Skills
101,174 packaged capabilities and playbooks for agents. Install any of them with one command.
All
Stress-test a crash analysis, hypothesis, or technical conclusion by launching three adversarial agents in parallel — one defends, one attacks, one hunts new evidence — and resolving disputes with physical proof instead of argument. Always use this skill when the analysis has 2+ competing explanations that both seem plausible, when a conclusion relies on a single log sample or inferred chain longer than 3 steps, or when someone says \"are you sure?\", \"that doesn't sound right\", or challenges an assumption. Triggers: 辩论、求真、挑战假设、质疑分析、multi-agent-debate, \"debate this\", \"challenge my analysis\", \"find holes in\", \"I'm not sure this is right\", \"verify this\".
Expert MCP (Model Context Protocol) orchestration with n8n workflow automation. Master bidirectional MCP integration, expose n8n workflows as AI agent tools, consume MCP servers in workflows, build agentic systems, orchestrate multi-agent workflows, and create production-ready AI-powered automation pipelines with Claude Code integration.
Provides nanoGPT, Karpathy's minimal PyTorch GPT implementation (model.py and train.py), with workflows for training character-level Shakespeare on CPU, reproducing GPT-2 124M on OpenWebText with multi-GPU torchrun, fine-tuning pretrained GPT-2 checkpoints, and training on custom text. Use when learning how GPT and transformer blocks work from scratch. Use when running a small training experiment on CPU or a single GPU. Use when modifying a transformer variant in plain PyTorch. Use when preparing character-level or BPE binary datasets. Use when troubleshooting out-of-memory or slow training in nanoGPT. Not for production or large-scale distributed training; use HuggingFace Transformers or Megatron-LM instead.
Interactive viewer for microscopy. Displays 2D/3D/4D arrays as Image, Labels, Points, Shapes, Tracks layers; supports annotation, plugin analysis, headless screenshots. Core visualization for Python bioimage workflows. Use ImageJ/FIJI for macro processing; napari for Python-native interactive visualization and DL segmentation review.
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.
Genre pack: neo-brutalism for the web — loud consumer energy with hard discipline. Bright warm paper, ink-black 2.5px borders on everything, hard offset shadows, one dominant hot accent plus one supporting accent, chunky pills and stickers, playful copy. Use for consumer apps, agencies, crypto/tech novelty brands, and anything wanting maximum personality. Stack on the design-taste core skill.
Apply these opinionated Next.js conventions whenever writing or reviewing App Router code (14+): Server Components by default, where the 'use client' boundary belongs, the caching defaults that changed in Next 15, tag-based revalidation, treating Server Actions as public endpoints, the Metadata API, and next/image and next/font.
Comprehensive Next.js SEO optimization skill for metadata, sitemaps, robots.txt, structured data, and Core Web Vitals. Use when building or auditing Next.js applications for SEO, implementing metadata API, generating sitemaps, adding Open Graph tags, creating JSON-LD structured data, optimizing for search engines, or improving search rankings. Works with Next.js 13+ App Router and Pages Router.
Integrate compiled NixieFX particle effects into PixiJS or Three.js applications using the public nixie-fx package. Use when an agent needs to install the runtime, load an out/vfx export bundle, implement texture or material providers, mount a renderer, drive effects from a game loop, inspect runtime diagnostics, or clean up NixieFX resources.
Node.js and Fastify performance, security, and best practices from Nearform. This skill should be used when writing, reviewing, or refactoring Node.js/Fastify code to ensure optimal performance, security, and maintainability. Triggers on tasks involving Node.js APIs, Fastify routes, database operations, error handling, or backend optimization.
Use when the user explicitly asks to "write/generate an NSFC budget justification", "write a budget justification", "generate budget.tex / budget.pdf", or "write an NSFC budget justification". Based on the user's proposal body or supplementary materials, outputs a submittable budget justification LaTeX project and renders `budget.pdf`. If the user has not specified a working directory, must pause and first ask them to specify one. ⚠️ Not applicable: the user only wants to understand budget principles; the user only wants budget table figures without writing a justification; or the user is on the 2026 Young Scientist A/B/C flat-fee (包干制) default with no need for a budget justification.
Use when the user explicitly asks to "generate an NSFC schematic diagram/mechanism diagram/schematic diagram/mechanism diagram" or needs to turn the research mechanism, algorithm architecture, or module relationships from a proposal into a "editable + embeddable in documents" diagram. By default it outputs editable source files (`.drawio`) and rendered files (`.pdf`/`.svg`/`.png`); when the user proactively mentions the Nano Banana image model, it can switch to PNG-only mode and is compatible with Gemini and OpenAI `gpt-image-2`. ⚠️ Not applicable: the user only wants to polish body text (should directly rewrite the text), only wants to change the format/size of an existing image (should use an image processing skill), or has no clear "schematic/mechanism diagram" intent.
Create and edit Obsidian Bases (.base files) with views, filters, formulas, and summaries. Use when working with .base files, creating database-like views of notes, or when the user mentions Bases, table views, card views, filters, or formulas in Obsidian.
Create and edit Obsidian Flavored Markdown with wikilinks, embeds, callouts, properties, and other Obsidian-specific syntax. Use when working with .md files in Obsidian, or when the user mentions wikilinks, callouts, frontmatter, tags, embeds, or Obsidian notes.
Resolve free-text scientific labels to ontology term IDs and validate existing CURIEs against the EBI Ontology Lookup Service (OLS4). Also look up prefixes in Bioregistry, resolve compact identifiers via Identifiers.org, map lab shorthand with ZOOMA, and build Ontobee term pages. Use whenever an ontology identifier must be produced or checked - annotating tissue, cell type, disease, phenotype, assay, chemical, organism, sex, or developmental stage fields; preparing metadata for GEO, ENA, BioSamples, CELLxGENE, HCA, or ISA-Tab submission; auditing a metadata table of term IDs; checking whether a term is obsolete and what replaced it; or deciding HPO vs HP. Triggers include "ontology term", "ontology ID", "CURIE", "controlled vocabulary", "UBERON", "CL:", "MONDO", "HPO", "EFO", "ChEBI", "NCBITaxon", "GO term", "PATO", "Zooma", "Bioregistry", "Identifiers.org", "Ontobee", "annotate this tissue/cell type/disease", and any request to emit or verify an identifier shaped like PREFIX:0001234.
High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.
Full inbox management across all channels — WhatsApp (wacli), Email (Gmail MCP), Slack (MCP), Telegram (user-auth MCP), Discord (webhook + REST read), Notion (MCP — comments, mentions, assigned tasks). Scans the full inbox (not just unread), identifies messages needing replies, and archives handled conversations.
Styles Rails views with RoleModel's Optics design system — its components, `--op-` design tokens, and theme overrides. Use when writing or reviewing CSS, styling a view or component, picking a color, spacing, font, radius, or shadow value, fixing hard-coded CSS values, adding a project token, or theming an app (brand colors, fonts, color scales, dark mode). Also use when the user says things like "style this", "make this look like the design", "what token should I use", "fix the hard-coded colors", "change the brand color", or "add dark mode".
Use when each language SDK lives in a separate repository. Covers cross-repo workflow dispatch, PR status reporting, PR reconciliation on merge/close. Triggers on "multi-repo SDK", "separate SDK repositories", "cross-repo workflows", "SDK PR synchronization", "spec repo triggers SDK repos".
Use this skill to organize Android test source sets — `src/test/`, `src/androidTest/`, the community `src/sharedTest/` convention, and the modern KMP-style `androidHostTest` / `androidDeviceTest` split that Compose itself adopted after AGP 7.2 broke the classic `sourceSets` wiring. Includes the `testImplementation` / `androidTestImplementation` / `debugImplementation` configuration matrix per Google's `/training/testing/local-tests` and `/training/testing/instrumented-tests` pages. Use when the user asks "where do I put my tests", "test or androidTest", "sharedTest setup", "AGP 7.2 broke my sourceSets", "androidHostTest vs androidDeviceTest", "testImplementation vs androidTestImplementation vs debugImplementation", or "why is my Robolectric test in androidTest".
Assess OpenTelemetry package and Collector upgrades across ecosystems. Use when choosing or validating API, SDK, instrumentation, exporter, Collector distribution, component, image, chart, Operator, OCB build, or configuration versions, including compatibility, required changes, telemetry and runtime behavior, and rollout risk.
太奶讀論文 — a centenarian grandmother uses Traditional Chinese, everyday metaphors, and anime references to help you understand academic papers. Always use this skill when the user provides a paper PDF, an arXiv link, or pasted paper text and wants to understand its content. Trigger words include: 讀論文、解釋論文、看不懂、幫我理解這篇、這篇在說什麼、paper reading、explain this paper. Suitable for an intuitive guided reading of any academic paper.
Academic paper review skill — perform an in-depth paper review through a structured four-step process, covering critical review, score prediction, key point refinement, and formal review output. Use this skill whenever the user needs to review a paper, simulate a reviewer's reaction, assess whether a paper can be accepted, or help judge a paper's strengths and weaknesses. Trigger words include: review this, review my paper, what would reviewers say, can this get in, paper review, give a score, find weaknesses. Applies to review simulation and evaluation of any academic paper.
Structured peer review of manuscripts and grants. 7-stage evaluation: initial assessment, section review, statistical rigor, reproducibility, figure integrity, ethics, writing. Covers CONSORT/STROBE/PRISMA and report structure. For evidence quality see scientific-critical-thinking; scoring see scholar-evaluation.
Display Oracle philosophy — the 5 Principles + Rule 6. Use when user asks about principles, "nothing deleted", "patterns over intentions", Oracle philosophy, or needs alignment check. Do NOT trigger for "who are you" (use /who-are-you), "what is oracle" (use /about-oracle), or session status questions.
PHP web full-chain white-box code security audit pipeline (multi-file orchestration). Serves as the master orchestration reference, calling each sub-audit skill by sink type (`php-route-mapper/php-auth-audit/php-route-tracer/php-*-audit`) and reusing unified output and severity standards.
CSRF audit tool for PHP web source code. Identifies whether state-changing endpoints are protected against CSRF, traces token generation, validation and bypass conditions, and outputs exploitability ratings, PoCs and remediation advice (omission not allowed).
Business logic vulnerability audit tool for PHP web source code. Identifies logic flaws other than authentication/authorization: Mass Assignment, flow bypass, race conditions, state machine flaws, payment/permission timing vulnerabilities, and more, and outputs evidence chains, ratings, PoCs and remediation advice (omission not allowed).
Adds or modifies a Physical AI Studio policy under library/src/physicalai/policies. Use when creating a new policy family with the config/model/policy split, registering it in the get_policy factory and package exports, or keeping a policy compatible with Lightning training and export. Covers Pi0.5, Pi0, ACT, GR00T, SmolVLA, and LeRobot-wrapped policies.
Comprehensive skill for project planning and prompt engineering. Covers hierarchical plans (briefs, roadmaps, phases), Claude-to-Claude meta-prompts, and multi-stage workflows. Use when: planning, prompt creation, agentic pipeline work, project roadmap, meta-prompts, research to implement workflow.
Generate leftover PlantUML types Mermaid cannot do easily (Salt wireframes, use case, timing, ArchiMate, nwdiag, WBS) and convert them to PNG/SVG. Use when asked for a wireframe, PlantUML image export, extract puml from markdown, or prepare Confluence uploads. Class, ER, state, sequence, and C4 default to design-doc-mermaid on GitHub wiki.
A/B experiment design skill. Starting from the experiment goal, output a complete experiment plan (hypothesis, groups, metric system, sample size estimation, stop-loss rules, decision rules). Triggers: the user mentions "A/B testing", "AB experiment", "experiment design", "controlled experiment", "traffic-split experiment", "gradual rollout plan", "experiment plan", "sample size calculation", "significance testing", "experiment evaluation", "p-value", "confidence interval", "MDE", "statistical power", "experiment duration", "stop-loss rules", and similar keywords. Also applies when: the user provides experiment goals / changeable points / data scale / duration constraints and requires an experiment plan; the user asks to evaluate whether an existing experiment design is reasonable; the user asks to define experiment decision criteria and decision rules. Typical input: experiment goal + changeable points + current data scale + acceptable experiment duration. Not suitable for: pure data analysis (use pm-analytics), pure tracking design (use tracking-spec-writer), pure PRD writing (use pm-prd-writer).
Use when: you need to analyze user feedback and comments, assess user satisfaction, collect product improvement suggestions, or perform sentiment analysis. Do NOT use when: there is very little user feedback data, or only collection is needed without analysis
Use when: you need to analyze the user conversion funnel, diagnose churn causes, generate optimization suggestions, or improve user conversion rates. Do NOT use when: the user conversion path is already very clear, or only single-metric monitoring is needed without full-funnel analysis
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.