Skills
97,725 packaged capabilities and playbooks for agents. Install any of them with one command.
Development
Version-aware guidance for PufferLib reinforcement-learning environments, vectorization, policies, PuffeRL training, evaluation, and safe checkpoint review. Use when adapting Gymnasium/PettingZoo environments to published PufferLib 3.0.0 or working with the redesigned native 4.0 source line.
Comprehensive internationalization skill for Ruby on Rails applications with proper English and Arabic translations, RTL support, pluralization rules, date/time formatting, and culturally appropriate content adaptation.
Production Ray (open-source, pinned to 2.57) for classic-ML workloads from training to serving — Ray Train, Tune, Data, Serve, Core, and cluster deployment on KubeRay. Corrects the older-corpus defaults a model reaches for (ray.air session reporting, Trainer-inside-Tuner, tune.run, map_batches concurrency=, DatasetPipeline/to_torch, max_concurrent_queries, RayServeHandle + ray.get, Deployment.deploy, ray.state, ray.get-in-a-loop) with the 2.57 idioms that replaced them (Train V2 defaults, driver-function tuning, compute strategies, streaming datasets, DeploymentHandle/DeploymentResponse, serve build/deploy, ray.util.state, KubeRay CRDs and Jobs API). Use when writing, reviewing, or productionizing Python code that touches Ray distributed training, data pipelines, hyperparameter tuning, model serving, or Ray cluster operations. LLM serving/batch-inference on Ray lives in the sibling ray-llm skill.
Electron desktop app reverse engineering: asar unpacking, main/renderer process JS, V8 bytecode (.jsc) boundaries, CDP dynamic debugging, and anti-debugging countermeasures. Triggers: Electron, asar, desktop app reverse engineering, .jsc, V8 snapshot, CDP, ELECTRON_RUN_AS_NODE, devtools detection.
Firmware/embedded/hardware analysis gateway. Workflow: extraction → rootfs → emulation → hardware interfaces. Subskills: [[re-fw-extract]] [[re-fw-rootfs]] [[re-fw-emulate]] [[re-hardware-io]] [[re-automotive]] [[re-uefi]] [[re-rtos]] [[re-tee]]. Trigger terms: firmware analysis, firmware, IoT, unpacking firmware, rootfs, firmware emulation.
Virtualization reverse engineering: VT-x/SVM, hypervisor detection, VMCS/EPT analysis, and partitioning and vdev semantics in Xen / QNX Hypervisor / Jailhouse / ACRN / Bao / Hyper-V·VMBus / XtratuM / LynxSecure / Quest-V. Trigger terms: hypervisor, VT-x, SVM, virtualization detection, EPT, Xen, grant table, event channel, Jailhouse, cell, ACRN, ivshmem, Bao, shmem_id, vdev, GPA/HPA, Hyper-V, VMBus, VSC, VSP, GPADL, SR-IOV, XtratuM, XM_CF, LynxSecure, Quest-V, sandbox kernel
Managed code reverse-engineering gateway. Orchestration: identify runtime → decompile → deobfuscate → route malware scenarios to [[re-malware]]. Subskills: [[re-dotnet]] [[re-java]] [[re-script-deob]] [[re-wasm]] [[re-ai-triage]] → [[re-ai-model]] [[re-blockchain]] [[re-python]]、[[re-browser-ext]]、[[re-electron]]、[[re-javacard]]. Trigger terms: .NET, dnSpy, ILSpy, Java, jar, bytecode, PowerShell obfuscation, VBA macro, JS deobfuscation, managed code.
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.
Transpile React Web applications (v18+) into Taro 4.x code optimized for WeChat Mini Program. Use this skill when converting React components, pages, or utilities to Taro-compatible code with proper JSX element mapping, event handler transformation, navigation/routing conversion, and API shim layer.
Write and edit human-facing project technical text for scanning, wherever it lives — README, specs, design notes, plans, reports, runbooks, changelogs, and project memory, not only docs/ or StrictDoc. Use whenever creating or revising these records, without waiting for an explicit style request; also for 说人话、去 AI 味、精简文档、短语分点、PPT 式表达. Prefer Chinese phrases, lists, and compact comparison tables while preserving facts, conditions, evidence, and required structure. Not a detector-evasion or creative-writing skill.
Use when building interactive 2D or 3D graph or network visualizations in React with the reagraph library (GraphCanvas, node-link diagrams, force-directed layouts). Covers data shapes (GraphNode, GraphEdge), 16 layout algorithms, selection, lasso, camera controls, sizing strategies, and theming.
Performance best practices for Relay applications. Use when optimizing data fetching, reducing re-renders, configuring caching, or improving time to first meaningful paint. Covers query placement, @defer, pagination, fetch policies, garbage collection, fragment granularity, and server-side filtering. Companion to the relay-best-practices skill which covers correctness and architecture.
Detects and prevents manual edits to release-please managed files (CHANGELOG.md, version fields in package.json, pyproject.toml, Cargo.toml). Provides conventional commit templates. Use when editing changelogs, version bumps, release files, or when user mentions "release", "changelog", "version bump", or "conventional commits".
Use this skill when releasing a new version of a GitHub project, covering the full process of version number management, CHANGELOG syncing, writing release notes, creating tags, CI build monitoring, release verification, and history cleanup. Applies to any GitHub-based software project such as desktop apps, CLI tools, web apps, and libraries/SDKs. Triggers when the user mentions "发布" (release), "release", "打 tag" (tagging), "新版本" (new version), "更新版本号" (updating the version number), "写 release notes" (writing release notes), "发布失败了" (release failed), "CI 挂了" (CI is down), "Actions 配额告急" (Actions quota running low), "短时间内多次发版" (multiple releases in a short time), "monorepo", "批量打包" (batch packaging), "多 skill 发布" (multi-skill releases), "skill zip", "专家套件 zip" (expert suite zip). Also used to reject the anti-pattern of treating a release as a CI verification mechanism ("打 tag 看一下", just tag it and see). Do not use for non-GitHub projects (such as pure GitLab / Gitea projects) or manual release scenarios that require no CI.
Turn a brief or existing feature into testable requirements: behavior, acceptance criteria, quality constraints, and data and permission rules. Use when notes or a concept need FR/NFR requirements or acceptance criteria before design or build, or when checking which requirements a shipped feature meets.
Request/execute structured code review: use after completing important tasks, at end of each execution batch, or before merge. Based on git diff range, compare against plan and requirements, output issue list by Critical/Important/Minor severity, and provide clear verdict on merge readiness. Trigger words: request code review, PR review, merge readiness, production readiness.
Reproduces, diagnoses, and fixes a defect in a Backstage plugin inside the rhdh-plugins or community-plugins repositories: resolve the report from a RHIDP, RHDHBUGS, RHDHPLAN or RHDHSUPP Jira key, a GitHub issue, or plain prose; locate the owning workspace; write a throwaway reproduction test that fails for the stated reason; record before and after Playwright video evidence for a UI defect; then apply the smallest fix and verify it. Use for "fix this plugin bug", "reproduce RHIDP-1234", a plugin that renders the wrong thing, or a regression that needs evidence before and after.
How to write idiomatic, efficient RxJS code. Use this skill whenever the user is writing, refactoring, reviewing, or debugging code that uses RxJS — including any file that imports from 'rxjs' or 'rxjs/operators'. Trigger on mentions of observables, subscriptions, RxJS operators, or reactive streams. Even if the user doesn't say "RxJS" explicitly, activate when you see patterns like `.pipe()`, `.subscribe()`, `Observable`, `Subject`, `BehaviorSubject`, `switchMap`, `mergeMap`, or similar.
Plan safe database schema evolution and zero-downtime change rollout for Kotlin plus Spring systems using Flyway, Liquibase, or equivalent migration tooling. Use when changing tables, columns, constraints, indexes, or data shape in systems with live traffic, rolling deploys, large datasets, or backward-compatibility requirements between old and new application versions.
Machine learning in Python with scikit-learn. Use for supervised learning (classification, regression), unsupervised learning (clustering, dimensionality reduction), model evaluation, hyperparameter tuning, preprocessing, or building ML pipelines. Provides comprehensive reference documentation for algorithms, preprocessing techniques, pipelines, and best practices.
Spectacle, budgeted. Scroll-driven animation with native CSS first and GSAP ScrollTrigger when needed, smooth scrolling with Lenis and when to turn it off, split-text reveals, pinned scenes, canvas and WebGL on a performance budget, shaders as material, pausing off-screen, loading strategy, and the reduced-motion and no-JS fallbacks that keep the page readable. Use for marketing pages, portfolios, launches and hero sections. Never for product UI.
Perform language and framework specific security best-practice reviews and suggest improvements. Trigger when the user requests: security review, security report, threat assessment, vulnerability scan, security audit, secure-by-default coding, hardening guidance, or security best practices guidance. Supported languages: Python, JavaScript/TypeScript, Go. Covers OWASP Top 10, supply chain security, secrets management, dependency auditing, and modern cloud deployment hardening.
Apply Semantic Versioning 2.0.0 (semver.org) to decide version bumps, validate version strings, and compare version precedence. Use whenever a task involves choosing the next version number, judging whether a change is a MAJOR/MINOR/PATCH bump, checking that a string is a well-formed version, ordering versions by precedence, or building a pre-release or build-metadata version. Often used as a building block by release, changelog, and publishing workflows — trigger whenever version numbers are being reasoned about, even if "semver" is never said.
Comprehensive fullstack development skill for building complete web applications with React, Next.js, Node.js, GraphQL, and PostgreSQL. Includes project scaffolding, code quality analysis, architecture patterns, and complete tech stack guidance. Use when building new projects, analyzing code quality, implementing design patterns, or setting up development workflows
Guided entry point for using Sentry through your agent. Orients you to your current setup and, for a new project, sets up Sentry end to end with sane defaults — provision a project, install the SDK (errors, tracing, and whatever it enables by default), and confirm real telemetry reaches Sentry. Routes other intents (adding more signals, fixing issues) to the right skill.
Serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production LLM APIs, optimizing inference latency/throughput, or serving models with limited GPU memory. Supports OpenAI-compatible endpoints, quantization (GPTQ/AWQ/FP8), and tensor parallelism
Fast structured generation and serving for LLMs with RadixAttention prefix caching. Use for JSON/regex outputs, constrained decoding, agentic workflows with tool calls, or when you need 5× faster inference than vLLM with prefix sharing. Powers 300,000+ GPUs at xAI, AMD, NVIDIA, and LinkedIn.
Write, debug, and optimize Shadertoy shaders and GLSL fragment shaders for procedural graphics. Use when creating visual effects, generative art, or WebGL shaders; when writing .glsl files; or when user says "write a shader", "make a Shadertoy effect", "help with GLSL", or "create a procedural animation".
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.