtime/live rooms, and policy tests. Use when writing or changing server-side Bounded logic, policies, or the rules that govern who can do what. Part of the Bounded skill family; see the bounded skill to route across frontend, deploy, onchain, and teams.
Add onchain to a Bounded app: Solana collections, embedded non-custodial wallets for email/social users (@user.address, Turnkey by default), client- and server-signed transactions, DEX/perps trading patterns, and crypto payments (accept USDC/crypto non-custodially). Use for wallet, token, on-chain transaction, or crypto payment work. Part of the Bounded skill family; policy and the actor model live in bounded-backend.
Turn a plain-language request into a working, tested Agenta agent through the Agenta API. Use when the user wants to build, create, schedule, or test an Agenta agent from their coding agent (Claude Code, Codex, Cursor). Writes one config and runs a few bundled shell scripts. Triggers on "build an Agenta agent", "create an agent on Agenta", "schedule an Agenta agent", or wiring an agent to a tool like GitHub or Slack via Agenta.
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.