Skills
98,067 packaged capabilities and playbooks for agents. Install any of them with one command.
Development
Detects anemic domain models, validates and refactors them into rich domain models, and enforces tactical DDD patterns (Entities, Value Objects, Aggregates, Domain Services, Domain Events). Use when the user asks to validate, review, or check domain models or DDD code; detect anemia; refactor domain objects; improve encapsulation; or mentions terms like "anemic model", "rich domain", "aggregate", "value object", "domain event", "ubiquitous language", "is this good DDD", "does this follow DDD", or "check my domain". Do NOT use for module or service boundary design, architectural decomposition, strategic DDD context mapping, or code outside the domain layer (DTOs, controllers, infrastructure adapters).
Find opportunities to improve web application code using TanStack libraries (Query, Table, Form, Router, etc.). Avoid man-with-hammer syndrome by applying TanStack after vanilla implementation works.
Prepare a self-contained task or agreed GitLab task set, apply shared review and scoped release planning, and produce a local Markdown plan with manual glab commands. Use when the user asks to prepare a task, an issue, or an agreed task set for execution or GitLab publication. Russian discovery terms: подготовить задачу, подготовить issue.
Acceptance-first TDD workflow for solo implementation work. Use when handling a coding change directly and you want to follow a disciplined sequence: read context, define acceptance criteria, write failing tests when feasible, implement, verify, and conclude with passed, failed, or skipped with reason.
Write technical tutorials, blog posts, and educational content with pedagogical structure. Covers concept explanations, how-to guides, deep dives, and developer education. Triggers on tutorial, technical article, blog post, explain concept, teach, educational content, developer guide.
Replaces the agent's reflex of piping a command straight through `tail -n N` or `head -n N` with `tee` to a file inside a `mktemp -d` directory followed by the same `tail` / `head`. The truncated view stays the same; the full output stays recoverable. Skipped for outputs expected to be very long.
Optimizes LLM inference with NVIDIA TensorRT for maximum throughput and lowest latency. Use for production deployment on NVIDIA GPUs (A100/H100), when you need 10-100x faster inference than PyTorch, or for serving models with quantization (FP8/INT4), in-flight batching, and multi-GPU scaling.
Build and optimize Three.js geometry with built-in primitives, BufferGeometry, BufferAttributes, indexed meshes, custom vertex data, lines and points, and instanced rendering. Use when creating meshes, editing vertices or UVs, generating shapes, reducing draw calls, or diagnosing geometry memory and culling behavior. Triggers on: Three.js geometry, BufferGeometry, BufferAttribute, vertices, indices, custom mesh, InstancedMesh, instancing, draw calls, line geometry, points, mesh optimization.
Load and manage Three.js assets with LoadingManager, GLTFLoader, DRACOLoader, KTX2Loader, texture/HDR loaders, async error handling, caching, progress reporting, and resource ownership. Use when importing GLB/GLTF models, textures, HDR assets, or compressed geometry, or when fixing slow, failed, duplicated, or leaky asset loads. Triggers on: Three.js loader, GLTFLoader, GLB, glTF, DRACOLoader, KTX2Loader, LoadingManager, texture loader, HDR loader, loading progress, asset cache.
Load, configure, and optimize Three.js textures: color space, wrapping, filtering, mipmaps, UV transforms, normal/AO/PBR maps, cube and HDR environments, data/canvas/video textures, render targets, texture memory, and disposal. Use when images look wrong, UVs repeat incorrectly, HDR/IBL is needed, or texture quality and GPU memory need tuning. Triggers on: Three.js texture, TextureLoader, UV mapping, color space, sRGB, mipmap, anisotropy, normal map, environment map, cubemap, HDR texture, render target, video texture.
[AUTO-INVOKE] MUST be invoked when designing or reviewing ERC20 token contracts that need flash loan protection. Covers token-level defense design patterns: cost tracking, same-block cooldown, progressive sell tax, minimum balance retention, EIP-7702 aware address checks, front-run protection, and referral binding. Trigger: any ERC20 token with anti-flash-loan, anti-bot, or tokenomics security design requirements.
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) with NVIDIA Megatron-Core using tensor, pipeline, sequence, context, and expert parallelism, plus FP8 on H100 and MoE configuration for Mixtral-style models. Covers choosing TP/PP/DP/CP sizes, launching distributed training, tuning micro-batch size, and fixing low MFU, out-of-memory errors, and diverging loss. Use when training models above 10B parameters on NVIDIA A100/H100 GPUs, when setting up 3D parallelism for a LLaMA-style model, when configuring expert parallelism for MoE training, or when trying to raise MFU toward 40-47%. Use PyTorch FSDP, DeepSpeed, or HuggingFace Accelerate instead for models under 70B or simpler setups.
Provides guidance for mechanistic interpretability research using TransformerLens to inspect and manipulate transformer internals via HookPoints and activation caching. Use when reverse-engineering model algorithms, studying attention patterns, or performing activation patching experiments.
Migrate an existing OTP / one-time-passcode flow from the Twilio Programmable Messaging API (self-built: generate code → send SMS → validate) to the Twilio Verify API (managed code generation, delivery, expiry, rate limiting, and Fraud Guard). Covers what Verify replaces, creating a Verify Service, replacing send + check logic, E.164 formatting with Lookup, bring-your-own custom codes, error-code translation, and a safe incremental cutover. Use when a developer is moving OTP/2FA delivery off Programmable Messaging onto Verify, or asks "should I use Verify instead of sending my own SMS codes?".
UiPath solution planner and designer. Always invoke for PDD / SDD files (`pdd.md`, `*-sdd.md`). Authors a Solution Design Document (SDD) from a Process Definition Document (PDD) or another process-knowledge source (Confluence page, SOP, BPMN model, meeting transcript), then derives the multi-skill, multi-project task list from an SDD, emitting live TaskCreate calls. Known-product single-project build, no PDD/SDD → that skill; ambiguous/“what should I build” → here. For `uip solution` lifecycle and `.uipx` → uipath-solution. For non-solution Orchestrator/IS/auth/traces → uipath-platform. For .xaml/.cs → uipath-rpa. For .flow → uipath-maestro-flow. For building/editing .bpmn → uipath-maestro-bpmn (a .bpmn as design input routes here). For agent.json/.py → uipath-agents. Sole case-SDD author. For caseplan.json build → uipath-maestro-case.
Systematic diagnosis and fixing of Unity bugs through code analysis. Classifies the bug, traces the execution path, proposes a fix with prevention. Triggers: /unity-debug, /debug, 'bug Unity', 'NullReferenceException', 'crash Unity', 'erreur Unity', 'ne marche pas', 'comportement bizarre', 'MissingComponentException', 'glitch visuel', 'physics bug'. Use when the user reports a bug, error, crash, or unexpected behavior in Unity. Produces a structured diagnosis: Symptom, Cause, Fix, Prevention.
Provides guidance on fine-tuning large language models with Unsloth, a library for faster, lower-memory training using LoRA and QLoRA, based on its official documentation (references/llms-txt.md). Use when setting up LoRA or QLoRA fine-tuning of an LLM with Unsloth, reducing GPU memory use during training, looking up Unsloth features or APIs, debugging Unsloth training code, or learning Unsloth best practices. Not for general model training with plain Hugging Face Transformers or PEFT without Unsloth.
Practices for when a dependency library has a bug or missing feature and you want to keep your downstream project running until an upstream PR is merged. Covers git ref pinning (branch HEAD SHA → merge SHA), deciding when to fall back to `link:`, handling SHAs, and pnpm v10 build script gating.
Symfony UX Map for interactive maps with Leaflet or Google Maps in Symfony. Covers markers, polygons, polylines, circles, info windows, and LiveComponent integration. Use when displaying maps, placing markers, drawing shapes or routes, handling map events, building store locators, using custom tile layers, or making maps reactive with LiveComponent. Code triggers: <twig:ux:map />, Map(), Point(), Marker(), Polygon(), Polyline(), Circle(), InfoWindow(), MapOptionsInterface, ComponentWithMapTrait, fitBoundsToMarkers, ux:map:marker:before-create, ux:map:connect, SYMFONY_UX_MAP_DSN. Also trigger when the user asks "how to display a map", "how to add markers", "how to draw a polygon on a map", "how to handle map click events", "how to make a reactive map", "how to use Leaflet in Symfony", "how to use Google Maps in Symfony", "map not showing", "map has zero height". Do NOT trigger for SVG icons (use ux-icons) or general frontend interactivity (use stimulus).
Turn any video into a cinematic scroll-driven landing page — Apple-style hero where scrolling progresses the visible frame through the video. Use when the user provides a video file and asks for "a landing page from this video", "scroll-frame website", "Apple-style scroll site", "hero that scrubs the video", "like the GitHub Copilot landing", or any equivalent. Extracts N evenly-spaced frames via ffmpeg, builds a self-contained HTML page with a sticky hero + JS scroll listener that swaps the visible frame as you scroll, plus headline, sections and CTA below. For YUV.AI projects, applies the yuv-design-system skill in Neon mode (pink/cyan/white, default for YUV.AI web) — Decks (purple/yellow) is reserved for slides only. For generic / non-YUV.AI projects, picks an appropriate palette per the source video. Output is one folder with `index.html` and a `frames/` directory — drop on any static host.
Vite 8+ development with Rolldown and Oxc. Use when configuring vite.config.ts, migrating rollup/esbuild-era configs, authoring Vite plugins with environments and hook filters, or troubleshooting SSR, Module Runner, and production build behavior in Vite 8. Triggers on vite, vite.config.ts, rolldownOptions, oxc, module runner, environments, hotUpdate, and lightning css.
MUST be used for Vue.js tasks. Strongly recommends Composition API with `<script setup>` and TypeScript as the standard approach. Covers Vue 3, SSR, Volar, vue-tsc. Load for any Vue, .vue files, Vue Router, Pinia, or Vite with Vue work. ALWAYS use Composition API unless the project explicitly requires Options API.
Logs and tracks machine learning experiments with Weights & Biases (W&B, wandb): metrics, hyperparameters, checkpoints, sweeps, artifacts with lineage, model registry, custom charts and shareable reports. Includes integrations for PyTorch, HuggingFace Transformers, PyTorch Lightning and Keras/TensorFlow. Use when logging training runs and comparing them across configurations, running automated hyperparameter sweeps, versioning datasets and models as artifacts, registering model versions, or sharing results with a team workspace. Use when setting up wandb.init for offline or unstable-connection training. Not for general data versioning outside ML runs.
Generate an interactive bash wizard that walks a person step by step through operations only a human can do. Use it to provision infrastructure, configure credentials or CI secrets, navigate unfamiliar third-party backends, or run a one-off migration/cutover. Don't use it for steps the agent can do itself.
Rails coding patterns derived from analysis of 37signals' Fizzy codebase. Use when writing Rails code in 37signals/Basecamp style or when asked to follow 37signals patterns. Covers controllers, models, views, Hotwire, testing, database, security, and team philosophy.
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.