Skills
96,981 packaged capabilities and playbooks for agents. Install any of them with one command.
Testing & quality
Review UX design intent and available implementation evidence for accessibility barriers, inclusive interaction, WCAG or Section 508 requirements, keyboard and focus behavior, semantics, dynamic announcements, content, motion, zoom/reflow, and recovery. Use when a designer asks for an accessibility review or needs accessibility behavior defined before engineering handoff.
Use when auditing or improving GitHub Actions workflows for a project. Covers cache setup (npm/pnpm/yarn), job parallelism, shard-based test splitting, artifact handling pitfalls, and Playwright-specific patterns. Trigger on: slow CI, cache miss, flaky shard jobs, merge-reports failures, or an explicit 'tune CI' request.
Design and audit AI feature product risks across user value, model limitations, hallucination, automation boundaries, human review, privacy, safety, evals, latency, cost, transparency, abuse, fallback, rollout, and support. Use when adding AI assistants, copilots, generators, classifiers, recommendations, agents, or automated decisions to a product.
axe-core - accessibility testing engine for automated WCAG compliance checks. USE WHEN: user mentions "axe", "automated accessibility testing", "a11y tests", asks about "integrating accessibility checks", "CI/CD accessibility", "Playwright accessibility", "Jest accessibility", "Vitest axe" DO NOT USE FOR: manual accessibility audits - use `wcag` skill instead
Validates animation durations, enforces typography scale, checks component accessibility, and prevents layout anti-patterns in Tailwind CSS projects. Use when building UI components, reviewing CSS utilities, styling React views, or enforcing design consistency.
Behavioral testing methodology — test what users experience, not how code is structured. Use when writing tests, reviewing test quality, planning test strategy for new features, or when existing tests are brittle/verbose/coupled to implementation details. Triggers: writing tests, TDD, test review, "tests keep breaking", "too many mocks", "tests are verbose", test coverage planning, behavior-driven development.
Find what a change could break somewhere else before it ships, beyond the diff, and prove the one fact it's safe because of by running real code instead of writing it up. Use for 'blast radius of X', 'what could this break', or reviewing a small diff you don't trust.
Structured bug fixing with reproduction, root cause analysis, fix design, regression risk analysis, and quality review. Only invoke this skill when the user explicitly asks to use it by name (e.g. "use bug-fix", "run bug-fix") or clearly requests a structured/systematic bug fixing process. Do not invoke for general bug reports, error messages, or broken behavior — handle those directly without this skill.
Set up and run Bugpatrol, a QA team of AI agents that explores a web, Electron, iOS, or Android app, finds and fixes bugs, and opens GitHub PRs and issues. Use when the user wants to add Bugpatrol to a repo, write or fix .bugpatrol/bugpatrol.yml, run an explore, judge, fix, or patrol cycle, or review bugs found by Bugpatrol.
This skill is required when initializing, maintaining, and executing the by-harness workflow. Applies when the user mentions by-harness, harness, initialization, continuous task breakdown, executing feat, plan/build/qa/fix, quick fix, session_close, automatic continuation, runtime upgrades, or needs to issue a Java master gate or distributed Java gate to constrain model coding. This skill generates an independent closed-loop scaffold, sharded task storage, a quick-fix triage tool, session close-out tooling, and runtime upgrade tooling, and issues Java hard-rule gates and a distributed Java coding contract; feature_list serves only as a legacy compatibility mirror.
Use this skill when the user asks to check Simulink model compliance against a standard (MISRA, MAB, JMAAB, ISO, DO, IEC, EN, CERT C/CWE, AUTOSAR, Simulink Code Inspector (SLCI)), wants to run Model Advisor checks, or needs a compliance report with fix suggestions. For JMAAB/MAB, supplement deterministic checks with agentic review of uncheckable guidelines.
Use this skill whenever the user asks for a code review, package review, or architectural audit. Delivers severity-ranked findings with evidence and fixes, not general advice. Always saves the review as a markdown file and returns a concise summary in chat.
Design, write, review, and test code samples in developer documentation, including executable examples, explanatory snippets, request and response pairs, placeholders, outputs, language choice, sandboxing, and autogenerated samples. Use when docs include API examples, SDK snippets, CLI commands, config examples, tutorials, or code blocks readers may copy.
Commit the current work — runs the quality, consistency and passport gates, branches off main if needed, stages specific files, and writes a commit whose subject states what is now true. Pushes and opens a pull request only with --pr or when the user asks; never merges — a merge happens only when the user explicitly says to merge. Use ONLY on explicit commit intent — user says "commit", "let's commit this", "open a PR", or prefixes with `/commit`. Do NOT auto-invoke on vague end-of-task phrases ("we're done", "wrap up") — those require explicit confirmation first. Never force-pushes or skips hooks.
Render a Jetpack Compose or Compose Multiplatform @Preview to PNG in one call (MCP render_preview, or the compose-preview CLI) and look at it. Use whenever someone asks to render, show, or see what a composable, @Preview or Compose component looks like (including library components such as Material 3 or Wear M3 EdgeButton, which come from the hosted catalog), after editing Compose UI to verify it, to compare before/after, or to review previews against catalog design guidelines.
Generate a project-local verification skill that drives your app the way a user does — any language, framework, or platform. Use for /create-verification-skill, "make a control skill for this repo", or when a project has no scripted way to prove UI/CLI/service behavior.
Design and audit creator, listing, plugin, skill, app, marketplace, content, or community ranking quality across relevance, quality signals, freshness, fairness, fraud resistance, diversity, monetization, review abuse, creator trust, and user outcomes. Use when marketplaces need discovery/ranking systems that reward quality without becoming pay-to-win or spam-prone.
Periodic dependency review for Node.js/pnpm projects — outdated package triage, security audit, update batching strategy (patch/minor/major), validation checklist. Run monthly or before major releases. Use when asked to review or update dependencies in a repo.
Review an Apache Doris PR from a local clone with the same multi-agent, shared-ledger convergence workflow as the CI Code Review Runner. Use when the user supplies a PR to `/doris-repo-review`, asks to review a Doris PR locally, or asks for the CI-style review flow. Safely align only the current worktree to the exact PR head, refuse to disturb tracked local changes, write equivalent English and Chinese review documents with verified path-line anchors, and never build, test, or edit source. Treat a review as pipeline-equivalent only when an explicitly selected qualified reviewer uses a supported effort (xhigh or max, plus ultra for Codex), preferring GPT-6 Astra on Codex and Fable 5.1 on Claude Code. After a converged APPROVE result, automatically post one machine-readable PASS comment bound to the reviewed commit.
Analyze a stack trace, crash report, panic, or error log; find the failing frame and the root cause; then propose a fix and a way to verify it. Covers Kotlin/JVM/Android (coroutines, ANRs, R8-obfuscated traces), KMP/Compose Multiplatform, Flutter/Dart (FlutterError, RenderFlex overflow), Swift on iOS/macOS (.ips reports, EXC_BAD_ACCESS), Rust panics, C++ segfaults and sanitizer output, plus Python, JS/TS, Go and Java. Use when someone pastes a stack trace, crash log or panic—even with no question attached—or says “why does this crash”, “what does this error mean”, “debug this exception”, “root-cause this trace”, “symbolicate this crash”, “deobfuscate this trace”, or the Hungarian “miért crashel”, “mit jelent ez a hiba”, “elemezd ezt a stack trace-t”, “nézd meg ezt a crash logot”. Not for auditing a whole project for latent bugs (that is code-analyzer), nor for a red CI run whose log has not been fetched yet (that is github-fix-ci-error).
Evaluate the quality of any AI-generated artifact — visualizations, code, documents, conversations, or any skill output. Works in 3 phases: (1) Generate evaluation specs tailored to the artifact type, (2) Run comprehensive evaluation against those specs, (3) Produce a beautiful visual report using the /visualize skill. Use after any skill produces output, or invoke directly with /evaluate <file-or-context>. Supports evaluating: HTML visualizations, code projects, documents, agent conversations, slide decks, dashboards, or any artifact with quality dimensions.
Provides guidance for automatically evolving and optimizing AI agents across any domain using LLM-driven evolution algorithms. Use when building self-improving agents, optimizing agent prompts and skills against benchmarks, or implementing automated agent evaluation loops.
Design and audit feature flags, staged rollouts, kill switches, targeting rules, exposure plans, holdouts, rollout gates, rollback paths, flag cleanup, and operational ownership for SaaS, apps, games, AI features, marketplaces, and desktop software. Use when releasing risky changes gradually or when a product needs safe experimentation, canaries, or emergency disablement.
This skill should be used when the user asks to "test for directory traversal", "exploit path traversal vulnerabilities", "read arbitrary files through web applications", "find LFI vulnerabilities", or "access files outside web root". It provides comprehensive file path traversal attack and testing methodologies
Frontend (including Electron desktop) performance optimization knowledge base, covering version-specific optimization knowledge for React 16-19, Angular 9-18+, and Electron 12-28+. Works together with the perf-workflow skill: perf-workflow drives the analysis process while this skill supplies frontend-specific quantitative standards, version-aware optimization strategies, bottleneck patterns, and tool references. Use when analyzing performance issues in Web frontend (React/Angular/Vue) or Electron desktop apps. Triggers — 「前端性能」「前端性能优化」「Electron 性能」 / frontend performance, frontend perf optimization, Electron performance.
Write tests that start with acceptance criteria, then add implementation tests for robustness. Use when writing unit tests (Vitest), end-to-end tests (Playwright), visual regression tests, or accessibility tests. Emphasizes user-centric testing, semantic locators, accessibility validation, and the balance between acceptance and implementation testing.
Use this skill whenever the user is writing, debugging, or configuring tests in an AdonisJS application with Japa. Trigger for any testing task: writing API tests, browser tests (Playwright), console/command tests, database state setup, test doubles (fakes, swaps, time), factory usage, or any question about the testing layer. Also trigger for "write a test", "how to test", "fake the mailer", "loginAs", "truncate database", "withGlobalTransaction", "emitter.fake", "hash.fake", "drive.fake", "swap container", "freeze time", "timeTravel", "useFake", "raw mode", "record mode", or any Japa-specific concept. For AdonisJS backend patterns used inside tests (models, factories, auth) use the adonisjs skill alongside this one.
LLM observability platform for tracing, evaluation, and monitoring. Use when debugging LLM applications, evaluating model outputs against datasets, monitoring production systems, or building systematic testing pipelines for AI applications.
This skill should be used when the user asks to "configure ls-lint", "set up filename linting", "enforce naming conventions", "create .ls-lint.yml", "lint file names", "lint directory names", "file naming rules", "directory structure linting", or mentions ls-lint, directory naming rules, or filename conventions.
Deploy MATLAB-generated code to embedded hardware using Embedded Coder. Use when configuring code generation for microcontrollers (STM32, Raspberry Pi, ARM Cortex), setting up PIL/SIL verification, disabling dynamic memory allocation, or configuring hardware-specific code generation settings. Covers ERT-based configurations, processor-in-the-loop testing, memory constraints, and the MEX→SIL→PIL verification progression.
Help users create and manage mirrord preview environments — running a modified service as an isolated pod in a shared Kubernetes cluster, scoped by an environment key and HTTP/queue traffic filtering, so teams can validate and review changes against real traffic without affecting live services. Use when a developer wants to run "mirrord preview" ad hoc, share a preview via a link (mirrord-share-ingress), or wire preview environments into CI with the metalbear-co/mirrord-preview GitHub Action (e.g. per-PR previews, least-privilege cluster access).
Evaluates LLMs across 100+ benchmarks from 18+ harnesses (MMLU, HumanEval, GSM8K, safety, VLM) with multi-backend execution. Use when needing scalable evaluation on local Docker, Slurm HPC, or cloud platforms. NVIDIA's enterprise-grade platform with container-first architecture for reproducible benchmarking.
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.