AI Skills, Agents & MCP

24 self-contained Claude Code skills, each distilled from one real engineering failure, with sanitised case studies and pre-flight checklists.

Web directory that indexes Claude Code plugin marketplaces themselves, so you can discover sources rather than individual plugins.

MCP server providing Rhino and Grasshopper automation so agents can build parametric 3D geometry.

MCP server from 21st.dev that generates production-ready React UI components from natural language descriptions.

Open-source visual editor that lets you style and edit a running React app directly and writes the code changes back.

tldraw project that turns a hand-drawn interface sketch on a canvas into working HTML you can iterate on.

Image prompt library packaged as an agent skill so a coding agent can pick and adapt prompts across models.

Ten-thousand-plus curated image prompts with preview renders and multilingual variants, aimed at Gemini image generation.

Hosted MCP endpoint that feeds shadcn/ui component and design-system context into Cursor, Claude Code and Windsurf.

Visual whiteboard that lets you annotate screenshots and feed the marked-up feedback straight back into the coding agent.

Open-sourced skill set behind ClaudeKit covering research, planning, shipping and repo hygiene for Claude Code projects.

MCP server that converts Figma files into simplified layout JSON so coding agents generate accurate markup from designs.

Skills that produce editorial-quality diagrams in HTML and SVG with deliberate typography and layout decisions.

The largest curated index of Claude Code resources: guides, skills, statuslines, hooks, clients, security tools and CLAUDE.md files.

Skill that attaches an agent to a real Chrome instance over CDP and adds a perception layer over the page.

MCP server that answers design-system questions about components, tokens, patterns and best practice from a curated corpus.

Official MCP server for Penpot enabling bidirectional design workflows against an open-source design tool.

MCP server exposing fifty-plus Photoshop automation tools for layers, filters, selections, text and export.

Design-engineering skill that critiques UI work against a scoreable rubric before and after implementation.

TypeScript starter for MCP servers with example tools, resources, prompts, multi-client install scripts and npm publish workflows.

Curated and tested ComfyUI workflows for image, video and voice generation with documented node graphs.

Official MCP server exposing the Magic UI animated component registry so agents install real components instead of inventing them.

Lightweight meta-prompting system for Claude Code that structures work into planning and execution passes without heavy scaffolding.

MCP server with a WebSocket bridge that lets an agent create and modify Figma elements from natural language.

Curated system-prompt collection framed as a builder's reference for people designing their own agents.

MCP server and CLI that injects current, version-specific library documentation and code examples into an agent's prompt.

Configuration framework adding thirty slash commands, twenty specialist agents, seven behavioural modes and eight MCP integrations to Claude Code.

Tool that converts existing Express or FastAPI applications into MCP servers without rewriting the API surface.

Real-time guardrails that detect and block secret leakage in agent output and tool calls.

Native macOS app for assembling code context into prompts, with file trees, diffs and an MCP server agents can call.

Claude Code plugin that converts raw terminal output and logs into styled, readable HTML explanations.

Fourteen-hundred image prompts scraped from X and ranked by engagement, portable across NanoBanana, Midjourney and GPT Image.

MCP server that translates design tokens between Tailwind, Figma, CSS variables, Material 3 and SwiftUI.

Curated system prompts and tool definitions harvested specifically from production AI coding agents.

Archive of extracted system prompts from ChatGPT, Claude, Grok, Gemini, Manus, v0, Lovable and Cursor.

Hook that intercepts destructive shell commands before an autonomous agent executes them.

Community-maintained list of plugins extending Claude Code with commands, agents, hooks and MCP servers.

Planning layer that lets several coding agents share specs, tasks and decision history inside one workspace before they write code.

MCP server bringing shadcn-vue and Tailwind component generation to Vue developers via an agent.

Official Expo skill pack teaching agents how to scaffold, build and debug Expo and React Native applications correctly.

Large transparency archive of system prompts across ChatGPT, Claude, Gemini, Grok, Perplexity, Cursor, Lovable and Replit.

Production-oriented Claude Code skill marketplace repo with installable skills for day-to-day development workflows.

Metaphor-driven product naming skill that runs structured divergence and convergence rather than dumping a word list.

Security-engineering skills that drive CodeQL and Semgrep static analysis from an agent, written by a real appsec firm.

Template teaching the PRP workflow — global rules, feature specs, example patterns and /generate-prp and /execute-prp commands.

Anthropic's official skill repo: docx, pptx, xlsx, pdf, canvas-design, algorithmic-art, mcp-builder, webapp-testing and skill-creator.

Opinionated cursor rules repo that also ships a maintained AGENTS.md translation of the same rule set.

Tool that crawls a documentation website and compiles it into a packaged Claude Skill with references and examples.

Anthropic's curated directory of vetted Claude Code plugins, used as the default marketplace source.

Workspace of databases, kanban boards, pages and calendars that Claude, Cursor and other agents read and write over MCP.

Browser automation harness that lets a coding agent open, drive and visually verify the app it just wrote.

Statusline that streams live World Cup football scores into your Claude Code terminal.

MCP server with fifty-seven FreeCAD tools and headless Docker execution for automated CAD generation.

Regularly updated extractions of Anthropic, OpenAI, Google and xAI system prompts including Claude Code and Codex.

Thinking-framework skills with a meta-router that picks the right reasoning approach for the problem at hand.

MCP server pairing Playwright with full Chrome DevTools access for autonomous debugging, performance and security inspection.

Playground repository with runnable side-by-side implementations comparing modern open-source agent frameworks on the same example tasks.

Bridge that lets coding agents drive arbitrary desktop software beyond the terminal and browser.

Slash-command system that interrogates your codebase, asks two rounds of yes/no questions, then writes a concrete requirements document.

Project template that adds pre-commit hooks, CI, linting, type checks, file-length limits and test-drift inventories to LLM-assisted work.

Skill that teaches Claude to build correct, idiomatic d3.js data visualisations instead of generic charting library output.

MCP server that drives Google NotebookLM so agents can query and build notebooks programmatically.

Curated list of Claude Code slash commands, subagents, MCP servers and hooks packaged as installable plugins.

Advanced .mdc rules for Cursor covering agent memory management and multi-step development workflows.

Microsoft's MCP server that drives Playwright using the accessibility tree rather than screenshots for deterministic browser control.

Modular real-world ComfyUI workflows organised by task, built to be composed rather than used monolithically.

Sprawling toolkit bundling agents, skills, commands, plugins, hooks, rules, templates and MCP configs in one repo.

Battle-tested Claude Code skill library covering brainstorming, TDD, debugging and code review, installed as a plugin marketplace.

Statusline showing live rate-limit consumption, token spend and cost so you can see budget burn in real time.

Experimental staging repo where new agent-skill techniques get prototyped in public before graduating into the main superpowers library.

Skill that generates favicons, app icons, OG images and social media assets from a single source graphic.

Highly customisable Claude Code statusline with powerline rendering, themes and configurable widgets.

All parts of the Claude Code system prompt, versioned and updated on each release for diffing.

Reference project demonstrating every Claude Code hook event with runnable example implementations and explanations.

Alternative curated list of Claude Skills, resources and tooling maintained by the Composio team with workflow-oriented grouping.

Large collection of .cursorrules and .mdc rule files organised by framework and language for Cursor and Windsurf.

Web and mobile progressive web app for driving Claude Code sessions remotely from a phone or browser.

Ready-to-use scientific computing skills spanning data analysis, bioinformatics and research pipelines for Claude agents.

Drop-in CLAUDE.md distilling Karpathy-style coding guidelines into behavioural rules for coding agents.

CLI that runs coding agents inside hardware-isolated virtual machines for safe autonomous operation.

Curated real-world AGENTS.md files, templates, guides and tooling for Codex-style and cross-vendor agents.

Claude Code skill bundle for modern web work, specialising in 3D graphics, animation and interactive web experiences.

Browsable web directory of Claude Skills and plugins with search and install instructions, rather than a flat markdown list.

Reference site documenting one hundred and ninety-plus patterns for building and operating coding agents.

Starter kit of commands for priming context, reviewing work, syncing docs and blocking secrets across Claude Code sessions.

Reference library of hook-based skill activation, modular skills, ten specialist agents and documentation patterns that survive context resets.

Fifty-plus motion graphics and animation skills for agents, covering easing, timelines, transitions and video-style sequences.

Machine-generated .mdc rule files derived from library documentation, one rule set per package.

Chrome automation MCP combining Playwright and CDP with network interception, HAR record and replay, and accessibility testing.

Collection of system prompts and full tool definitions from Augment, Claude Code, Cursor, Devin, Kiro and VSCode Agent.

Converts Claude Code session transcripts into an interactive HTML replay you can scrub and share.

Skill with a detection engine that strips AI-isms, filler transitions and hedging from generated prose.

Automatically refreshed daily index of MCP servers, clients and frameworks with activity signals baked in.

Focused collection of Claude Code plugins bundling specialised agents, skills and slash commands for development workflows.

Curated index of official and community Claude Skills with per-entry notes, plus authoring tools and marketplaces for skills.

Catalogue of agentic workflow patterns documented with Mermaid diagrams and implementation notes.

Curated index of AI agent frameworks, platforms and protocols with categorisation by use case.

Statusline that reacts to what the agent is doing with kaomoji expressions and mood states.

Curated collection of Claude Code hooks for event-driven automation across the agent lifecycle.

Walkthrough of the Ralph technique: running a coding agent in a repeating loop against a spec until every item is finished.

Long-running curated index of MCP servers organised by domain, widely mirrored across the ecosystem.

Plugin that turns CSS selectors into generated visual documentation of a component's states and variants.

Community gallery of Claude Code status lines where every card preview is rendered from the real script before you copy it.

Browsable catalogue of Claude Code plugins and marketplace collections with install commands and categorisation.

Official shadcn skills giving agents accurate component context and enforcing shadcn conventions during generation.

Claude Code plugin pane plus a local Codex dashboard showing the agent and subagent tree, live status, token use and the current request's stage-grouped task flow.

Self-hosted Go sandbox service for agent-written code that reports the isolation tier of every run, refuses runs below a floor and brokers API calls without exposing credentials.

Claude Code skill that builds standalone interactive isometric biology figures, like dopamine synapses or GLP-1 on a beta cell, with toggles and a citation for every number.

Read-only MCP server on your own Cloudflare Workers account that links as a WhatsApp device so agents can search and read your personal chats.

Deterministic Python CLI plus nine Claude Code skills that audit how well ChatGPT, Perplexity and AI Overviews can crawl, quote and cite a site.

End-to-end encrypted shared scratchpad where Claude Code, Codex, Cursor and other agents leave notes, files and handoffs for each other across machines.

Claude Code skill that turns a one-line machine description into a photoreal three.js simulation in one HTML file, driven by a small real physics solver.

Agent skill that turns a one-line brief into a concept-led site with GSAP, Lenis and WebGL, then screenshots and scores it until each category hits 8/10.

Single Rust binary that builds a tree-sitter code graph and persistent memory your coding agents query over MCP, a port of Graphify.

Claude Code skill that routes any two places and bakes real OpenStreetMap buildings, terrain and landmarks into a scrollable pseudo-3D pixel-art fly-through page.

MCP server plus Figma plugin that lets Claude Code, Cursor and Codex read and edit the open Figma Desktop file as XML with CSS, tokens and components.

Claude Code skill plus runtime that takes a live UI component apart into paint layers, states and tokens, measured from the real DOM rather than redrawn.

Agent skill that turns a short JSON spec into animated, explorable diagrams in one HTML file, exportable as SVG, PNG, GIF or MP4.

Claude Code plugin that shows images and plays videos inside the Ghostty terminal: design options side by side, Playwright test recordings, before and after QA.

Claude Code mod that picks reasoning effort per prompt with a Haiku judge, and adds a bar showing context fill, prompt cache warmth and one-click compaction.

macOS CLI plus Claude Code and Codex skill that lets agents draw big click-through arrows, boxes and text over any window, then vanish.

Interface design skill for AI agents that explores several distinct visual directions side by side, then refines the chosen one against real screenshots.

Open-source parametric 3D CAD app in pure Rust with sketches, constraints and a feature timeline, running on desktop, web, and via a built-in MCP server.

Claude Code plugin with writing rules that cut reply length roughly in half and strip mannered prose from replies, docs, commits and subagents.

Agent skill that turns hard questions into a readable one-page offline HTML explainer, rendered by a bundled CLI from a short Markdown draft.

A free public tracker for Claude Code usage reset timing and status.

Open-source runtime layer for AI coding agents that intercepts shell, secrets, MCP, and installs locally.

PuppyOne keeps your project files and AI in one local-first workspace, so your assistants can understand the whole project, work across real files, and continue without starting from scratch.

A Product Hunt-style directory for AI coding plugins across Claude, Codex, Cursor, and MCP.

Tale is an open-source project workspace where teams assign tasks to AI agents in persistent sandboxes and review their deliverables together.

Marketplace of hook-based Claude Code plugins: context bar, usage meter, review watcher, Markdown preview, rm -rf blast radius, Spotify now-playing and more.

Hosted MCP server for ASCII Magic that lets Claude, Cursor or any MCP client turn images into ASCII art, halftone, LEGO, voxel and 100+ looks.

Local video studio for coding agents: describe a video in one sentence and your agent directs motion graphics, voice-over, music and captions, rendered on your own machine.

Saves decisions, rules and failed approaches from Claude Code chats to a JEVMEM.md in your repo, recalls them next session, and checks commands against saved rules.

Linux command runner with a time limit that guarantees no process an agent's command started survives, using cgroup v2 kill or a subreaper plus pidfds.

Agent skill that opens your running local app as live iframe screens on a design canvas, so you can compare routes and push visual edits back through your coding agent.

Eleven Claude skills that reverse-engineer an app's screens and data model, rebuild it clean-room, test it, then mine user complaints to fix what the original gets wrong.

Claude Code skill that clones a website from a URL: Playwright captures three viewports, extracts fonts, tokens and keyframes, rebuilds section by section, then grades the clone by pixel diff.

Claude Code plugin of rules, hooks and skills that makes the agent answer ten questions about a change against real code, scope its work, and prove it before claiming done.

Claude Code plugin that opens a Doom deathmatch pane beside the transcript while Claude works, with everyone else waiting on Claude, and hands you back when it finishes.

Curated list of 230 open-source skills and toolkits that let Claude Code, Codex and other agents make video, grouped by type with GIF previews and security grades.

Skills, a CLI and a fal MCP server that let coding agents mod PC games: detect the engine, read the code, build the mod, generate assets and test it in-game.

Agent skill that runs a full logo process with SVG craft, audit tools and a searchable library of 1,400+ reference marks.

Local-first Rust CLI and MCP that catches intent conflicts between parallel coding agents before they write conflicting code.

Claude Agent Skill that builds continuous-take product launch films with measured continuity, motion blur, and a rejection oracle.

Claude skill that strips AI tells from Chinese drafts by merging three rule sets and adjudicating seven contradictions.

Workflow DSL that turns repeatable agent work into inspectable steps with Jev decisions and optional full agent loops.

Prebuilt and DIY agents that use Jev, Laya, or Cua-S1 System-1 decision models for browser, desktop, games, and routing.

Claude Skill with 700-plus cinematography terms, prompt formula, and ready combos for Veo, Kling, Sora, and friends.

Menu-bar and TUI app that switches every local coding agent's model through one local gateway and shared catalog.

Claude Code skills that turn a Lichess game plus think-aloud audio into Stockfish-checked annotated PGN, HTML board, and narrated video.

Self-hosted versioned skills library with HTTP MCP, scoped client keys, immutable revisions, and optional Jev recommendations.

Turns the Figma desktop app into a local MCP server so agents get the full Plugin API with no cloud quota.

Four-hundred-plus rule library that steers Claude Code and other agents away from generic-looking websites.

Claude Skill for liquid-glass and glassmorphism UI with material tiers, accessible contrast and platform fallbacks.

Claude Code status line that shows session context size, next-turn cost, and rate-limit headroom before you type.

Human-curated marketplace for paid and free agent skills across coding, writing, design, research, and ops.

Claude Code and Codex skill that turns any topic into a Remotion motion-graphics explainer with TTS, subtitles, and QC.

Research workspace where chat, files, PDFs, and notebooks share one dock, and the agent can spawn custom panes when needed.

Open-source extractor that pulls chat histories from Claude Code, Cursor, Windsurf, Aider, Cline, and related coding assistants.

Agent skill that turns PDFs, books, and papers into interactive learning webpages with quizzes, notes, and source traceability.

Local-first desktop inbox for background AI agents, with unread results, approval queues, and bring-your-own model providers.

Multi-agent orchestrator that turns one goal into an ordered graph of coding-agent tasks with per-task runners and verified markers.

Agent skill that turns a prompt or reference into a holographic 3D flash-card page with Blender scene, layered PNGs and Three.js viewer.

Local desktop app that detects, configures, and runs coding-agent CLIs like Claude Code, Codex, Gemini, and OpenCode in one workspace.

Agent-first CLI framework with one grammar and canonical names across SaaS providers, loading command schemas only when asked.

Agent skill that writes codebase-specific contracts, ratchets, and CI gates so AI-written code cannot quietly skip the rules.

Dev-server MCP bridge so agents can inspect and edit live Vue, React, Svelte, or Solid app state from inside the running page.

Agent skill that maps architecture and planned file scope into a standalone HTML page before any code edit starts.

Agent skill that turns an idea into an explorAble Three.js or Blender scene, with hundreds of recipes and runnable studies.

Self-hosted, read-only MCP server that lets Claude or Ollama query your own bank accounts via Enable Banking open banking.

Compiler-accurate Find Usages and symbol graph for C# 15 that coding agents can query instead of grepping the repo.

Claude Code skill that turns one topic into a ranked YouTube research workbook with channels, transcripts, hooks, and replay hotspots.

Curated catalogue of Italian MCP servers for public data, legal-tech, invoicing, and public administration, with readiness scores.

Native macOS workspace that keeps one local knowledge graph shared across Claude Code, Codex, OpenCode, and Cursor.

SEO and GEO MCP with Search Console, GA4, PageSpeed, structured data, llms.txt, and WordPress tools for Claude, Codex, and Cursor.

Open-source simulation testing and monitoring platform for voice agents, with BYO STT/LLM/TTS keys and zero markup on inference.

Read-only CLI that inventories every agent, MCP server, plugin, and AI extension on a machine and what each can touch.

Pure-Go git-native agent memory using OKF markdown, in-memory BM25 search under 300µs, and an embedded MCP server.

Agent design skill with research, moodboard, tokens, register dials, and a slop linter so UI stops looking generic.

Bridges Claude Design into Cursor, Codex, and other agents so designs sync from real components and tokens both ways.

Agentic loop where coding agents draw in p5.js, then refine via adversarial visual review — skills for Claude Code and Codex.

MCP server and CLI that makes deterministic tools judge every AI claim against ground truth, with evidence and session rollover that keeps facts intact.

Open-source runtime guardrails that allow, warn, or block every LLM and MCP tool call before it commits, with a hash-chained audit trail.

Local Rust daemon that schedules agent loops (Claude, Codex, Hermes, Pi) with REST, MCP, and a web UI — no cloud queue required.

Claude/Codex skill that turns place and culture photos into art-directed Chinese editorial posters with composition, type, and reference gates.

Agent OS for Blender 5.2: skills, verified bpy knowledge, AGENT_OK/FAIL contract, and a production gate for 3D-printable parts.

Rust SRE agent harness that investigates production incidents with scoped workers, disk-backed context, and human-gated remediations.

Self-hostable personal context MCP: one endpoint for accounts, memory, skills, vault, and audit that every agent you authorize can share.

Claude Code org chart of 16 independently installable departments and 172 skills, addressed as department:skill so names never collide.

Agent skills from Brad Frost covering design-system inspection, product health, DS adoption plans, and enforceable AI-usage limits.

Source-linked archive of documented incidents where AI coding agents and chatbots deleted data, leaked secrets or burned money, each with the lesson.

Pair of agent skills that build a slide deck from a repository, citing exact file line ranges and hashes so a script can flag slides that drifted.

Claude Skill that searches for existing open-source solutions before your agent writes code, vetting license, maintenance and dependency CVEs.

Agent skill that generates images by rendering HTML/CSS instead of calling an image model, with 32 named styles and zero-dependency PNG export.

Claude Code skill that builds scroll-driven sites using eight mutually exclusive page grammars, then screenshots its own scroll to grade contrast and dead scroll.

Claude Code skill for long-form generative art: hash-seeded PRNGs, trait and rarity tables, plus scripts that render contact sheets and check determinism.

Claude Code skill that reads your local session logs to explain where tokens went, why you hit a limit, and what to change.

Claude Code skill pack that rewrites agent output as plain-language reports, straight yes/no verdicts on completion, and numbered steps you can follow.

Rust binary giving agents web fetch, search and crawl plus scanned-PDF OCR as MCP tools, with no API keys or accounts.

Keyless MCP server exposing 42 tools over Overture Maps open data: place search, geocoding, routing, isochrones and area comparisons.

Local-only MCP server plus Svelte Flow viewer for planning system architecture as a typed, linted diagram before any code is written.

Paste a public GitHub repo of SKILL.md files and get a stable streamable-HTTP MCP URL for Claude Code, Cursor or Codex.

Agent skill packaging 18 production CSS transitions plus commands to audit a project for hardcoded durations and swap them for motion tokens.

MCP server giving agents six design-lookup tools: UI inspiration galleries, component code, shaders and motion, plus scraping a site's type and palette.

Two-skill agent library: shader-driven WebGL UI components with real fallbacks, and Tailwind container queries used in place of viewport breakpoints.

Local MCP server driving your real logged-in Chrome so agents can read and act on login-gated pages across X, Reddit, Zhihu, Xiaohongshu and Bilibili.

Agent-driven design canvas running beside Cursor, VS Code, Claude Code or Codex, storing designs in an open format that lives inside your repo.

Append-only SQLite store of typed facts for agents, exposed over MCP, returning a freshness and assurance verdict on every recalled belief.

macOS app that floats a strip of eight virtual LEDs beside the Dock, signalling whether coding agents are working, waiting, finished or broken.

Native macOS app built in Rust and GPUI that drives Claude Code, Codex, Cursor, OpenCode and Grok CLIs in one local timeline.

MCP server giving agents persistent memory as Open Knowledge Format markdown files on disk, indexed in SQLite FTS5 for fast lookups.

Desktop IDE running Claude Code, Codex, OpenCode, Cursor and Grok in parallel, each task isolated in its own git worktree.

Local read-only dashboard for token usage and reported cost across OpenCode, Claude Code, Codex and Cursor sessions, with per-call cache-miss drilldown.

Emacs shell for chatting with any ACP-driven agent — Gemini CLI, Claude Agent, Auggie, Mistral Vibe — inside a native buffer.

Agent skill that turns one concept into a sparse abstract poster using precise geometry, matte fields and a single spatial event.

Go deep-research CLI and MCP server with a hard spend ceiling, quote-verified claims, and local CSV analysis that never sends rows to the model.

LLM-friendly JSON design language with a validation core and a Figma adapter that turns the JSON into editable native Figma frames.

Library of DESIGN.md files capturing real brand design systems — colour, type, spacing — in a format coding agents can read directly.

Local-first canvas where graph edges define the context sent to an LLM; branch, prune and merge nodes, then inspect the exact message sequence before sending.

Agent skill that encodes Material Design 3 decision rules — semantic tokens, component choice by behaviour, adaptive width classes, accessibility — with progressive-disclosure reference files.

Agent skill teaching motion design judgement — timing, easing, choreography and Disney's 12 principles adapted for UI — independent of any animation library.

Claude Code skill encoding the ISO 24495-1 plain-language rules, rewriting text for reader outcomes in English and Traditional Chinese.

Agent skill for picking CJK typefaces: licence-tiered commercial-safe font list, scene-by-mood selection matrix, and CJK–Latin pairing rules.

Directory of Cursor plugins, MCP servers and framework rule files, ranked by what the community is installing.

Maintained index of seventy-five-plus high-quality Claude Code repositories with descriptions and categories.

Desktop platform where specialised frontend, backend and database agents collaborate to generate full applications from a plain description.

Design-judgment skill that teaches an agent refined visual principles so its UI output stops defaulting to generic templates.

MCP server exposing a local markdown second brain to agents with search and structured retrieval.

Forty-six-plus documented Claude Code techniques plus status-line scripts, conversation utilities and a bundled developer-experience plugin.