Browse skills

Uber's enterprise security system for AI coding agents, covering observability, security benchmarking and threat detection.

Microsoft's policy enforcement, identity, sandboxing and audit logging for autonomous AI agents.

Production-grade engineering skills that make coding agents follow senior-engineer steps: spec, plan, build, test, review and ship.

An open book on building AI agents, from fundamentals to engineering practice, with runnable code for each chapter.

15 marketing skills with parallel subagents. Audit a website, write copy, build email sequences and produce client-ready PDF reports.

14 sales skills and 5 parallel agents. Research prospects, qualify leads with BANT and MEDDIC, write outreach and prepare for meetings.

Runs 70B models on a single 4 GB GPU without quantization, by keeping only one layer on the GPU at a time.

An agent skill that turns a codebase or system description into a polished, interactive architecture diagram in one HTML file.

A suite of MCP servers from AWS Labs for AWS docs, the AWS API, infrastructure as code, support and many individual services.

Control Blender with an LLM. Model objects, build scenes and change materials by describing what you want.

Turns a technical book or document into a structured Claude Code skill that your agent loads only when it needs it.

The popular open-source library for AI agents that operate a real browser to click, type and finish web tasks.

A CLI that reads your local Claude Code logs and shows token use and estimated cost by day, month and session.

The official Chrome DevTools server for coding agents. Control and inspect a live Chrome with console, network and performance tools.

Anthropic's agentic coding tool that lives in your terminal, understands your codebase and handles routine tasks and git workflows.

Route Claude Code and other coding agents to the model providers you choose, with fail-over and per-task routing.

An Anthropic GitHub Action that uses Claude to review the changes in each pull request for security vulnerabilities.

Record the talking head and Claude Code does the rest, from the cut and visuals to the thumbnail and the upload.

Persistent memory for coding agents. It records what the agent did in each session and feeds the useful parts into the next one.

An open-source memory platform that gives agents long-term memory across sessions through a self-hosted knowledge graph.

Pulls up-to-date, version-specific library docs and code examples straight into your agent's prompt.

An open-source, LLM-friendly web crawler and scraper that turns pages into clean Markdown and structured data.

A DeepSeek-native coding agent for the terminal, built around prefix-cache stability to keep long sessions cheap.

Gives Claude Desktop terminal control, file system search and diff-based file editing on your own computer.

Stanford's framework for programming, not prompting, language models, with optimizers that tune your prompts for you.

antirez's focused engine for running DeepSeek V4 Flash and a few other large open models locally on Metal, CUDA and ROCm.

A large, battle-tested kit of agents, skills, hooks and commands that tunes Claude Code for quality, security and cost.

The Anthropic skill for opening, fixing and building spreadsheets with working formulas and clean formatting.

The fast, Pythonic way to build MCP servers and clients, now a full MCP application framework from Prefect.

Gives Cursor and other coding agents the layout data from your Figma files, so they build designs more accurately than from screenshots.

The web context API that searches, scrapes and interacts with pages at scale, returning clean Markdown or structured data.

The official Firecrawl MCP server. Search, scrape and crawl the live web from any MCP agent and get clean content back.

Microsoft's chart language that lets AI agents produce polished charts from short, human-editable specs.

A local proxy that runs Claude Code, Codex and other agents on free-tier providers, falling back automatically when one goes down.

GitHub's official MCP server. Let agents read code, manage issues and pull requests, and check workflow runs.

Control Gmail, Calendar, Drive, Docs, Sheets, Slides and more with natural language from any MCP client.

An open deep-research agent that searches the web and your local files and writes a detailed report with citations.

Type /graphify in your coding assistant and it maps code, docs, PDFs, images and video into a knowledge graph you can query.

A design skill for Claude Code, Cursor and Codex that refuses to make interfaces look AI-generated.

Nous Research's self-improving agent, which creates skills from experience, improves them in use and remembers you across sessions.

HeyGen's framework that turns HTML, CSS and seekable animations into deterministic MP4 videos, built for agents.

An MCP server for Kubernetes and OpenShift that talks to the Kubernetes API directly, with no kubectl needed.

A Python framework for building realtime voice and video AI agents, with telephony, job scheduling and MCP support built in.

One command to find which open-source LLMs actually run on your hardware, scored for fit, speed, quality and context.

A Microsoft Research library that compresses long prompts by dropping low-value tokens, so the model reads less and costs less.

Lightricks' open audio-video model that generates synchronized video and sound from one prompt, with an official LoRA trainer.

Microsoft's lightweight Python tool that converts PDFs, Office files and more into clean Markdown for LLMs.

The Anthropic skill that guides Claude through designing and building high-quality MCP servers in Python or TypeScript.

The official tool for testing and debugging MCP servers, with a web UI, a CLI and a terminal UI.

The official reference MCP servers, such as Filesystem, Git, Fetch and Memory, from the Model Context Protocol project.

Train a 64M-parameter language model from scratch in about two hours, with every step written in plain PyTorch.

Modular's open platform for AI development and deployment: the MAX framework, the Mojo language and an OpenAI-compatible server.

A 45M-parameter tool-calling model in a single 14 MB binary that runs a session in about 28 MB of RAM.

The official Notion MCP server. Let your agent search, read and update the pages and databases in your workspace.

A terminal coding agent with the IDE wired in: hash-anchored edits, LSP-aware renames, subagents and a browser.

A free AI gateway with one OpenAI-compatible endpoint that routes across hundreds of providers, many with free tiers.

Alibaba's AI code review CLI that mixes deterministic pipelines with an LLM agent for precise, line-level comments on Git diffs.

An open-source agentic video production system that turns your coding assistant into a studio: research, script, assets, edit and render.

An open-source context database for agents that stores memories, resources and skills under a viking:// filesystem you can browse.

A desktop workspace for running a fleet of coding agents in parallel, each in its own git worktree, tracked in one place.

Microsoft's MCP server that lets an agent drive a real browser through the accessibility tree instead of screenshots.

The Anthropic skill for building and editing slide decks, and for reading the text out of existing presentations.

Test and compare prompts, agents and RAG pipelines, and red-team them for security problems, from the command line.

The official Remotion skills that teach coding agents the best practices for making videos with React.

Packs a whole repository into one AI-friendly file, with token counts, so you can hand a codebase to any LLM.

A CLI proxy that compresses shell output before your coding agent reads it, so every command costs fewer tokens.

An agent meta-harness for Claude Code and Codex that runs coordinated swarms of agents with shared memory.

Sentry's MCP server for coding agents. Pull real errors, stack traces and issues into the chat while you debug.

An MCP toolkit that gives coding agents IDE-like, symbol-level code search and editing instead of reading whole files.

A high-performance serving framework for LLMs and multimodal models, from a single GPU to large distributed clusters.

An autonomous AI pentester for web apps and APIs that only reports the vulnerabilities it can prove with a working exploit.

The official Anthropic skill for writing new Agent Skills, improving existing ones and measuring them with evals.

Record yourself doing a task once, and this Microsoft desktop app turns the session into a reusable agent skill.

Microsoft's optimizer that trains agent skills like neural networks, with epochs and validation gates, without touching weights.

NVIDIA's security scanner for agent skills that finds vulnerabilities, prompt injection and malicious patterns before you install a skill.

Fine-tune LLMs from one YAML file; layer streaming trains an 8B model on a 4 GB laptop GPU.

GitHub's toolkit for spec-driven development. Write the spec first, then let the agent plan, split the work into tasks and build.

Stripe's official SDKs, plugins and skills for connecting payments and billing to LLMs and agent frameworks.

Open-source AI penetration testers that run your app, find vulnerabilities and prove each one with a working exploit.

Connect your Supabase projects to Cursor, Claude, Windsurf and other assistants to design tables, run queries and more.

A complete software-development method for coding agents, built from composable skills like planning, TDD and debugging.

NVIDIA's Rust proxy that routes each LLM call to the cheapest model that can still do the job, keeping native OpenAI and Anthropic APIs.

A team-level memory hub that turns conversations, docs and code into reusable agent memory: chat memory, skills, a wiki and a code graph.

HashiCorp's official MCP server that connects agents to the Terraform Registry and HCP Terraform for accurate infrastructure code.

Google Research's pretrained time-series foundation model, so you can forecast in a few lines of Python instead of training your own.

Security research, vulnerability detection and audit skills for Claude Code from the security firm Trail of Bits.

A Rust vector index with Python bindings that, the project says, fits 10M documents in about 4 GB and searches faster than FAISS.

Run, fine-tune and serve LLMs and diffusion models on your own hardware, with an OpenAI-compatible API for your agents.

A high-throughput, memory-efficient engine for serving LLMs, with PagedAttention, continuous batching and prefix caching.

The Anthropic skill that lets Claude create, read and edit real Word documents and templates.