Skills
97,501 packaged capabilities and playbooks for agents. Install any of them with one command.
Testing & quality
Deep accessibility scorecard for a single Figma component or component set — state coverage (default/hover/focus/disabled/error/active/loading), focus-indicator quality and contrast, non-color state differentiation, target size, annotation completeness, and color-blind simulation (protanopia/deuteranopia/tritanopia), all rolled into per-category 0–100 scores and prioritized recommendations. Use when the user wants to vet ONE component before shipping it: triggers 'audit this component', 'accessibility score for this button', 'is this component set accessible', 'check the states on this component', 'does my button have a focus state', 'color blind check', 'rate this component for a11y', 'component accessibility scorecard'. Works on ANY Figma plan. For a broad sweep over a whole page/frame use lint-design-figma; for CODE-side (HTML) checks use scan-code-accessibility-figma.
Cross-discipline interface review: routes a screen, flow, feature, or product interface to every `better-*` domain skill and consolidates one ranked verdict. Use when asked for a holistic review rather than a single domain, and when `interface-review` hands up a change to route. Triggers on better-interface, holistic interface review, review the whole screen.
Code review automation for TypeScript, JavaScript, Python, Go, Swift, Kotlin, C#, .NET, Java, C, C++, Rust, Ruby, PHP, and Dart/Flutter. Analyzes PRs for complexity and risk, checks code quality for SOLID violations and code smells, and generates review reports. Use when reviewing pull requests, analyzing code quality, identifying issues, or generating review checklists
Critical rules and common mistakes to avoid in Convex development. Use when reviewing Convex code, debugging issues, or learning what NOT to do. Activates for code review, debugging OCC errors, fixing type errors, or understanding why code fails in Convex.
Automatic compliance audit for the Hexagone design system (PrimeVue 4 + Tailwind CSS 4 + Hexagone tokens). Analyses the source code and the rendered UI of a page or component, detects violations, fixes them automatically and produces a report — all without user intervention. Use when the user wants to audit a page or component's design system compliance, fix PrimeVue/Tailwind/Hexagone tokens deviations, or mentions "design compliance" / "conformité design".
Use when evaluating projects for production or enterprise readiness, implementing supply chain security (SLSA, cosign, SBOMs, pnpm), hardening CI/CD pipelines, establishing quality gates (TYPO3: CI matrix PHP 8.2-8.5 x TYPO3 12.4/13.4/14.3 LTS), pursuing OpenSSF Best Practices Badge (Passing/Silver/Gold) or OSPS Baseline levels, reviewing code quality, writing ADRs, or configuring Git hooks and CI pipelines.
Use the published FinalRun CLI to install, configure, validate, run, troubleshoot, and inspect reports for repo-local mobile test workspaces. Trigger this skill for requests involving finalrun, check, test, suite, doctor, runs, start-server.
Create production-ready GitHub Actions workflows for automated testing, building, and deploying applications. Use when setting up CI/CD with GitHub Actions, automating development workflows, or creating reusable workflow templates.
Plain-language fix guide for failing GTM tests. Use when tests have failed and the developer needs step-by-step instructions to fix each issue. Reads gtm-test-results.json and writes a fix guide in plain English - no GTM expertise required to follow. No API calls. Trigger on - "fix guide", "how do I fix failing tests", "tests failing", "gtm test failures", "fix my tracking", "what broke".
Use this skill to drive an Android device from the host shell — inject taps, swipes, text, key events, and drag-and-drop via `adb shell input`; resize the display with `wm size` / `wm density`; flip hermetic-test settings with `settings put global window_animation_scale 0`; reset app state with `pm clear` vs `am force-stop`; launch Activities with `am start -n pkg/.Activity`; and use the modern `cmd <service>` wrapper (`cmd wifi`, `cmd connectivity`, `cmd uimode night`, `cmd locale set-app-locales`) instead of the deprecated `svc` aliases on API 30+. If the user mentions "input tap", "input swipe", "input text spaces", "KEYCODE_BACK", "wm size for screen-size matrix", "settings put animation_scale", "force-stop vs pm clear", "svc wifi no-op", "cmd wifi set-wifi-enabled", or "am start intent extras", use this skill.
This skill MUST be used when a bug's root cause is unknown and requires investigation, or when the user explicitly requests investigate or root-cause analysis. SHOULD be used for unexplained failures, regressions, or intermittent behavior: 根因調查, 為什麼會這樣, 壞掉了, failing, broken, flaky. MUST NOT treat an apparently obvious cause as confirmed without evidence. When evidence already establishes the cause, proceed to targeted verification and repair without restarting the investigation.
Implement comprehensive testing strategies using Jest, Vitest, and Testing Library for unit tests, integration tests, and end-to-end testing with mocking, fixtures, and test-driven development. Use when writing JavaScript/TypeScript tests, setting up test infrastructure, or implementing TDD/BDD workflows
Implement comprehensive evaluation strategies for LLM applications using automated metrics, human feedback, and benchmarking. Use when testing LLM performance, measuring AI application quality, or establishing evaluation frameworks.
Checks local maintainability hotspots: complexity, long methods, god modules, signatures, algorithms, and constants. Also flags identifier drift across API/DTO/DB layers. Use when auditing code hotspots.
Generate and run MATLAB unit tests using matlab.unittest and matlab.uitest. Parameterized tests, fixtures, mocking, coverage analysis, CI/CD with buildtool, app testing with gestures. Use when creating tests, writing test classes, running test suites, checking coverage, testing apps, or validating MATLAB code.
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.
Use when the user asks to review code/a PR with opencode or with a non-Claude model (e.g., "review with opencode", "review using gpt-5.6-sol", "/opencode-review"). Runs a fresh opencode session in a subagent, tells it to use the review-code skill, and relays its review.
Use when improving performance, latency, throughput, memory usage, or general efficiency. Start by defining target metrics, measuring comprehensively, attributing bottlenecks, validating with static analysis, and prioritizing macro-optimizations before micro-optimizations.
Use when the user wants to verify paper claims against code or data, audit numerical accuracy, check formula-code alignment, or validate citation accuracy. Triggers on phrases like "verify claims", "check numbers", "do the numbers match", "formula vs code", "audit the paper", or "cross-check results".
Configure and interpret the Phoenix plugin for Harbor agent evaluations. Use when adding `arize-phoenix` to Harbor jobs, choosing ATIF tracing, mapping Harbor tasks and rewards to Phoenix experiments, comparing agents or models, resuming jobs, or troubleshooting Harbor records in Phoenix.
Use when writing E2E tests with Cucumber BDD and Playwright, setting up BDD test infrastructure, writing Gherkin feature files, implementing step definitions, configuring hooks and World, or integrating Cucumber with CI/CD. Invoke for Cucumber, BDD, Gherkin, feature files, step definitions, Given When Then, hooks, tags, World, scenario outline, @cucumber/cucumber.
Report a Decentraland SDK bug, engine/Explorer runtime limitation, or docs gap to the SDK team before applying a workaround. Use whenever you or the user find that an SDK7 API, component, the Explorer, the Creator Hub, or the `@dcl/sdk` toolchain doesn't behave as documented, is missing something a scene reasonably needs, or forces a workaround—even when the workaround is easy. Asks the user for consent once per scene, then reports automatically and never sends the same issue twice. Do not use for bugs in the user's own scene code, questions about how an API works, or issues a skill already documents as a known limitation
Use when reviewing or judging a finished api-spec—the engineering wire contract (operations, request/response schemas, auth, error model, pagination, versioning, examples)—to decide whether a client can call every operation and a server can implement it from the contract alone. An acceptance gate, not authoring. Judges a single-sourced 11-condition contract-completeness bar: style + base + versioning + deprecation policy; every operation listed + traced; every operation fully typed on both sides + status codes (happy-path-only fails); auth + per-operation authorization; a complete error model (one shape + every failure case + retryability); shared types referencing the data-model; pagination + a tie-breaker; examples matching the schemas; one-directional vs the api-reference; consistent with the shipped API; delta-scoped amend. Style-agnostic (REST/GraphQL/gRPC, no OpenAPI reflex). Emits exactly `VERDICT: approve|revise`. Not authoring, the data-model, the api-reference, or generic design docs (design-review).
Generates structured test assertions and failure diagnostics for skill packages from a definition and task prompt. Triggers on: "verify this skill", "generate assertions", "surrogate verification", "diagnose skill failure". NOT for code review, use pr-review.
Test-driven development. Use when the user wants to build features or fix bugs test-first, mentions TDD or "red-green-refactor", asks for tests that prove a behavior, wants a bug reproduced as a failing test, or wants integration tests. Not for only running an existing test suite.
Test Temporal workflows with pytest, time-skipping, and mocking strategies. Covers unit testing, integration testing, replay testing, and local development setup. Use when implementing Temporal workflow tests or debugging test failures
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.
Use for a Yandex KIT store launch-readiness review: «Можно запускать?», «Проверь готовность», «Что мешает открытию?», «Можно вести покупателей?». Model-invoked; the default workflow is fully read-only and distinguishes proven blockers, risks and unverified checkout links.
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.