Skills
101,174 packaged capabilities and playbooks for agents. Install any of them with one command.
All
Screenshot-to-HTML prototype generator with iterative refinement and learning memory. Use this skill whenever the user provides a screenshot, mockup, wireframe, or image of any UI and wants it reproduced as a working HTML prototype — or when they want to modify an existing prototype they previously generated. Also triggers on: "照这个做原型", "参考这个图", "把这个页面画出来", "this UI needs to be prototyped", "replicate this design", "convert this mockup to HTML", "帮我出个原型", "根据截图输出 HTML", or any image attachment combined with requests like "输出 HTML", "做成页面", "帮我实现". Even if the user just sends a screenshot with a brief instruction like "加一个字段" or "这个也一样", this skill applies — it means they want you to modify or replicate the UI shown. When in doubt, if there's a UI screenshot in the conversation, use this skill.
Use when: new requirements need to be entered and tracked, requirement status needs changing, the requirement pool needs unified management, or the growing number of requirements needs structuring. Do NOT use when: requirements are few and can be managed directly, or only one-off analysis is needed without ongoing tracking
Live preview of product documentation — preview all MD documents under docs/ in the browser, with automatic refresh when documents are updated. Use when: you need to preview product documentation or check how MD documents such as PRD/BRD/MRD render. Do NOT use when: the documents have not been generated yet (docs/ is empty)
Create research posters using HTML/CSS that can be exported to PDF or PPTX. Use this skill ONLY when the user explicitly requests PowerPoint/PPTX poster format. For standard research posters, use latex-posters instead. This skill provides modern web-based poster design with responsive layouts and easy visual integration.
Generate high-quality Pull Request descriptions in Markdown. Supports issue, feature, and big-feature PR types with structured, production-ready output. Also generates PR titles and supports delivery as chat output or downloadable .md file. Proactively detects open GitHub PRs for the current branch and offers to update their title and description directly via the gh CLI.
Integrate Prava's payment SDK into AI applications — create server-side payment sessions, embed or host the PCI-compliant checkout, retrieve one-time Visa credentials, and report real processor outcomes. For application integrations, not the Prava agent CLI.
Use this skill to diagnose "no node matched", "found N nodes", and "useUnmergedTree" failures by dumping the actual semantics tree with `printToLog(tag, maxDepth)` and `printToString(maxDepth)`. Covers the output grammar (`Node #<id> at (l, t, r, b)px, Tag: '...'`, alphabetically sorted config, `Actions = [...]`, `[FlagKey]` for `Unit`-valued semantics, `MergeDescendants`, `ClearAndSetSemantics`), the framework hint that suggests `useUnmergedTree = true` when matches are found only in the unmerged tree, and `fetchSemanticsNode` / `fetchSemanticsNodes` for custom matchers. Use when the developer reports "tag not found", "multiple nodes matched", "merged tree vs unmerged tree", "test fails with assertion error and a long error message", or wants to debug a finder without adding `Thread.sleep`.
Programmatic PubMed access via NCBI E-utilities REST API. Covers Boolean/MeSH queries, field-tagged search, endpoints (ESearch, EFetch, ESummary, EPost, ELink), history server for batches, citation matching, systematic review strategies. Use for biomedical literature search or automated pipelines.
Manages cloud infrastructure through natural language conversations with Pulumi Neo, an AI agent for platform engineers. Enables infrastructure analysis, resource provisioning, stack deployment, and configuration management via conversational AI. Use when creating Neo tasks, requesting infrastructure analysis, automating cloud deployments, managing infrastructure as code (IaC), provisioning AWS/Azure/GCP resources, managing infrastructure through natural language prompts, reviewing PRs with Neo, handling Neo approval workflows, or checking Neo task status and events. Also use when the user mentions "Pulumi Neo", "Neo task", "Neo agent", or wants AI-assisted infrastructure management.
Guidance for advanced Microsoft Purview DLP capabilities - Adaptive Protection (risk-based DLP driven by Insider Risk), Exact Data Match, contextual conditions, and advanced Endpoint DLP egress controls. Covers dynamic, risk-adaptive enforcement and high-precision matching for mature programs. WHEN: advanced DLP, Adaptive Protection, risk-based DLP, Exact Data Match DLP, dynamic DLP enforcement, Insider Risk driven DLP, advanced Endpoint DLP, contextual DLP conditions, dynamic block actions.
Guidance for assessing and remediating oversharing before and during Microsoft 365 Copilot adoption, using SharePoint Advanced Management (SAM), sensitivity labels, restricted content discovery, and DLP for Copilot so Copilot only surfaces content users should access. Covers data access governance reports, EEEU cleanup, and ongoing governance. WHEN: Copilot oversharing, prepare data for Copilot, restrict Copilot access, SharePoint Advanced Management, data access governance, oversharing remediation, Copilot readiness data security, limit Copilot content, Copilot is surfacing files users should not see, too many people have access to SharePoint sites, tighten up permissions before Copilot rollout, how do I make SharePoint Copilot-ready. DO NOT USE when the goal is monitoring what sensitive data users are actively sending in AI prompts (use purview-dspm-ai).
Guidance for Microsoft Purview data classification and sensitivity labels - sensitive information types (SITs), trainable classifiers, exact data match, and sensitivity label taxonomy with auto-labelling. Covers building a label taxonomy, picking detection methods, and rolling out auto-labelling in simulation. WHEN: data classification, sensitivity labels, sensitive information types, SIT, trainable classifier, exact data match EDM, auto-labelling, label taxonomy, classify data, information protection labels, MIP labels.
Guidance for Microsoft Purview Data Lifecycle Management and Records Management—retention labels and policies to retain and delete content across Microsoft 365, plus declaring records. Covers retention strategy, adaptive vs. static scopes, principle of precedence, disposition review, and regulatory records. WHEN: data lifecycle management, retention policy, retention label, records management, declare record, disposition review, adaptive scopes, retain and delete content, file plan, regulatory retention, immutable records.
Guidance for the Microsoft Purview Data Map—the foundation that scans and maps data sources across multicloud and on-premises estates to power cataloging and governance. Covers source registration, integration runtimes (managed vs. self-hosted), scan rule sets, classifications, collections, and cost control. WHEN: Purview Data Map, scan data sources, register data source, data discovery, integration runtime, map enterprise data, multicloud data scanning, classification scan, collections, glossary.
Uses the PyTDC package (import tdc, Therapeutics Data Commons) to discover therapeutic ML tasks from tdc.metadata, plan and load approved datasets, apply task-aware splits (random, scaffold, cold_split, combination, time), run evaluator metrics, evaluate benchmark groups such as admet_group, and run bounded molecular-oracle scoring. Use when selecting a TDC task or dataset, planning a download-free split, scoring predictions with TDC evaluators, running a benchmark group evaluation, or checking dataset licenses and cache effects before downloading. Use when scoring molecules with TDC oracles like QED. Not for generic RDKit cheminformatics or training molecule generators.
High-level PyTorch framework with Trainer class, automatic distributed training (DDP/FSDP/DeepSpeed), callbacks system, and minimal boilerplate. Scales from laptop to supercomputer with same code. Use when you want clean training loops with built-in best practices.
Audit UI/UX in a SENIOR style — four sections: accessibility · visual-hierarchy · consistency · antipattern, each giving a SCORE/10 + MOST SEVERE ISSUE + HOW TO FIX, then a conclusion of PASS-to-next-step or NEEDS-FIXING. The MEASURABLE part (WCAG contrast, tap-target, geometric overlap/misalignment, missing-label) runs deterministically through the visual-qa engine (headless, 0-token, NO LLM); the part needing EYES (hierarchy, design-system drift, dark-pattern) is an LLM audit. Call when the user says "qc uiux", "audit ui", "inspect the interface", "clean up UI antipatterns", "/qc-uiux", or AFTER there is a mockup/UI. MIRRORS /qc-code (expensive = manual LLM, cheap = automatic deterministic tests). DIFFERENT from /redesign (aesthetic renewal) and /visual-qa (screenshot + pixel guarding) — qc-uiux is a 4-section senior verdict.
Builds semantic code indexes, maps dependency graphs, and performs intelligent code search across large codebases. Use when understanding unfamiliar code, tracing call chains, analyzing import dependencies, or reducing context window usage through targeted retrieval.
French tax radar and cash pilot for Qonto accounts. Builds a dated, amount-estimated tax schedule (VAT, corporate tax instalments, CFE, dividend flat tax), projects the next 90 days of cash, computes the month's tax provision, and—with explicit user consent—creates a transfer request to a dedicated tax sub-account that the user approves with 2FA in the Qonto app. Use for “où en sera ma tréso fin septembre ?”, “quand tombe ma TVA et combien ?”, “provisionne mes impôts”, “can I afford this purchase?”, “what if my client pays late?”.
Build Retrieval-Augmented Generation (RAG) systems for LLM applications with vector databases and semantic search. Use when implementing knowledge-grounded AI, building document Q&A systems, or integrating LLMs with external knowledge bases.
Use when the user asks for new work such as a feature, component, integration, service or dependency, especially when the request names a solution instead of a problem, the scope looks larger than the user may expect, or the motivation is that other products have it
Distributed training orchestration across clusters. Scales PyTorch/TensorFlow/HuggingFace from laptop to 1000s of nodes. Built-in hyperparameter tuning with Ray Tune, fault tolerance, elastic scaling. Use when training massive models across multiple machines or running distributed hyperparameter sweeps.
Keep README files in sync with the repo using symbolic HTML comments. Use this skill whenever making a repo change that could affect documentation — code (exported APIs, configuration, examples, CLI commands, dependencies), repo structure (adding/removing/renaming files, directories, packages, or other sibling entries referenced by a listing comment), or metadata (package.json, manifests). Also use when authoring or editing README files to add or update symbolic comments. Trigger on any change — code or structural — that touches a construct referenced by an HTML comment in a markdown file.
Validate Regle forms with schema libraries via `@regle/schemas` — Zod, Valibot, ArkType, and any Standard Schema spec library using `useRegleSchema`, `useRules`, `refineRules`, and `InferInput`. Use when a form's validation is defined by an external schema instead of `@regle/rules`. For native rules see regle-rules; for form setup see regle.
Revit 2026 API documentation lookup and reference. Triggered when the user asks about Revit API classes, methods, properties, or enum usage, looks up APIs related to Revit development, or needs to confirm API signatures and parameters. Covers: (1) looking up members and usage of a specific class/interface (2) searching for the APIs needed to implement a feature (3) viewing method signatures and parameter descriptions (4) understanding namespace structure and inheritance (5) confirming the correct way to call the Revit API. Keywords: RevitAPI, Autodesk.Revit, Element, Document, Transaction, FilteredElementCollector, ExternalCommand, ExternalApplication, Wall, Floor, FamilyInstance, FamilySymbol, Parameter, BuiltInParameter, BuiltInCategory, Selection, GeometryElement, Solid, Face, CurveLoop, Level, View, ViewSheet, XYZ, Line, Arc, IExternalCommand, IExternalApplication, UIApplication, UIDocument, Ribbon, PushButton, ExtensibleStorage, SubTransaction, TransactionGroup, ElementId, Reference, TaskDialog.
Transform photos, film stills, or new scene descriptions into the observed visual grammar of the 2026 animated film 《牛来》. Use when the user asks for 牛来风格、牛来正片风格、粗粝低模动物 3D、把真人画面改成牛来式动画, or requests iterative comparison against 《牛来》 references. Default to the rough film-3D mode; use the separate ink-poster mode only when the user explicitly asks for the movie poster look.
Use for Roblox Engine API lookup during implementation: finding the correct engine class or service, confirming properties, methods, events, callbacks, datatypes, enums, globals, and built-in libraries, and verifying parameter, return, and property usage for a known engine task. Prefer this skill when the problem is referential rather than architectural.
Use for Roblox persistent data and cross-server state design: choosing between DataStoreService, OrderedDataStore, MemoryStoreService, and MessagingService; designing save and load flows, schema shape, versioning, metadata, retries, quotas, observability, and concurrency-safe coordination across servers.
Use when errors occur deep in execution and you need to trace back to find the original trigger - systematically traces bugs backward through call stack, adding instrumentation when needed, to identify source of invalid data or incorrect behavior
Guidance for scaffolding new Rust projects. Use when: (1) starting a new Rust project or workspace, (2) configuring Cargo.toml best practices, (3) setting up CI pipelines for Rust, (4) organizing a multi-crate workspace, (5) configuring clippy, rustfmt, and linting.
Covers the RWKV (Receptance Weighted Key Value) architecture, an RNN/Transformer hybrid with O(n) inference and no KV cache, including RWKV-7, its parallel GPT-mode training and sequential RNN-mode inference, state passing, fine-tuning with DeepSpeed, and CUDA kernel setup. Use when generating text token by token with constant memory, processing very long contexts of 100K+ tokens, fine-tuning an RWKV model, comparing RWKV memory and speed against Transformers, or debugging RWKV state handling, loading, or out-of-memory errors. Prefer a standard Transformer when peak accuracy matters more than memory, and Mamba for state-space models.
RYUX - design intelligence for AI agents and designers. Tell it what you are doing and it routes to Analyze (understand an existing interface), Design (create or improve UI and UX in Figma, pen.dev, or mockups), Build (implement in code), Critique (find what to change first, with evidence, or run a heuristic evaluation), or QA (verify a build against the design), reads only the knowledge the task needs (product, UX, UI, interaction, forms, content, accessibility, responsive, design system, frontend, anti-slop), and closes with quality gates. Use for any UI, UX, copy, frontend, design review, or design analysis task, and for product visuals: illustration, 3D, imagery, icons, logos, and motion, even when the request does not name RYUX.
Interprets buying signals and prioritize accounts for outreach. Use when drowning in intent data and not sure what matters, can't tell which accounts to prioritize, buying signals aren't translating to pipeline, job change alerts are noisy, intent topics feel too broad, lead scores don't match actual deal quality, or not sure how to act on signals in outreach. Do NOT use for building prospect lists (use /sales-prospect-list), enriching contacts (use /sales-enrich), ZoomInfo-specific intent config (use /sales-zoominfo), or general Apollo platform help (use /sales-apollo).
Perform codebase analysis and architecture mapping as the first phase of a security assessment. Explores the tech stack, frameworks, entry points, data flows, and trust boundaries. Outputs sast/architecture.md. Run this before any vulnerability detection skill. Use when asked to analyze a codebase for security or when sast/architecture.md does not yet exist.
Configure Static Application Security Testing (SAST) tools for automated vulnerability detection in application code. Use when setting up security scanning, implementing DevSecOps practices, or automating code vulnerability detection.
Detect path traversal vulnerabilities in a codebase using a three-phase approach: recon (find file-loading sinks with dynamic paths), batched verify (trace user input and mitigations in parallel subagents, 3 sinks each), and merge (consolidate batch results). Requires sast/architecture.md (run sast-analysis first). Outputs findings to sast/pathtraversal-results.md. Use when asked to find path traversal, directory traversal, or file disclosure bugs.
This skill MUST be used when applying SCAMPER (奔馳法) to generate or review innovation ideas for products, services, processes, business models, or concepts. It SHOULD also be used for 創意發想、創新思考、產品改良、服務優化、創意激盪、系統化創新 when the user wants to explore multiple creative angles for an existing target, even without naming the framework.
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.