Skills
101,174 packaged capabilities and playbooks for agents. Install any of them with one command.
All
Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.
Design and implement purposeful, performant, and accessible motion in Flutter. Use for transitions, micro-interactions, gesture feedback, animated state changes, shared elements, list choreography, reduced-motion behavior, or diagnosing animation quality and performance.
Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, artifacts, posters, or applications (examples include websites, landing pages, dashboards, React components, HTML/CSS layouts, or when styling/beautifying any web UI). Generates creative, polished code and UI design that avoids generic AI aesthetics
Review code changes or requirements documents with evidence-backed findings and a verdict. Use for integration seams, public contracts, security, migrations, or a requested deep review, or in gate mode for a machine-readable merge verdict; keep local task reviews focused.
Gangtise open platform investment research data service. Provides knowledge base search (broker research reports, announcements, meeting minutes), company earnings forecasts, economic data indicator queries, conference call lists, and research report source download. Use when users ask about stocks, companies, industries, macroeconomic data, or research reports.
Generate complete Markdown documentation for a Figma component — anatomy/layer tree, design tokens (colors, spacing, typography), states/variants matrix, accessibility notes, content guidelines, and optional code-parity + YAML frontmatter. Use when the user wants a docs page or handoff spec for a component or component set. Triggers: 'document this component', 'generate component docs/spec', 'create a docs page for the Button', 'write up the anatomy and variants', 'component handoff doc from Figma', 'turn this component into Markdown docs'. Reads the node tree, bound variables/tokens, and designer annotations, then the agent assembles Markdown. Requires the Figma Desktop app (Plugin API).
Use when adding authentication to a project scaffolded with genesis — login, sessions, SSO, admin users. Asks for an auth tier (basic/standard/strict), derives the mechanism from requirements, and implements it in place following the project's existing conventions.
Queries 20+ bioinformatics databases and analysis services through the gget CLI and Python package, covering Ensembl gene search, info and sequences (ref, search, info, seq), BLAST, BLAT, MUSCLE, DIAMOND, PDB, AlphaFold, ELM, ARCHS4, CELLxGENE, Enrichr, Bgee, OpenTargets, cBioPortal, COSMIC, viral sequence downloads (virus), and 8cube mouse specificity and expression data. Use when looking up gene or transcript details, running a quick BLAST/BLAT search, fetching AlphaFold or PDB structures, running enrichment analysis on a gene list, downloading viral sequences with filters, or exploring disease and drug associations interactively. Not for batch processing or fine-grained BLAST control (use biopython) or multi-database Python pipelines (use bioservices).
Software development engine for coding. A production-grade skills system for web apps and mobile architecture, APIs, databases, web, mobile, AI, security, UX, SaaS, and ERP engineering. Encodes reusable engineering judgment through layered baseline and specialist skills to help generate secure, scalable, maintainable, high-performance software systems
Google Search Console data analysis and site management via google-search-console-cli. Use when the user wants to check search performance, analyze queries, inspect URL index status, manage sitemaps, or view site properties. Triggers: "Search Console", "search performance", "search queries", "GSC", "index status", "URL inspection", "sitemap", "search analytics", "impressions", "CTR", "search appearance".
Google Search Console API integration for search analytics, URL inspection, sitemap management, and site verification. Use when working with search performance data, checking indexing status, managing sitemaps, or analyzing SEO metrics.
Quantizes LLMs to 4-bit (also 3-bit) with GPTQ using group-wise quantization (group size 128 by default), via AutoGPTQ and transformers. Covers loading pre-quantized GPTQ models, quantizing your own model, choosing group size, selecting ExLlamaV2, Marlin, or Triton kernels, and QLoRA fine-tuning with PEFT. Use when fitting 70B-class models onto limited or consumer GPUs, cutting memory about 4x versus FP16, speeding up inference, finding pre-quantized checkpoints on HuggingFace, or fine-tuning a quantized model with LoRA. For slightly better accuracy on newer GPUs use AWQ instead, and for simple 8-bit or on-the-fly quantization use bitsandbytes.
A high-intensity questioning interview about a plan or design (adversarial design review / grill session) that exposes flawed assumptions and missing constraints, while maintaining a domain model (glossary and ADRs) throughout. Invoke manually with /grill-me.
Use this skill whenever the user wants to use Grok or xAI through the grok-cli command line, including chat, X search, image generation/editing, video generation/editing/extension, text-to-speech, speech-to-text, OAuth login/status, or local usage stats. This skill should also be used when grok-cli may need to be installed first; it runs the user's requested grok-cli command directly when possible, repairs install/auth/permission problems only after an actual failure, retries the original task after recovery, and returns Grok/grok-cli results without assistant-side rewriting unless the user explicitly asks for transformation.
Guides GRPO (Group Relative Policy Optimization) fine-tuning of language models with the TRL library, including GRPOTrainer configuration, composing multiple reward functions (correctness, format, length, style), dataset prep in chat format, Unsloth setup, LoRA merging, and monitoring reward, reward_std and KL. Use when training a model to follow a strict output format such as XML or JSON. Use when teaching verifiable tasks like math or code with objective correctness rewards. Use when improving chain-of-thought reasoning with custom reward functions. Use when debugging flat rewards, mode collapse, or OOM in GRPO runs. Do not use for plain supervised fine-tuning (use SFT) or when high-quality preference pairs exist (use DPO or PPO).
Compares current live GTM container against last known state from gtm-implementation-log.json. Shows what tags, triggers, and variables have been added, deleted, or modified since the last skill run. Read-only, no changes made. Prevents accidentally overwriting manual GTM changes. Trigger on - "what changed in GTM", "gtm diff", "check for manual changes", "gtm drift", "what's different in my container", "has anything changed".
Constrains LLM output during generation with Guidance (Microsoft Research), using regex, select() choices, context-free grammars, token healing, and @guidance functions, with Anthropic, OpenAI, Transformers, and llama.cpp backends. Use when you need generated text to match a regex or fixed format such as dates, emails, or IDs. Use when you need guaranteed valid JSON, XML, or code from a model. Use when building multi-step generation workflows or ReAct-style agents with Python control flow. Use when classifying text into fixed categories with select(). Use when running local models and wanting grammar constraints. Not for Pydantic validation with automatic retries; use Instructor for that.
Write prompts for MiniMax-H3-Image text-to-image generation and image editing, and optionally render them through the MiniMax API. Use when the user wants an image such as a poster, infographic, chart, product shot, UI screen, character, scene, storyboard or grid, wants to improve an image prompt, or wants to edit, combine or restyle reference images with MiniMax-H3-Image.
Harness Engineering phase two: set up cross-session state management so the agent stops forgetting everything between conversations. Creates three files: tasks.json (task list), progress.md (progress log), and init.sh (environment setup script). Use immediately when the user says "set up task management", "make the agent remember progress", "create tasks.json", "keep state across sessions", "the agent never remembers what it did last time", "create a progress file", or "initialize state management". Prerequisite: harness-step1 and harness-step2 are complete (the project has AGENTS.md and a docs/ knowledge base).
Context compression for AI agents — 60–95% fewer tokens by compressing tool outputs, logs, files, and conversation history before they reach the model. MCP tools headroom_compress/retrieve/stats plus an optional traffic-compressing proxy. Wraps the open-source headroom-ai; on Windows install via WSL2.
Reasoning skill for HEDIS measure specification, enrollment logic, exclusion evaluation, NCQA audit requirements, and care gap prioritization. Use when the user asks about HEDIS measure definitions, denominator/numerator/exclusion logic, continuous enrollment rules, Star Rating impact, or care gap closure strategies.
Half-Quadratic Quantization for LLMs without calibration data. Use when quantizing models to 4/3/2-bit precision without needing calibration datasets, for fast quantization workflows, or when deploying with vLLM or HuggingFace Transformers
Wraps existing PyTorch training scripts with HuggingFace Accelerate (Accelerator class, accelerate config, accelerate launch) so the same code runs on CPU, single GPU, multi-GPU, multi-node, TPU, or Apple MPS. Covers device placement, FP16/BF16/FP8 mixed precision, gradient accumulation, distributed checkpointing, and switching between DDP, DeepSpeed ZeRO, FSDP, and Megatron backends. Use when converting a single-GPU script to multi-GPU, enabling mixed precision, configuring DeepSpeed ZeRO or FSDP, or setting up gradient accumulation. Use when one script must run on different hardware. Not for callback-heavy training loops (use PyTorch Lightning) or multi-node orchestration with hyperparameter tuning (use Ray Train).
When the user wants to brainstorm YouTube video ideas, score existing ideas, validate topic choices, or batch-generate content concepts. Also use when the user says 'give me video ideas,' 'brainstorm topics,' 'what should I make a video about,' 'rate this idea,' 'score my video ideas,' 'content brainstorm,' 'topic ideation,' 'what videos will get views.' For title optimization, see title-craft. For scheduling ideas, see content-calendar. For viewer insight to inform ideas, see audience-research.
Create or continue IELTS Academic Reading practice, produce printable passage sessions, and review answers using private persistent progress and passage deduplication. Use for IELTS reading study and continuation, not unrelated English conversation practice.
Generate images with Gemini (default) or fal.ai FLUX.2 klein 4B (--cheap for fast/low-cost). Generate videos with Grok Imagine (default) or fal.ai LTX-2 (--cheap). Use for: create image, generate visual, AI image generation, poster, video generation.
Query and download public cancer imaging data from NCI Imaging Data Commons. Invoke for any question about IDC collections, cancer imaging datasets, DICOM data access, radiology (CT, MR, PET) or pathology AI training sets, metadata queries, visualization, or license checks — even when the user doesn't explicitly mention "IDC". No authentication required.
API design and implementation across REST, GraphQL, gRPC, and tRPC patterns. Use when building backend services, public APIs, or service-to-service communication. Covers REST frameworks (FastAPI, Axum, Gin, Hono), GraphQL libraries (Strawberry, async-graphql, gqlgen, Pothos), gRPC (Tonic, Connect-Go), tRPC for TypeScript, pagination strategies (cursor-based, offset-based), rate limiting, caching, versioning, and OpenAPI documentation generation. Includes frontend integration patterns for forms, tables, dashboards, and ai-chat skills.
Implements and trains LLMs using Lightning AI's LitGPT with 20+ pretrained architectures (Llama, Gemma, Phi, Qwen, Mistral). Use when need clean model implementations, educational understanding of architectures, or production fine-tuning with LoRA/QLoRA. Single-file implementations, no abstraction layers.
Analyze increases and decreases in holdings by directors, supervisors, senior managers, and major shareholders in the A-share market to identify companies with significant management confidence signals. Use this skill when the user asks about director and executive increases, major shareholder buys, insider trading analysis, management buying signals, or shareholder holding changes.
Extracts structured, validated data from LLM responses using the Instructor Python library with Pydantic response models, including nested models, enums, custom validators, automatic retries with validation error feedback, and streaming of partial objects or iterables. Works with Anthropic, OpenAI, and local Ollama models. Use when pulling typed fields or entities out of free text, when classifying text into fixed categories, when an LLM must return JSON that passes schema validation, when failed extractions need automatic retry, or when streaming partial structured results. Not for prompt optimization (use DSPy) or building multi-step chains (use LangChain).
Pattern library for Copilot Studio agent design. Contains proven and recommended implementation patterns with YAML examples. Used by the Advisor agent to suggest patterns and by the Author agent to reference YAML structure during implementation. USE FOR: design guidance, pattern suggestions, review against best practices, troubleshooting known pitfalls, YAML reference during implementation. DO NOT USE FOR: general YAML schema reference (use int-reference), topic creation mechanics (use new-topic).
Accepts any text, link, document, or question and upgrades "explain this" into a real learning loop: first judge whether it is worth learning and how deep to go, then use minimal verification to force out genuine understanding, and dynamically generate checkpoint HTML when needed. Use this skill proactively whenever the user clearly wants to "understand it", "master it", "guide me through learning it", or "help me verify whether I really understand it", rather than just a summary, rewrite, or ordinary explanation. It also applies when the user wants to export learning cards, checkpoint pages, retrospective pages, or learning status pages.
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.