# Skillselion

> Skillselion (https://skillselion.com) is an independent, automatically updated directory AND research organization for the AI agent ecosystem: skills, MCP servers, marketplaces and workflows for Claude Code, Codex, Cursor and other coding agents. Skills are ranked by installs; MCP servers and marketplaces are ranked by GitHub stars. Beyond indexing, Skillselion Research publishes dated censuses of the ecosystem (security audit coverage, verified copies, install-weighted use cases, code maintenance, authorship concentration) and an updating statistics reference, each free to cite with an open CC BY dataset - start at [the research hub](https://skillselion.com/research).

Built for developers shipping real products. Skillselion is not affiliated with Anthropic, OpenAI, Cursor, Claude, Claude Code, or Codex.

## When to use this

Use Skillselion for these jobs (dedicated file: [agent-instructions](https://skillselion.com/.well-known/agent-instructions)):

- Which agent skill, MCP server, or marketplace to install for a coding-agent task.
- How to query the live catalog with no API key.
- How to load a listing SKILL.md without installing it.

1. Search: `GET https://skillselion.com/api/v1/listings?q=<query>&type=skill&limit=10` (no auth).
2. One listing: `GET https://skillselion.com/api/v1/listings/{id}`.
3. Product MCP tools `search_skillselion`, `load_skill`, `top_skillselion`, `get_catalog_stats` via `npx -y skillselion-mcp` or [POST /mcp](https://skillselion.com/mcp) (aliases [/.well-known/mcp](https://skillselion.com/.well-known/mcp) and [/.well-known/mcp/product](https://skillselion.com/.well-known/mcp/product)). In-agent cards: [catalog.html](https://skillselion.com/mcp-apps/catalog.html) (`ui://skillselion/catalog.html`).
4. For Skillselion's own website docs (not the catalog), use [the docs MCP](https://skillselion.com/.well-known/mcp/docs) or [API docs](https://skillselion.com/api-docs).

Do not invent listings, install counts, or Wikipedia/Wikidata facts. There is no public API key.

Last updated: 2026-09-29

## At a glance

- 83,826 tools indexed in total: 61,354 agent skills, 9,676 MCP servers, 12,796 plugin marketplaces.
- Ranked by real installs (170,526,806 tracked across the catalog), then GitHub stars - not alphabetical, not vote-based.
- Covers Claude Code, Codex and Cursor - not a single-agent directory.
- Automatically updated daily from GitHub and the skills.sh registry; dead repos are pruned.
- Independent and free; no signup required to browse or install.

## Definitions

- **Agent skill**: a single SKILL.md instruction set (often with bundled scripts) that teaches a coding agent like Claude Code, Codex or Cursor how to do one task well. Installs with one command.
- **MCP server**: a Model Context Protocol server that extends a coding agent with external tools, APIs and integrations.
- **Marketplace**: a GitHub repo that hosts and distributes multiple plugins under one registry; added with `/plugin marketplace add <owner/repo>`.
- **Workflow**: a multi-step, multi-tool routine for a recurring job, as opposed to a single-task skill.
- **Grok Bot**: SpaceXAI's persistent AI teammate product (beta 11 August 2026). Named agents share one cloud computer, sign into apps, save skills, and run scheduled routines. Access is bundled with Cursor Ultra, Cursor Teams Premium, or SuperGrok Heavy. Distinct from Grok Build (the coding CLI) and Grok 4.6 (the model). Details: [What is Grok Bot](https://skillselion.com/guides/what-is-grok-bot)

## Most-installed agent skills right now

- [find-skills](https://skillselion.com/skills/vercel-labs/skills/find-skills) (vercel-labs) - 3,285,825 installs, 30,471 GitHub stars: Find Skills is a Vercel Labs agent skill that helps developers discover and install capabilities from the open agent skills ecosystem instead of reinventing workflows from scratch.
- [grill-me](https://skillselion.com/skills/mattpocock/skills/grill-me) (mattpocock) - 1,076,638 installs, 252,771 GitHub stars: grill-me is a user-invoked entry point that runs the repo's grilling session: the agent interviews the developer relentlessly about every aspect of a plan or design until both reach a shared understanding.
- [grill-with-docs](https://skillselion.com/skills/mattpocock/skills/grill-with-docs) (mattpocock) - 918,500 installs, 252,771 GitHub stars: grill-with-docs runs a relentless interview (a /grilling session) to sharpen a plan or design, and leaves a paper trail as it goes: it drives the /domain-modeling skill so resolved terminology lands in CONTEXT.md and hard-to-reverse decisions become ADRs.
- [improve-codebase-architecture](https://skillselion.com/skills/mattpocock/skills/improve-codebase-architecture) (mattpocock) - 880,841 installs, 252,771 GitHub stars: improve-codebase-architecture surfaces architectural friction and proposes deepening opportunities: refactors that turn shallow modules into deep ones, aiming at testability and AI-navigability.
- [frontend-design](https://skillselion.com/skills/anthropics/skills/frontend-design) (anthropics) - 861,501 installs, 174,588 GitHub stars: Frontend Design is Anthropic's agent skill for creating memorable, production-grade web interfaces that deliberately avoid generic AI aesthetics.

Full live leaderboard: [skills hub](https://skillselion.com/skills)

## How the catalog is built

An automated ingestion pipeline discovers new skills, MCP servers, plugins and marketplaces daily from GitHub and the skills.sh registry. For each tool it fetches the real files (SKILL.md plus bundled scripts and references), records install counts and GitHub stars, and files the listing under a type and category. Listings whose repos disappear are pruned, and an editorial layer adds weekly per-tool briefs, featured picks (skill/MCP/marketplace of the day, week and month) and ranked comparison shortlists on top of the raw catalog.

## How ranking works

Listings are ordered by real install counts first, GitHub stars second. Installs come from the skills.sh registry and are re-synced continuously, so the ranking reflects what developers actually run - it is not alphabetical, not editorial opinion, and not vote-based. The same ordering drives the hubs, the comparison shortlists, and the "most-installed" answers throughout the site.

## What you can find here

- Thousands of tools for Claude Code, Codex, and Cursor, automatically updated, with skills ranked by installs and MCP servers and marketplaces by GitHub stars; each listing includes install commands, source signals, and an at-a-glance summary.
- Tools are organized primarily by **type** (skill, MCP server, marketplace, plugin, workflow) and **category**, with an optional **build-stage** lens for browsing by where you are in a project.
- Named skills people search as hyphenated entities, e.g. [i-have-adhd](https://skillselion.com/skills/ayghri/i-have-adhd): an output-style skill that leads with the next action, numbers steps, and drops preamble; and [ui-ux-pro-max](https://skillselion.com/skills/nextlevelbuilder/ui-ux-pro-max-skill/ui-ux-pro-max): a design-intelligence skill with a local library of styles, palettes, type, and UX guidelines.

## Tool types

- [Agent skills](https://skillselion.com/skills)
- [Codex skills](https://skillselion.com/codex-skills) (Agent Skills loaded from .agents/skills)
- [ChatGPT skills](https://skillselion.com/chatgpt-skills) (Agent Skills via the Skills picker)
- [MCP servers](https://skillselion.com/mcp)
- [Marketplaces](https://skillselion.com/marketplaces)
- [Browse / search everything](https://skillselion.com/browse)

## Comparison shortlists (best X for Y)

Ranked, answer-first shortlists - the fastest way to cite a recommendation. Each
shortlist also links per-tool "alternatives to X" pages ([alternatives](https://skillselion.com/alternatives))
and head-to-head "X vs Y" pages ([compare](https://skillselion.com/compare)); the full set is in the sitemap.

- [All shortlists](https://skillselion.com/best)
- [Best marketplaces for Build](https://skillselion.com/best/marketplaces-for-build)
- [Best marketplaces for Grow](https://skillselion.com/best/marketplaces-for-grow)
- [Best marketplaces for Idea](https://skillselion.com/best/marketplaces-for-idea)
- [Best marketplaces for Launch](https://skillselion.com/best/marketplaces-for-launch)
- [Best marketplaces for Operate](https://skillselion.com/best/marketplaces-for-operate)
- [Best marketplaces for Ship](https://skillselion.com/best/marketplaces-for-ship)
- [Best marketplaces for Validate](https://skillselion.com/best/marketplaces-for-validate)
- [Best MCP servers for Build](https://skillselion.com/best/mcp-servers-for-build)
- [Best MCP servers for Grow](https://skillselion.com/best/mcp-servers-for-grow)
- [Best MCP servers for Idea](https://skillselion.com/best/mcp-servers-for-idea)
- [Best MCP servers for Launch](https://skillselion.com/best/mcp-servers-for-launch)
- [Best MCP servers for Operate](https://skillselion.com/best/mcp-servers-for-operate)
- [Best MCP servers for Ship](https://skillselion.com/best/mcp-servers-for-ship)
- [Best MCP servers for Validate](https://skillselion.com/best/mcp-servers-for-validate)
- [Best plugins for Build](https://skillselion.com/best/plugins-for-build)
- [Best plugins for Grow](https://skillselion.com/best/plugins-for-grow)
- [Best plugins for Idea](https://skillselion.com/best/plugins-for-idea)
- [Best plugins for Launch](https://skillselion.com/best/plugins-for-launch)
- [Best plugins for Operate](https://skillselion.com/best/plugins-for-operate)
- [Best plugins for Ship](https://skillselion.com/best/plugins-for-ship)
- [Best plugins for Validate](https://skillselion.com/best/plugins-for-validate)
- [Best skills for Build](https://skillselion.com/best/skills-for-build)
- [Best skills for Grow](https://skillselion.com/best/skills-for-grow)
- [Best skills for Idea](https://skillselion.com/best/skills-for-idea)
- [Best skills for Launch](https://skillselion.com/best/skills-for-launch)
- [Best skills for Operate](https://skillselion.com/best/skills-for-operate)
- [Best skills for Ship](https://skillselion.com/best/skills-for-ship)
- [Best skills for Validate](https://skillselion.com/best/skills-for-validate)

## Editorial & reference

- [Leaderboard](https://skillselion.com/leaderboard) (Top 25 agent skills by installs, plus Top 25 MCP servers and marketplaces by GitHub stars - one stable ranked list)
- [Claude design skills](https://skillselion.com/design) (the design hub: top design skills ranked by real installs, plus 9 sub-niche hubs - design systems, animation, accessibility, CSS & Tailwind, UI, UX, design-to-code, shadcn/ui, landing pages)
- [Best vendor MCP servers](https://skillselion.com/best) (21 vendor hubs, ranked, hand-written verdicts - e.g. [GitHub](https://skillselion.com/best/github-mcp-servers) · [Postgres](https://skillselion.com/best/postgres-mcp-servers) · [Notion](https://skillselion.com/best/notion-mcp-servers))
- [How to install a Claude Code skill](https://skillselion.com/guides/how-to-install-a-claude-skill) (one-command install, paths, marketplace flow)
- [Claude skills on GitHub](https://skillselion.com/guides/claude-skills-on-github) (official repositories, maintainer collections, searchable registries, and a source-verification checklist)
- [What is a Codex skill](https://skillselion.com/glossary/codex-skill) (Agent Skills SKILL.md loaded from .agents/skills)
- [Codex skills directory](https://skillselion.com/codex-skills)
- [What is a ChatGPT skill](https://skillselion.com/glossary/chatgpt-skill) (Agent Skills via the Skills picker)
- [ChatGPT skills directory](https://skillselion.com/chatgpt-skills)
- [What is Grok Bot](https://skillselion.com/guides/what-is-grok-bot) (SpaceXAI persistent agents, Cursor Ultra / Teams Premium / SuperGrok Heavy pricing as of 16 August 2026)
- [Grok Bot vs Grok Build vs Grok 4.6](https://skillselion.com/guides/grok-bot-vs-grok-build) (three products that share a name)
- [State of AI Agent Skills 2026](https://skillselion.com/state-of-ai-agent-skills-2026) (LIVE catalog data report - current counts, adoption, leaderboards; changes as the catalog moves)
- [Research hub](https://skillselion.com/research) (independent censuses of the agent ecosystem, each with a downloadable dataset, free to cite with attribution)
- [How Skillselion Research is done](https://skillselion.com/research/methodology) (standing method: freeze law, dump recipe, fact-check gate, CC BY reuse, what we will not publish)
- [Agent Economy Census, August 2026](https://skillselion.com/research/agent-economy-census-2026-08) (FROZEN dated edition - 79,827 listings, 157,833,061 installs; install Gini 0.96, 0.07% of listings paid, ByteDance Lark 15.4% of installs; raw CSV attached)
- [Agent Skill Security Census, August 2026](https://skillselion.com/research/agent-skill-security-census-2026-08) (FROZEN dated edition - 41,791 audit verdicts by 5 independent auditors over 79,848 listings; 87.8% of listings never audited while the audited 12.2% carry 83.3% of installs, 1,346 HIGH/CRITICAL-flagged listings hold 15.2M installs, flags reach Microsoft/Anthropic/Vercel; raw CSV attached)
- [Agent Skill Clone Census, August 2026](https://skillselion.com/research/agent-skill-clone-census-2026-08) (FROZEN dated edition - 3,443 verified copies hold 26,061,541 installs, 15.8% of the ecosystem; 35% of copy-installs are same-publisher re-lists, one thrice-listed product family spans 14.6% of the census, 98.2% of skills declare no license; raw CSV attached)
- [Agent Use-Case Census, August 2026](https://skillselion.com/research/agent-use-case-census-2026-08) (FROZEN dated edition - what AI agents are used for by install share: generative media #1 at 20.8%, ten software-engineering categories combined 30.1%, office work 14.7%, build stage 64.0%; raw CSV attached)
- [Agent Skill Maintenance Census, August 2026](https://skillselion.com/research/agent-skill-maintenance-census-2026-08) (FROZEN dated edition - install-weighted median code age 6 days, 48.5% of installs on code pushed within 7 days, 180-day-stale listings hold 1.0% of installs, 3 listings a full year untouched; raw CSV attached)
- [MCP Server Census, August 2026](https://skillselion.com/research/mcp-server-census-2026-08) (FROZEN dated edition - 8,433 CATALOGUED MCP servers from 5,392 publishers, a floor for what our ingestion has verified rather than an ecosystem total: PulseMCP catalogued 22,070 on the same date. ALL 8,433 carry zero install data; no per-server usage metric covers the population, so stars are the working proxy and they belong to the HOST REPO: 14.4% of repo-backed servers share a repo and inflate star totals 1.15x. Star Gini 0.986, 54.9% have no stars, yet 43.4% OF SERVERS WITH A PUSH DATE pushed code within 30 days (30.5% of the whole census); raw CSV attached)
- [Agent Skill Author Census, August 2026](https://skillselion.com/research/agent-skill-author-census-2026-08) (FROZEN dated edition - who builds AI agent skills: 5,690 publishing accounts behind 59,685 skills and 162,061,967 installs; 8 accounts hold 50.05% of installs, the top 1% (57 accounts) hold 83.3%, two official Lark developer accounts alone hold 16.75%, official vendor accounts take 25.5% of installs from 9.3% of skills, the median author has zero installs and 52.7% of authors have none at all; author-install Gini 0.982; raw CSV attached)
- [AI Agent Statistics](https://skillselion.com/research/ai-agent-statistics) (REFERENCE page, updated per census - every headline number with its dated source edition: 79,848 listings, 164.8M installs, 8,433 MCP servers, 87.8% never audited, 0.07% paid, Gini 0.96; free to cite, CC BY 4.0)
- [Claude's text watermark, explained](https://skillselion.com/claude-text-watermark) (EXPLAINER, updated as facts move - Anthropic announced on 2026-08-14 that FUTURE Claude models will watermark their text, to comply with the EU AI Act - applied at each new model's launch, NOT retrofitted instantly; models launched before 2026-08-02 fall under an EU transition period and follow "over the coming months", so a Claude model in use today is probably not watermarked. Mechanism: a key plus the preceding words settle LOW-STAKES WORD CHOICES; a version of DeepMind's SynthID-Text (Nature 2024), lineage back to Scott Aaronson 2022. NOTHING is added to the text and there are NO HIDDEN CHARACTERS; no extra tokens, no price or speed change, no measured quality loss. Carries NO identifying information - cannot be traced to a person, organization or chat, and changes nothing about ownership or legal responsibility. CODE: not applied where an exact output is required; it can attach to arbitrary choices such as CODE COMMENTS but has "a negligible effect on the actual code produced", so it cannot tell you whether a pull request was agent-written. PROVES ONLY the likelihood Claude was partly involved - it cannot distinguish "Claude wrote this" from "Claude heavily edited this", does not confirm human authorship, and cannot identify a different AI. Weak on short samples. Proofreading a human draft may leave too few choices to register; TRANSLATIONS do carry a watermark. IMAGES/FILES (.png/.jpg/.svg) get no watermark at all - they get a C2PA content credential, a signed note in file METADATA that leaves the file itself unchanged (Skillselion's read, not Anthropic's claim: therefore lost whenever a pipeline strips metadata). DIFFERENT from AI detectors like Pangram, which have no key and read stylistic tells. Detection needs Anthropic's key; the detection API had NOT shipped at publication. Driven by EU AI Act Article 50(2), applicable 2026-08-02, with a transition period for models launched before that date)
- [Glossary](https://skillselion.com/glossary) (what is an MCP server / agent skill / marketplace …)
- [Learn](https://skillselion.com/learn) (curated videos & articles)
- [Weekly brief](https://skillselion.com/brief) (what's new per tool)
- [About](https://skillselion.com/about) (how ranking works, independence)
- [FAQ](https://skillselion.com/faq)

## Machine access

- [Developer resources](https://skillselion.com/developers) (catalog API, OpenAPI, MCP, CLI note)
- [Skillselion API docs](https://skillselion.com/api-docs) (version 1, `GET /api/v1/listings`, rate limits)
- [Read-only sandbox](https://skillselion.com/sandbox) (same public catalog; no writes)
- [Agent instructions / when-to-use](https://skillselion.com/.well-known/agent-instructions)
- [Product MCP Streamable HTTP](https://skillselion.com/mcp) (JSON-RPC initialize / tools/list / resources/list — browser GET stays the HTML hub)
- [Product MCP alias](https://skillselion.com/.well-known/mcp/product) (same catalog tools as [/.well-known/mcp](https://skillselion.com/.well-known/mcp))
- [OpenAI-style function tools](https://skillselion.com/functions.json)
- [Web Bot Auth JWKS (empty keys)](https://skillselion.com/.well-known/jwks.json)
- [Docs MCP](https://skillselion.com/.well-known/mcp/docs) (search_docs / get_doc / list_docs — not the catalog product MCP)
- [MCP Apps catalog view](https://skillselion.com/mcp-apps/catalog.html) (ui://skillselion/catalog.html)
- [Developer map (modular llms.txt)](https://skillselion.com/developers/llms.txt)
- [Docs map (modular llms.txt)](https://skillselion.com/docs/llms.txt)
- [AGENTS.md](https://skillselion.com/AGENTS.md) (site rules for agents; repo copy: [icidab/prism AGENTS.md](https://github.com/icidab/prism/blob/main/AGENTS.md))
- [OpenAPI spec](https://skillselion.com/openapi.json) (versioned alias [/api/v1/listings](https://skillselion.com/api/v1/listings) · [/v1/listings](https://skillselion.com/v1/listings); also [/api/openapi.yaml](https://skillselion.com/api/openapi.yaml))
- [API versioning policy](https://skillselion.com/api/v1) (Sunset / Deprecation / API-Version)
- [RFC 9727 API catalog](https://skillselion.com/.well-known/api-catalog)
- [AI Catalog (ARD)](https://skillselion.com/.well-known/ai-catalog.json)
- [Agent skills index](https://skillselion.com/.well-known/agent-skills/index.json)
- [MCP server card](https://skillselion.com/.well-known/mcp/server-card.json)
- [A2A agent card](https://skillselion.com/.well-known/agent-card.json)
- [OAuth protected resource (public / unauthenticated)](https://skillselion.com/.well-known/oauth-protected-resource)
- [OAuth authorization server (does not issue tokens)](https://skillselion.com/.well-known/oauth-authorization-server)
- [plugin.json](https://skillselion.com/plugin.json)
- [mcp.json](https://skillselion.com/mcp.json)
- [MCP discovery alias](https://skillselion.com/.well-known/mcp.json)
- [MCP server card alias](https://skillselion.com/.well-known/mcp-server.json)
- [MCP SSE remote](https://skillselion.com/.well-known/mcp/sse)
- [Web Bot Auth directory (no signing keys)](https://skillselion.com/.well-known/http-message-signatures-directory)
- [Homepage markdown](https://skillselion.com/index.md)
- [Pricing (free)](https://skillselion.com/pricing.md)
- [Authentication (no public OAuth)](https://skillselion.com/auth.md)
- [NLWeb ask](https://skillselion.com/ask) · [schemamap](https://skillselion.com/schemamap.json) · [NLWeb schemamap (well-known)](https://skillselion.com/.well-known/nlweb/schemamap.json) · [catalog DataFeed](https://skillselion.com/feeds/listings.json)
- [Full URL list](https://skillselion.com/sitemap.xml)
- [Skillselion MCP server (install it)](https://skillselion.com/skillselion-mcp) - gives a coding agent on-demand access to this catalog via search_skillselion, load_skill, top_skillselion, get_catalog_stats. Install: npx -y skillselion-mcp. Live handshake: [POST /mcp](https://skillselion.com/mcp) (aliases [/.well-known/mcp](https://skillselion.com/.well-known/mcp)).
- [Contact](https://skillselion.com/contact)
- [日本語ページの一覧 (Japanese-language pages index)](https://skillselion.com/ja/llms.txt)

## Contact

- [Contact page](https://skillselion.com/contact)
- info@skillselion.com