Skills
96,981 packaged capabilities and playbooks for agents. Install any of them with one command.
Development
Use this skill when auditing, reviewing, or fixing thread-safety issues in ArkCompiler Runtime Core, especially ArkTS-Sta ETS stdlib and plugin code under static_core/plugins/ets. It covers static mutable state, singleton initialization, shared maps/caches/counters/timers, taskpool/EAWorker concurrency, TSAN follow-up, and concurrent test design.
OpenRouter unified-LLM-API expertise. Use when code POSTs to `openrouter.ai/api/v1`, imports `@openrouter/sdk`, `openrouter` (Python), or `@openrouter/agent`; when the OpenAI SDK is pointed at `https://openrouter.ai/api/v1`; when the user asks about routing across hundreds of LLMs (OpenAI, Anthropic, Google, DeepSeek, Meta, xAI, Mistral, etc.) through one endpoint; when working with provider routing, model fallbacks (`models` array), the Auto Router (`openrouter/auto`), `:nitro`/`:floor`/`:online` model suffixes, presets (`@preset/...`), prompt caching with `cache_control`, structured outputs, tool calling, the `openrouter:web_search` server tool, reasoning tokens, multimodal (image/PDF/audio/video) inputs, BYOK provider keys, OAuth PKCE, the `openrouter:` plugins (`web`, `file-parser`, `response-healing`, `context-compression`), or generation/usage accounting via `/api/v1/generation` and `/api/v1/key`.
Generate initial OpenWiki documentation (openwiki/ directory) for the current repository — quickstart entrypoint plus focused section pages, grounded in source and git evidence. Use when the user wants to initialize, bootstrap, or (re)generate a repository wiki from scratch.
Main entry point for working with Opper (https://opper.ai), the AI gateway for agents. Use when the user mentions Opper without choosing a surface: "set up Opper", "try Opper", "build with Opper", "migrate to Opper", or "help me with Opper". Explore context and route to opper-mcp for agent-assisted setup, platform operations, and model tests; opper-cli for explicit CLI and shell workflows or launching coding agents through Opper; opper-api for application code (the OpenAI, Anthropic or Google SDKs and agent frameworks on the compat endpoints), raw HTTP, and platform concepts; or opper-multimodal for media, files, and realtime voice. Own setup, verification, and follow-up.
Optimize and compress 3D assets for the web — shrink glTF/GLB models and textures using gltf-transform, gltfpack/meshopt, Draco, and KTX2/Basis, and optionally convert a model to an R3F component with gltfjsx. Use when a model is too heavy, loads slowly, or needs to be web-ready before adding it to a Three.js/R3F scene.
Run and configure oxfmt — the high-performance JavaScript/TypeScript formatter built on the Oxc compiler stack, ~30x faster than Prettier. Use this skill whenever working in a project that has oxfmt installed (check for `oxfmt` in package.json devDependencies or an `.oxfmtrc.json` / `.oxfmtrc.jsonc` config file). This includes when you need to format code after making changes, set up oxfmt in a new project, migrate from Prettier, configure formatting rules, set up editor integration (VS Code, Neovim, Zed, JetBrains), or add formatting checks to CI/CD pipelines. Also use when the user mentions "oxfmt", "oxc formatter", "format with oxc", or asks about Prettier alternatives with better performance.
Hardware-agnostic quantum ML framework with automatic differentiation. Use when training quantum circuits via gradients, building hybrid quantum-classical models, or needing device portability across IBM/Google/Rigetti/IonQ. Best for variational algorithms (VQE, QAOA), quantum neural networks, and integration with PyTorch or JAX. For hardware-specific optimizations use qiskit (IBM) or cirq (Google); for open quantum systems use qutip.
Official NVIDIA-authored guidance for PhysicsNeMo ShardTensor domain parallelism — integrate domain parallelism into training/inference scripts (new or existing) with DDP or FSDP2, write and register shard patches to enable new layers/ops, and bootstrap multi-GPU correctness tests. Use when working with ShardTensor, scatter_tensor, domain parallelism, sequence/spatial sharding, ring attention, DeviceMesh + DDP/FSDP2 hybrid parallelism, or physicsnemo.domain_parallel. Do NOT use for generic PyTorch DDP/FSDP setup without domain parallelism, picking a PhysicsNeMo model or example (use physicsnemo-discover), or non-distributed training questions
Use when handling creation, version management, instantiation, permission control, PAC templates, and store integration for BK-CI pipeline templates. Prefer this when users mention template reuse, template versions, batch instantiation, template migration, store templates, or PAC templates.
Use this skill when working with Pixi workspaces, pixi.toml, Pixi-enabled pyproject.toml, or pixi.lock. Covers workspace initialization and import; conda/PyPI dependencies; features and environments; task execution; lock policies; troubleshooting; and Pixi global tools.
Write a Guided Review (a chaptered walkthrough of a changeset with the real diffs inline) and turn it into one portable HTML file with `plannotator guide export`. Use when someone asks for a guide, walkthrough, tour, or explainer of a diff, branch, PR, or commit, or wants a shareable review guide. Needs git and the plannotator CLI.
Browser automation with Playwright for Python. Use when testing websites, taking screenshots, filling forms, scraping web content, or automating browser interactions. Triggers on browser, web testing, screenshots, selenium, puppeteer, or playwright.
Manage stacked PRs using whichever stacking tool is installed: Graphite (`gt`), git-town (`git town`), or GitHub's native `gh stack` extension. Detects which is available in the current repo, then drives the matching one via its per-tool reference (`references/gt.md`, `references/git-town.md`, or `references/gh-stack.md`). Use when the user asks to "create a stack", "stack a branch", "restack", "submit a stacked PR", "rebase the stack", "sync the stack", "track this branch", "the bottom PR merged, what now", "clean up after a stack merge", or invokes /pr-stack. If none is available, stop and tell the user — never fall back to ad-hoc `git push` chains. Do NOT use for staging or committing — hand off to /commit. Do NOT use for PR review traffic, comments, or CI checks — hand off to /gh.
This skill should be used when creating, validating, or converting Product Requirements Documents (PRDs) to LLM-native format. Use this skill when the user asks to write a PRD, review a PRD for compliance, convert an existing PRD to machine-readable format, or ensure PRD quality for AI agent consumption. This skill is optimized for users working with agentic coding frameworks who need strict requirement specifications to prevent context poisoning, hallucination, and mesa-optimization.
Prepare Sprint 0 (walking skeleton) for a project based on its Tech Design and ADRs: classify ADRs as cross-cutting or story-specific, define the route ownership map by role, write the skeleton's Definition of Done, propose a plan, and stop for human approval. Use when the LEAD (or user) needs to prepare, plan, or initialize a project's Sprint 0 / walking skeleton before any user story is built.
Project documentation structure creation assistant. Acting as a senior product manager, it uses multi-turn dialogue to deeply understand project requirements, analyse their soundness and offer best-practice suggestions, and finally generate a complete technical documentation system. **When to use:** - A new project starts and a documentation system needs to be set up (recommended first step) - An existing project lacks standard documentation and needs to be completed - Code already exists but documentation is missing, and documentation needs to be backfilled from the code - Professional product and architecture analysis advice is needed **Two supported working modes:** - Mode A: new project documentation creation (from requirements to documentation) - Mode B: code-based documentation backfill (from code to documentation) **Relationship with other skills:** - This skill creates complete documentation (PRD, SAD, development guide, etc.) - project-workflow executes development tasks based on the documentation - Recommended flow: project-docs-setup → project-workflow **Triggers:** - "创建项目文档" / "初始化项目文档" / "设置项目文档" - "根据代码生成文档" / "回补项目文档" / "补充项目文档" - Directly describing project requirements, e.g. "帮我创建一个在线教育平台的文档"
PyQt6 UI design assistant that generates and refines modern, professional desktop interfaces using a consistent design system (DESIGN.md tokens). Use this skill whenever the user wants to: build or improve a PyQt6 GUI, write QSS stylesheets, add light/dark theme support, generate sidebar navigation, data tables, forms, cards, modals, or any other PyQt6 component. Also trigger this skill when the user uploads or mentions existing PyQt6 code that needs a "refresh", "redesign", "modern look", or "professional UI". Use even for partial UI tasks like "style my button" or "make a dark sidebar" — any PyQt6 styling task benefits from this skill.
Use when defining or evolving public interfaces, schema boundaries, or pydantic usage in Python. Also use when annotations are missing on public APIs, pydantic models appear everywhere instead of at trust boundaries, contract changes lack migration guidance, or Any/object types are overused across module boundaries.
Set up a Quiver bolt's three-layer bootstrap stack (public bootstrap repo, private dotfiles repo, private workspace repo, an age key, and an SSH key) and create the bolt. Use when a user wants to configure a new Quiver persona or dotfiles stack, onboard to Quiver, or spin up a bolt dev environment from scratch. Works on macOS, Linux, and Windows.
Use when planning or reviewing production database migrations, adding columns, indexes, constraints, backfills, renames, table rewrites, or concurrent operations. Covers phased rollouts, lock behavior, rollback strategy, strong_migrations compliance, and deployment ordering for schema changes.
Runs autonomous loop fetching stories from GitHub Issues. Implements and closes issues as done. Triggers on "loop through my PRDs", "work on my issues", "start the autonomous loop", "implement my PRDs", or requests to work through GitHub issues autonomously.
Route requests to the focused Randroid development workflows and explain which capability to use. Use when the user asks what Randroid skills are available, invokes the Randroid namespace without naming a workflow, or needs help choosing among loop execution, PR feedback handling, VibeReview playtesting, visual de-slopping, and gameplay interaction or feedback work.
Expert implementation of @vis.gl/react-google-maps library for Google Maps in React. Use when building maps, markers, pins, infowindows, places autocomplete, geocoding, draggable markers, polygons, circles, polylines, drawing tools, or any Google Maps JavaScript API integration in React/Next.js applications.
Build type-safe validated forms using React Hook Form v7 and Zod v4. Single schema works on client and server with full TypeScript inference via z.infer. Use when building forms, multi-step wizards, or fixing uncontrolled warnings, resolver errors, useFieldArray issues, performance problems with large forms.
React component patterns and architectural best practices. Covers compound components, render props, higher-order components, custom hooks, composition, state machines, and container/presentational patterns. USE WHEN: user mentions "compound components", "render props", "HOC", "higher-order component", "composition pattern", "component architecture", asks about "reusable components", "component patterns", "slots pattern", "polymorphic components" DO NOT USE FOR: Basic React components - use `react` skill instead, hooks implementation - use `react-hooks` skill instead, performance patterns - use `react-performance` skill instead
Advise on and review React code through the lenses of "Advanced React" (Makarevich) — unnecessary re-renders, composition-before-memoization, the memo/useMemo/useCallback traps, Context performance, refs & stale closures, debounce/throttle, useLayoutEffect flicker, portals & stacking context, data-fetching waterfalls & race conditions, and error boundaries. Use when writing or reviewing a React component/hook/context/provider, when chasing "why does this re-render" or "should I memoize this", or when dealing with stale state in a callback, UI flicker, modal clipping, fetch race conditions, or error handling. Detects the target repo's React version and state/UI libraries before applying version-specific rules. Distinct from architecture-lens (architecture) and design-patterns (GoF).
All-in-one Redis expertise covering data modeling and key naming, client connections (pooling, pipelining, client-side caching, timeouts), clustering and replication, Redis Search (FT.CREATE, vector and hybrid search, RAG pipelines), observability and incident triage, production security hardening, semantic caching for LLM responses via LangCache, and agent memory via Redis Iris (session events, long-term memory, background promotion). Use whenever designing, writing, reviewing, or debugging anything that touches Redis: choosing a data structure, naming keys, configuring a client, sharding across a cluster, building or tuning a search index, monitoring or diagnosing performance, hardening a deployment, caching LLM completions, or wiring persistent memory for an AI agent, merged into one skill so the right domain loads automatically without picking between separate ones.
Remix v2 data loading and mutations. Use when writing loaders, actions, deferred data, revalidation logic, or pending state. Triggers on loader, action, useLoaderData, useActionData, json(), defer(), <Await>, shouldRevalidate, useRevalidator, useNavigation, useTransition (v1 holdover).
Reviews Remix v2 error-handling code for the unified ErrorBoundary, isRouteErrorResponse narrowing, throw-vs-return, root boundary scaffolding, and v1 holdovers (CatchBoundary, useCatch). Use when reviewing routes that throw or define ErrorBoundary in a Remix v2 codebase.
Remix v2 form submissions and mutations. Use when implementing forms, optimistic UI, file uploads, or multi-action routes. Triggers on <Form>, useFetcher, useSubmit, useNavigation for pending state, unstable_parseMultipartFormData, fetcher.formData, intent-based actions, encType multipart.
Reviews Remix v2 form code for manual fetch() mutations, native <form> misuse, wrong useNavigation/useFetcher choice, missing pending state, unbounded uploads, and intent-pattern violations. Use when reviewing form/mutation code in a Remix v2 codebase.
Reviews Remix v2 code for v1-shape meta exports (BREAKING in v2), cookie security gaps (httpOnly, secure, secrets rotation), auth gates in wrong layer, and missing CSRF. Use when reviewing meta/SEO, session, auth, or form-mutation code in a Remix v2 codebase.
Deep analysis of Git history: identify frequently changed hotspot files, analyze code ownership by contributor, and scan for leaked secrets. Triggered when users ask about Git analysis, code hotspots, who owns what code, secret scanning, security audits of commit history, or optimizing code review assignments.
Use when working with, adding to, or updating the shared cross-project research library at $RESEARCH_HOME (vendor repo clones, reference PDFs/epubs, mirrored docs pages) — before fetching the same material from the web, when cloning a reference repo for a project, or when asked to update/refresh the library. Also owns judging a named package or tool before depending on or installing it — from PyPI, npm, crates.io or apt, or shipped only as GitHub release binaries: whether it is still maintained, who actually commits to it, how often it releases on its stable line, what a release really ships and costs (wheel or sdist, npm platform packages and install scripts, prebuilt binaries), whether its Python, node or glibc floor fits this machine, how far an apt package lags upstream, whether it ships types, how much test suite backs it, whether a version cap will hold you back — read from the registry, the GitHub API and the project's own source rather than from a search summary.
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.