Skills
98,873 packaged capabilities and playbooks for agents. Install any of them with one command.
All
Cloud GPU processing via RunPod serverless. Use when setting up RunPod endpoints, deploying Docker images, managing GPU resources, troubleshooting endpoint issues, or understanding costs. Covers all 5 toolkit images (qwen-edit, realesrgan, propainter, sadtalker, qwen3-tts).
The scriptwriter of the creator's videos - short-form scripts (YouTube Shorts, TikTok, Instagram Reels, 30 to 60 s) in their own voice from video-agent/voice.md, each on a concept proven by the niche's outliers, with 3 scored hooks from 22 formulas, a re-hook, triggers, offer proximity, quality gates, timing at their pace and visual notes, then a humanize pass (invisible characters, AI words in English and French, AI sentence shapes, facts kept). Modes - the week (default), one topic, a funnel or divisive script, a making-of, hook printer, rewrite, humanize, and viral (what over-performs in their niche, scored against each account's median, YouTube free with yt-dlp, TikTok and Instagram through Treg, into video-agent/swipe.md). Use when the user asks for scripts, ideas, this week's videos, a funnel or divisive (clivant) script, a making-of, a hook, a better hook, to rewrite a script, what goes viral, viral hooks, what works in my niche, competitor research, swipe file, outliers, or to humanize any text (sounds like AI, make it sound like me, a caption before posting).
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
Find the reels that are working right now in the user's niche, rank them by how much each outperformed its own account, name the hook formula each used, and turn that into a swipe file they can shoot from. Use when the user says “find viral videos,” “what's working right now,” “what are people posting in my niche,” “reverse engineer this account,” “build me a swipe file,” “why is this reel doing numbers,” or asks what to make next without evidence to guide the answer.
Draft process letters and bid instructions for sell-side M&A processes. Covers initial indication of interest (IOI) instructions, final bid procedures, and management meeting logistics. Triggers on “process letter”, “bid instructions”, “IOI letter”, “bid procedures”, “final round letter”, or “management meeting invite”.
Framework for state-of-the-art sentence, text, and image embeddings. Provides 5000+ pre-trained models for semantic similarity, clustering, and retrieval. Supports multilingual, domain-specific, and multimodal models. Use for generating embeddings for RAG, semantic search, or similarity tasks. Best for production embedding generation.
Configures trace spans, defines custom metrics, sets up log exporters, and optimizes sampling strategies for OpenTelemetry instrumentation. Use when instrumenting applications with traces, metrics, or logs. Triggers on requests for observability, telemetry, tracing, metrics collection, logging integration, or OTel setup.
Implement Terraform provider ephemeral resources with the Plugin Framework: the Open/Renew/Close lifecycle, ephemeral schema design, registration via EphemeralResources, renewal for expiring credentials, and how ephemeral values flow into write-only attributes and provider configuration. Use when adding an ephemeral resource, exposing secrets/tokens/certificates that must never persist in state or plan, deciding between an ephemeral resource and a data source, or wiring short-lived credentials from one provider into another.
OpenAI's model connecting vision and language. Enables zero-shot image classification, image-text matching, and cross-modal retrieval. Trained on 400M image-text pairs. Use for image search, content moderation, or vision-language tasks without fine-tuning. Best for general-purpose image understanding.
Fast tokenizers optimized for research and production. Rust-based implementation tokenizes 1GB in <20 seconds. Supports BPE, WordPiece, and Unigram algorithms. Train custom vocabularies, track alignments, handle padding/truncation. Integrates seamlessly with transformers. Use when you need high-performance tokenization or custom tokenizer training.
Write the Instagram caption—the line that survives the "... more" cut, the body, the single ask, the search terms, and the three hashtags—and lint it before it goes out. Use when the user says "write the caption", "caption this", "what do I put in the description", has a reel or a carousel ready and needs the text, or asks about hashtags.
Build an Instagram carousel - the cover that earns the swipe, slide-by-slide copy, and the 1080x1350 files to upload. Use when the user says "carousel", "slides", "swipe post", "turn this into a carousel", or has a list-shaped or step-shaped idea that would die as a single image.
Open Graph and social sharing image design with platform specs, text placement, and branding. Covers OG meta tags, Twitter cards, LinkedIn previews, and dynamic generation. Use for: social sharing images, blog thumbnails, link previews, social cards. Triggers: og image, open graph, social sharing image, twitter card, social card, link preview image, og meta, sharing preview, social thumbnail, meta image, og:image, twitter:image, linkedin preview
Migrate Terraform provider resources and data sources from Plugin SDKv2 to the Plugin Framework: muxing both plugins in one provider (terraform-plugin-mux, tf5to6server), per-resource migration workflow, SDKv2-to-Framework schema mapping (ForceNew, ValidateFunc, DiffSuppressFunc, Default, Timeouts, blocks), null-vs-zero-value behavioral traps, and state-compatibility verification. Use when converting or translating SDKv2 resources to the Framework, setting up a muxed provider server, deciding whether a resource should be migrated at all, or debugging plan diffs and state errors that appeared after a migration.
Discover and install related skills from inference.sh skill registry. Helps find complementary skills for your AI workflow. Use for: skill discovery, workflow expansion, capability exploration. Triggers: related skills, find skills, skill discovery, complementary skills, expand workflow, more capabilities, similar skills, skill suggestions
Creative coding skills for Claude Code, claude.ai and Cowork. Motion, micro-interactions and design systems across Web (GSAP, Motion, Three.js, CSS), Android (Jetpack Compose) and Apple (SwiftUI). Proposes an interaction thesis you approve before it writes any code. Anti-AI-slop.
Guides development of Fusion portal shells — scaffolding, module configuration, app loading, routing, header/context integration, analytics, and deployment using the Fusion Framework CLI portal commands. USE FOR: create portal, scaffold portal, configure portal modules, portal app loading, portal routing, portal header, context selector, portal analytics, portal telemetry, portal manifest, ffc portal dev, portal deployment, embed apps in portal. DO NOT USE FOR: app-level feature development (use fusion-app-react-dev), backend service changes, Fusion Help Center integration, skill authoring.
Generate and conversationally edit short videos with stable Google Gemini Omni Flash (`gemini-omni-1.1-flash`). Use when: (1) iterating on a clip with natural-language edits instead of regenerating ("make the phone invisible, keep everything else the same"), (2) generating 3-10s 720p clips with synthesized audio, rendered on-screen text, or timecoded beats, (3) binding reference images to roles with <FIRST_FRAME>/<IMAGE_REF_N> prompt tags, (4) editing an existing uploaded video. Accessed via the `gemini_omni_video` tool using the project's GEMINI_API_KEY/GOOGLE_API_KEY — the same key as Imagen and Google TTS
Extend context windows of transformer models using RoPE, YaRN, ALiBi, and position interpolation techniques. Use when processing long documents (32k-128k+ tokens), extending pre-trained models beyond original context limits, or implementing efficient positional encodings. Covers rotary embeddings, attention biases, interpolation methods, and extrapolation strategies for LLMs.
Build, run, and visualize multi-step AI generation workflows. The AI architect translates natural language descriptions into connected node graphs — chain image generation, video creation, enhancement, and editing into automated pipelines.
Educational GPT implementation in ~300 lines. Reproduces GPT-2 (124M) on OpenWebText. Clean, hackable code for learning transformers. By Andrej Karpathy. Perfect for understanding GPT architecture from scratch. Train on Shakespeare (CPU) or OpenWebText (multi-GPU).
Adds PyTorch FSDP2 (fully_shard) to training scripts with correct init, sharding, mixed precision/offload config, and distributed checkpointing. Use when models exceed single-GPU memory or when you need DTensor-based sharding with DeviceMesh.
Convert PRDs to beads for ralph-tui execution using beads-rust (br CLI). Creates an epic with child beads for each user story. Use when you have a PRD and want to use ralph-tui with beads-rust as the task source. Triggers on: create beads, convert prd to beads, beads for ralph, ralph beads, br beads
Deploy, import, interact with, and monitor smart contracts using Circle Smart Contract Platform APIs. Supports bytecode deployment, template contracts (ERC-20/721/1155/Airdrop), ABI-based read/write calls, and webhook event monitoring. Keywords: contract deployment, smart contract, ABI interactions, template contracts, event monitoring, contract webhooks, bytecode, ERC-1155, ERC-20, ERC-721.
Activation-aware weight quantization for 4-bit LLM compression with 3x speedup and minimal accuracy loss. Use when deploying large models (7B-70B) on limited GPU memory, when you need faster inference than GPTQ with better accuracy preservation, or for instruction-tuned and multimodal models. MLSys 2024 Best Paper Award winner.
A workflow for creating viral Douyin scripts. The full process, from analyzing competitor videos to generating scripts: download Douyin videos → Gemini video analysis → extract viral formulas → generate scripts and storyboards → review for AI-like phrasing. Use this skill when users mention “Douyin script,” “viral content breakdown,” “competitor analysis,” “livestream commerce script,” “Qianchuan creative,” “product seeding script,” “video breakdown,” or “Douyin video analysis.”
Use when creating a new OpenCLI adapter from scratch, adding support for a new website or platform, exploring a site's API endpoints via browser DevTools, or when a user asks to automatically generate a CLI for a website (e.g. "帮我生成 xxx.com 的 cli"). Covers automated generation, API discovery workflow, authentication strategy selection, TS adapter writing, and testing.
Install, load, and unload SAS packages with the SAS Packages Framework (SPF) and the SASPAC archive — package fileref rules, %installPackage / %loadPackage / %helpPackage / %unloadPackage, and version pinning. Use when installing a package such as sasjscore or debugging an installation that reports rc=20006 or a missing package zip
Use this skill when making a Phaser 4 game responsive or handling display scaling. Covers ScaleManager, scale modes (FIT, RESIZE, EXPAND, ENVELOP), auto-center, fullscreen, and browser resize handling. Triggers on ScaleManager, responsive, resize, fullscreen, FIT, scale mode.
UI/UX taste for apps — dashboards, lists, tables, forms, settings, modals, picker pages. By default it works like a designer: brief, 2–3 wireframes, and it only builds once the user picks. Other routes when the request is explicit: inspect the existing UI ("take a look", "review"), compare the new version with the original ("what's better"), rebuild while keeping the brand ("keep the brand") or follow the skill's taste ("drop the old style"), refactor without changing the look, just build it, design system first ("UI kit", "component library"), a simple logo, or fix one component. Flat by default; can also do glass, gradients, elevation, dark backgrounds. Landing pages, online stores and blogs aren't taught yet — it still builds them, but warns first. Use when building, redoing or fixing an app's interface, when looking at an app screenshot or link, or when it says "make the UI pretty", "build a screen", "design", "redo the UX", "looks messy", "build a page", "design this screen", "redesign", "make it look good", "ecommerce", "ui-ux", "evon".
Salesforce Data Cloud product orchestrator for connect→prepare→harmonize→segment→act workflows. TRIGGER when: user needs a multi-step Data Cloud pipeline, asks to set up or troubleshoot Data Cloud across phases, manages data spaces or data kits, or wants a cross-phase `sf data360` workflow. DO NOT TRIGGER when: work is isolated to a single phase (use the matching sf-datacloud-* skill), the task is STDM/session tracing/parquet telemetry (use sf-ai-agentforce-observability), standard CRM SOQL (use sf-soql), or Apex implementation (use sf-apex).
Detect texture-atlas regions and map each source part to its own UV rectangle or projected UV surface. Use when a texture pack must fit a Blender model 1:1, when textures look off/stretched, when alpha decals or lightmaps are provided, or before exporting a textured GLB for a brand mascot/logo.
This skill should be used when the user asks to "add MCP App support to my web app", "turn my web app into a hybrid MCP App", "make my web page work as an MCP App too", "wrap my existing UI as an MCP App", "convert iframe embed to MCP App", "turn my SPA into an MCP App", or needs to add MCP App support to an existing web application while keeping it working standalone. Provides guidance for analyzing existing web apps and creating a hybrid web + MCP App with server-side tool and resource registration.
Creative coding skills for Claude Code, claude.ai and Cowork. Motion, micro-interactions and design systems across Web (GSAP, Motion, Three.js, CSS), Android (Jetpack Compose) and Apple (SwiftUI). Proposes an interaction thesis you approve before it writes any code. Anti-AI-slop.
Extract structured data from LLM responses with Pydantic validation, retry failed extractions automatically, parse complex JSON with type safety, and stream partial results with Instructor - battle-tested structured output library
Train Mixture of Experts (MoE) models using DeepSpeed or HuggingFace. Use when training large-scale models with limited compute (5× cost reduction vs dense models), implementing sparse architectures like Mixtral 8x7B or DeepSeek-V3, or scaling model capacity without proportional compute increase. Covers MoE architectures, routing mechanisms, load balancing, expert parallelism, and inference optimization.
Clarify ambiguous requirements through focused dialogue before implementation. Use when requirements are unclear, features are complex (>2 days), or involve cross-team coordination. Ask two core questions - Why? (YAGNI check) and Simpler? (KISS check) - to ensure clarity before coding.
Provides guidance for mechanistic interpretability research using TransformerLens to inspect and manipulate transformer internals via HookPoints and activation caching. Use when reverse-engineering model algorithms, studying attention patterns, or performing activation patching experiments.
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.