Skills
99,951 packaged capabilities and playbooks for agents. Install any of them with one command.
All
Runs a fast pre-build risk review on a product idea, feature request, or scope change, naming the single assumption most likely to make it fail and returning a clear verdict (build small, validate first, pivot first, or don't build yet) with a no-code validation step. Use before committing build effort, when triaging whether to honor a feature request, or when deciding whether to expand scope, ahead of writing a PRD. For a launched product's pivot-or-persevere decision, use iterate-pivot-decision instead.
Guides developers using Fusion Framework feature flags with MCP-backed framework retrieval first and bundled public-source fallback assets when MCP is unavailable. USE FOR: helping with `enableFeatureFlag`, `enableFeatureFlagging`, `useFeature`, rollout/cleanup guidance, and finding Fusion Framework feature-flag examples. DO NOT USE FOR: generic SaaS flag platforms, backend-only rollout systems, or inventing framework APIs.
Write and edit professional content in Hebrew, including marketing copy, UX text, articles, emails, and social media posts. Use when the user asks to write in Hebrew, "ktov b'ivrit", create Hebrew marketing content, edit Hebrew text, write Hebrew UX copy, or optimize Hebrew content for SEO. Covers grammar rules, register from formal to dugri, mixed Hebrew/English, gendered language, nikud and numerals, and Hebrew SEO best practices. Do NOT use for Hebrew NLP/ML tasks (use hebrew-nlp-toolkit) or translation (use a translation skill).
Runs LLM inference on CPU, Apple Silicon, and consumer GPUs without NVIDIA hardware. Use for edge deployment, M1/M2/M3 Macs, AMD/Intel GPUs, or when CUDA is unavailable. Supports GGUF quantization (1.5-8 bit) for reduced memory and 4-10× speedup vs PyTorch on CPU.
Analyze and reduce the bundle size of rspeedy/Lynx (ReactLynx) apps — the size / `.lynx.bundle` lever (not runtime perf, types, or CI). Use when the user wants to optimize bundle size, find what is making a rspeedy bundle large, decide which size lever to pull, or shrink `main-thread.js` by stripping background-only code (jsb / telemetry / request / monitor SDKs) out of the render path via `__BACKGROUND__` / 'background only' gating. Covers media, JS-background, and main-thread layers, plus mangle/eval-poisoning bloat, i18n-locale duplication, duplicate-package dedup, extractStr/compile knobs, and lazy/dynamic-component splitting (first-screen vs total size). Measure-first and evidence-based — produces a prioritized report; changes code only when asked.
Use when a 2D side-scroller shot, platformer cutscene, game trailer or parallax animation has to be built in Blender from Scenario-generated pictures (flat image cards, an orthographic camera, keyed jumps, rigid-body props), or when a finished shot needs a new theme or look without re-animating. Also use when image planes in Blender float above the ground, a tiled parallax layer shows a seam, or physics props leave the picture plane.
Onboards an AI agent into the Senpi trading platform by creating an account, generating an API key, and configuring the Senpi MCP server connection. Supports wallet, Telegram, or agent-generated wallet identity. Use when the user says "set up Senpi", "onboard to Senpi", "connect to Senpi", "install Senpi", "register with Senpi", or when the agent needs to self-register with Senpi for autonomous trading. Do NOT use for trading operations, strategy management, or market queries -- those require the Senpi MCP server to already be connected
Use this skill when a Steel automation appears unreliable because of bot detection, CAPTCHA loops, 403/access-denied or verify-you-are-human pages, proxy failures, profile or identity problems, credential or login persistence issues, suspicious pacing, redirect loops, or retry strategy, or when the user asks about stealth, proxies, CAPTCHA solving, profiles, credentials, or reliable logins. Do not use as a first step for unknown failed sessions; collect evidence with steel-session-debugging first when needed.
Systematically add test coverage for all local code changes using specialized review and development agents. Add tests for uncommitted changes (including untracked files), or if everything is committed, then cover the latest commit.
Threat modeling workflow for software systems: scope, data flow diagrams, STRIDE analysis, risk scoring, and turning mitigations into backlog and tests. Use when designing new features, reviewing architecture changes, handling sensitive data, or hardening auth/payment/multi-tenant flows.
Use when a developer wants to iterate on ONE specific Agent Observability / LLM Obs trace whose output they didn't like—rerunning it against local code, viewing a concise diff of the old and new output, and looping (change code → replay → diff) until satisfied. Invoked as /agent-observability-replay-trace <trace-id> [changes to test]. Signals: "replay this trace"; "iterate on a trace"; "this trace's output is wrong, fix it and re-run"; "re-run trace <id> with <change>"; or pasting a trace ID from the Agent Observability UI with a description of what to fix. It fetches the trace via the datadog-llmo MCP or the pup CLI, edits code, reruns the app to emit a new trace, and diffs the two—no local server or browser. For Agents traced with ddtrace / LLM Obs (Python first-class), with JSON-serializable entry input. Do NOT use for: scored Experiments or the browser "Replay" button (that's agent-observability-replay-experiment), building an experiment from a dataset/CSV, writing evaluators, root-causing failed traces, or RUM/HTTP session replay.
Use when comparing physical Android benchmark configurations, investigating inconsistent rankings, or selecting an Android default from measured results. Do not use for code-level Compose performance diagnosis without a configuration comparison
AI-powered web search via Exa with content extraction. Use when the user says "exa search", "web search with content", "find similar pages", or needs broad web results beyond academic databases such as arXiv and Semantic Scholar.
SSH job queue for multi-seed/multi-config ML experiments with OOM-aware retry, stale-screen cleanup, and wave-transition race prevention. Use when user says "batch experiments", "队列实验", "run grid", "multi-seed sweep", "auto-chain experiments", or when /run-experiment is insufficient for 10+ jobs that need orchestration.
Enforces staged execution discipline on large tasks: a written stage plan, parallel delegation where the runtime supports it, a failable verification check at each stage, and a skeptical self-review before delivery. Trigger when the user explicitly asks OR when the task objectively spans multiple files, sources, or sessions. Do NOT trigger on ordinary multi-step requests that a direct attempt handles fine (autonomous triggering only; explicit invocation always runs).
Turns any source artifact (spec, discovery, research, GTM plan, experiment results, retro, or raw notes) into one canonical master document plus a set of audience-tailored briefings, each re-pitched to a stakeholder lens (executive, board, engineering, UX, PMM, sales, CS, legal, data, or a custom audience). Every briefing is a traceable projection of the master, so the versions never disagree. Use when one piece of work must reach several audiences who each need a different framing, decision, and level of detail.
Expert patterns for input handling covering InputMap actions, InputEvent processing, controller support, rebinding, deadzones, and input buffering. Use when setting up player controls, implementing input systems, or adding gamepad/accessibility features. Keywords InputMap, InputEvent, gamepad, controller, rebinding, deadzone, input buffer.
Draws original hand-written SVG spot illustrations in the One-line style, joyful action figures doing design and coding work. One continuous uniform black line (2.3 px at 480 wide) loops and doubles back to draw the whole figure. Flat coral, periwinkle and white patches sit behind the line, wobbly and deliberately knocked 2–6 px out of register like a misprinted riso. Tops carry coral stripe lines, and there is no ground plane, only a squiggle, a curl or a few speed strokes, all on warm beige paper. Covers the palette, a route-building toolkit (tube limbs, hand and shoe loops, wobble patches, masks), poses, composition and a render-critique loop. Use when asked for one-line, single-line, continuous-line, misregistered, riso-style or playful doodle illustrations, for empty states, onboarding, 404 pages, blog headers, feature spots, launch announcements or marketing cards.
TypeScript type safety for Inertia Rails (React, Vue, Svelte): shared props, flash, and errors via InertiaConfig module augmentation in globals.d.ts. Use when setting up TypeScript types, configuring shared props typing, fixing TS2344 or TS2339 errors in Inertia components, or adding new shared data.
Use when editing .lean files, debugging Lean 4 builds (type mismatch, sorry, failed to synthesize instance, axiom warnings, lake build errors), searching mathlib for lemmas, formalizing mathematics in Lean, finding a counterexample to, refuting, or disproving a Lean statement, or learning Lean 4 concepts. Also trigger when the user asks for help with Lean 4, mathlib, or lakefile. Do NOT trigger for Coq/Rocq, Agda, Isabelle, HOL4, Mizar, Idris, Megalodon, or other non-Lean theorem provers.
Simulates real users searching Amazon’s storefront to retrieve live search results, product positions for keywords, prices, ratings, review counts, ad labels and delivery information, and supplement SellerSprite competitor data as needed. Suitable for Amazon product searches, ASIN ranking checks, competitor discovery, price comparisons, sponsored product analysis, new product monitoring, and comparing search results across marketplaces or devices. Not suitable for historical search volumes or ranking trends, ad management, standalone sales trend analysis, review sentiment analysis, Listing copy optimization, or inventory and supply chain management.
Connect an AI coding agent to Proxyman MCP. Use when Proxyman is installed but the agent cannot see tools, or when configuring Codex, Claude, Cursor, VS Code, or Copilot and troubleshooting bridge, handshake, or tool-discovery errors.
Classic three.js materials (non-Node): MeshStandardMaterial, MeshPhysicalMaterial, Phong/Lambert/Toon/Basic, Line/Points/Sprite materials, MeshMatcapMaterial, MeshNormalMaterial, depth/distance materials, ShaderMaterial and RawShaderMaterial. Use when tuning PBR maps, transparency, depth write, skinning flags, or writing GLSL in ShaderMaterial; for TSL/NodeMaterial/WebGPU shader graphs use threejs-node-tsl instead—not both for the same authoring path unless migrating.
Install the Zapier SDK CLI for Zapier Workflows Early Access and bootstrap the workflows companion skills. Use when the user wants to set up Zapier Workflows, get started building durable workflows, install workflow skills, or configure the Zapier SDK CLI.
Build websites and web apps using 21st.dev — the largest marketplace of shadcn/ui-based React Tailwind components with 1400+ components. Use when user mentions "21st.dev", "21st", "ใช้ 21st", "จาก 21st", "build with 21st", "21st component", "21st magic MCP", or references the 21st.dev component registry in any language. Covers building complete sites (landing pages, dashboards, portfolios, e-commerce storefronts, auth pages), browsing and installing individual 21st.dev components into existing projects, and setting up the 21st.dev MCP server. Also trigger when user wants premium UI components specifically from the 21st.dev registry. Do NOT trigger for general web development, standard shadcn/ui, plain Tailwind CSS, React Native, API development, testing, deployment, or any task that doesn't explicitly reference 21st or 21st.dev.
Search free commercial-use creative assets across multiple online sources—photos, illustrations, icons, video footage, music, sound effects, and fonts. Use when the user needs to find assets for videos, PPTs, articles, or any content creation. Trigger on phrases like "find me a photo of", "search for icons", "I need background music", "find video footage", "search fonts", "找一张图片", "搜素材", "帮我找图标/视频/音乐/字体". Covers Pexels, Unsplash, Pixabay, Iconify, Freesound, Google Fonts and more—with API-backed search where available
Product content designer for UI copy. Use when writing, reviewing, or auditing user-facing text: button labels, error messages, tooltips, empty states, modal copy, placeholder text, confirmation dialogs, onboarding flows, or i18n strings. Also use when the user says /copy, /content, or /ux-copy.
Use when tackling complex reasoning tasks requiring step-by-step logic, multi-step arithmetic, commonsense reasoning, symbolic manipulation, or problems where simple prompting fails - provides comprehensive guide to Chain-of-Thought and related prompting techniques (Zero-shot CoT, Self-Consistency, Tree of Thoughts, Least-to-Most, ReAct, PAL, Reflexion) with templates, decision matrices, and research-backed patterns
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.