# A site that aggregates others' skills into 900 pages: UI Skills climbed from 870k to 240k global rank in three months · Web

> Source: https://stringers.io/en/web/web_ui-skills.com.html (Stringers / 什么值得做, generated 2026-11-08)
> Stringers answers one question every day: what small product can one or two people build in a few months and actually make money from. Numbers come from 15 databases we collect daily, each with source and date.
> For agents: these verdicts and the underlying databases (verified revenue, App Store / Google Play charts, ad libraries, GitHub, Product Hunt…) are queryable live over MCP — tools today / hot / products / directions / direction / data. Setup: https://github.com/komako-workshop/stringers (the user logs in at stringers.io → "配置 agent" to get a personal URL). Site index: https://stringers.io/llms.txt

> In the first half of the year, the author's name in design-engineering circles and the pain point of "fixing cheap AI-written UI" brought the first users and 10k stars. From 8 to October, agents mass-produced an SEO site of 900+ pages, and a command-line tool let agents keep coming back. Rank tripled in three months.

[Web](https://stringers.io/en/web/index.md) · 11-6 written

UI Skills(ibelick/ui-skills) · ui-skills.com

# A site that aggregates others' skills into 900 pages: UI Skills climbed from 870k to 240k global rank in three months

**What it is** A curated directory of AI agent skills for design engineers, picking the ones that make Claude Code and Cursor output less "AI-looking" UI, and letting agents pull them directly via CLI and MCP.

**Why it took off** In the first half of the year it rode its author's reputation in the design engineering crowd and the pain point of "fixing cheap-looking UI from AI" to its first users and 10k stars. From 8 to October it used agents to batch out 900 pages of SEO content, and the CLI let agents keep coming back. Rankings tripled in three months.

**C** **How you could play it** Don't fight for UI as a niche. Pick a field where you have credibility, build a "hand-picked" skills directory, and add a routing skill, an npx CLI, and MCP so agents can fetch from it. Then batch-generate pages by skill / author / agent / topic and tie them to search.

Grade C: the starting point was the author's personal reputation, which followers can't copy. The SEO expansion can be copied in weeks, so clones and same-format directories already exist, and marketing, legal and other niches are already covered. Hiring and sponsorships have just opened, with no outside paying customer seen yet. Chatter in small circles

Global rank 245k · 90 days ago, top 1M+

UI Skills (ui-skills.com) is a directory of AI agent skills for UI, motion, and accessibility. The content comes from people like Anthropic, shadcn, and Emil Kowalski. The owner only curates and organizes it, and adds a CLI that lets Claude Code and Cursor pull the skills themselves. It launched in January, first broke into the global top 1M websites at the end of September, and hit 245k ranking by November 5. In November 2 the owner shared traffic for the first time: 240k pageviews and 88k unique visitors. Below: how it got going, where the money comes from, and whether this playbook can be copied.

### One line of "I'm sick of it"

In 2026 January 7, French design engineer Julien Thibeaut (@Ibelick on X, 11.6k followers) created a GitHub repo whose README was just a few lines:

> UI Skills is a small set of opinionated constraints to guide agents when building interfaces. Every rule comes from things that bored me when using agents to build UI.

Roughly: this is a small set of hard rules for agents writing interfaces, and every rule comes from "the things that annoyed me when an agent wrote UI." The contribution rules are strict too: only MUST / SHOULD / NEVER; don't write what the model already knows; "only accept things that reduce the cheap feel (slop) of AI-generated UI."

That line hit a feeling many people shared. Early in 2026, more and more people were writing front ends with Claude Code and Cursor, and the pages were recognizably AI-made: the same gradients, the same rounded cards, the same centered big headlines. The frontend-design skill Anthropic released in October 2025 also says in its description that it should "avoid the generic AI aesthetic." Fixing that look became a concrete need.

Thibeaut was already known in this circle. He had built motion-primitives (6.9k stars), a React motion component library, and prompt-kit. In January 30 he tweeted that he would build "a directory that picks the best agent UI skills," a tweet with about 28k views (the figure shown in search previews). In June 16 he released the CLI, saying "have Codex, Cursor, Claude Code or any agent use this CLI, and it will find the right design engineering skill for you," with 64k views and 739 likes.

### What it actually is

| Project | Metric | Sources |
|---|---|---|
| Global rank | 9-24 first entered the top 1M (874k) → 10-15 429k → 11-05 245k | Tranco |
| Chrome real users | Globally in the top 1M tier in August, rising to the top 500k tier in September; Taiwan, Vietnam and Thailand entered the top 100k tier | Chrome User Experience Report (CrUX), monthly |
| Self-reported by the site | 240k+ page views, 88k+ unique visitors (no period given; the sponsorship is sold by the 30-day, so this is likely the past 30 days) | /sponsor page, 11-02 |
| GitHub | 10,722 stars, recently gaining 30 per day, now 65 | ibelick/ui-skills,11-07 |
| CLI tool | npm monthly downloads May 975 → June 5,385 → October 11,042 | npm |
| Page count | Sitemap lists 933 URLs, of which 791 are individual skill detail pages | sitemap |

The homepage opens with two cards at the top: "CLI: run npx ui-skills in your terminal" and "MCP: connect your agent to the UI Skills directory," with the skill list below. The first skill listed is his own routing skill "ui-skills-root," which picks the fewest UI skills an agent needs based on topic, tech stack and intent. Then come improve-ui (his own), frontend-design (Anthropic), improve-animations (Emil Kowalski) and improve (shadcn). Each skill gets its own page with a one-line install command and a link back to the original author's repo.

In other words, **most of the content on this site was not written by it** . What it sells is selection: out of tens of thousands of skills, telling design engineers which few are worth installing.

### Why it only took off after September

Chrome's real-user data shows that by 5, June it already had substantial traffic in Taiwan and Vietnam, and from June on 20 countries entered the top 500k tier. So in the first half of the year, on the author's name and those few tweets, it had its first group of users.

What pushed rankings from 870k to 240k was the expansion from 8 through October. Over those three months the repo saw about 300 commits (August 102, September 109, October 88), nearly all adding pages:

- 8-17: Playbook (48 pages) and CLI pages, with the commit message literally saying "update SEO";

- 9-14: Collections (19 pages);

- 9-17: agents (29 pages in one day, one each for Claude Code, Cursor, Codex and OpenCode), MCP and Design.md;

- 9-18: a hiring section, together with "optimize jobs page SEO";

- 10-09: "add guides, agent setup pages, and unique /skills metadata", giving each skill page its own title and description;

- 10-30: a design system directory, importing a public design system list in full.

The commit history shows branches opened by a Cursor agent, so this round of expansion was most likely also batch-generated with agents (inference).

The results are visible: searching "UI skills for AI agents" on Brave puts its /agents page at 1, and searching "frontend design skill cursor" puts /agents/cursor at 6. Broader terms, such as "best design skills for claude code," have not reached the top 10. Rankings started climbing in late September, matching the timing of the expansion.

Some traffic also comes from agents themselves. Every CLI run requests ui-skills.com, and MCP, llms.txt and registry.txt are pulled directly by agents. People who have installed it don't need to come back and browse the site; their agents come back over and over. How much these requests lift rankings can't be told from the outside.

Another comparison makes the point: its repo star count is mid-to-low among its peers (mattpocock/skills 316k stars, pbakaus/impeccable 105k stars), yet its site rankings have caught up with impeccable's official site (254k). **Its traffic was earned by the site itself, not carried over by the GitHub repo.**

### Where the money comes from

At first, no public revenue.

- **Hiring** : launched 9-18, a posting costs $199 / month. In early November the listings had only two entries, a product designer at Deepgram and a design engineer at Cohere. The links carry utm parameters, so it's unclear whether these are paid postings or scraped from elsewhere.

- **Sponsorship** : launched 11-02, one card across "site + GitHub README" at $3,000 per 30 days, up to 3 sponsors at a time, with the card fixed in the corner of every page. The page is candid: "no click guarantee." The booking email goes to the author's own studio, Interface Office; that spot now shows his other product, Mesurer.

- There is also a newsletter that collects emails; logged-in users can bookmark skills.

This is a very old niche-community playbook: the narrower the audience, the more hiring and sponsorships are worth. Around 88k design-engineer-level visitors a month, a $3,000 sponsorship price is not unreasonable, but no first external paying customer has been seen yet.

### Who it's competing with

| Competitors | What it is | Scale | How it differs from UI Skills |
|---|---|---|---|
| skills.sh(Vercel) | Cross-category skill rankings + installer | Globally ranked 64k; repo 38k stars | Collects everything, no curation. UI Skills' install command uses its `npx skills add`, so the two are upstream and downstream |
| SkillsMP, agentskills.io, LobeHub, tessl, etc. | General skill marketplaces | Ranked anywhere from 46k to 277k; SkillsMP has been declining since October | All categories, no editorial selection; all of them list UI Skills |
| impeccable.style, ui-ux-pro-max, taste-skill | Single design skill + official site | 105k / 146k / 111k stars | Promote only their own set; UI Skills is a curated multi-source collection |
| getdesign.md | Only collects DESIGN.md files | Globally ranked 447k | UI Skills added a Design.md section in September too, a direct overlap |
| aiuxplayground.com, lazyskills.sh | Directories that likewise only cover UX / front-end design skills | Not yet in the global top 1M | Same format, but no author reputation and no CLI / MCP |

A pixel-perfect clone of the whole site already sits on GitHub (niklasp/ui-skills, zxzvsdcj/ui-skills). Broad directories are crowded. A directory that is design-engineering only and curated by a respected name is the earliest and most complete one.

### If it were you

**Verdict C: worth studying, but copying it isn't worth it.** The first half-year ran on Thibeaut's own reputation in the design engineering scene, which followers can't get. The second half's expansion and SEO is something anyone can copy, and it takes a few weeks. Clones and copycats already exist in this same niche, and others are doing it too (marketing-skills.com for marketing, agentskills.legal for legal). Monetization has only just started and hasn't proven it brings in money yet.

**Who it suits:** people who already have some standing in a professional circle, with thousands of peers following them and people trusting what they recommend. For example, indie devs working in data analysis, growth, legal tech or game art who already have a set of skills they use and like.

**Who it doesn't suit:** anyone without industry credibility who just wants an "XX skills directory" to live off SEO traffic; anyone trying to take on UI / design head-on as a niche; anyone hoping to earn money from it in the short term.

**How to do it:** don't build a directory of everything, build "the ones I picked." Pick a niche you know well and first write one or two of your own skills that explain "what mistakes AI keeps making in this field" (like his line about "the parts that annoyed me"). Then bring in skills from well-known people in the field and link back to the original authors. Most important: build a routing skill plus a CLI (`npx <name> start`) and MCP so users' agents can fetch from it directly, instead of waiting for people to come browse. Once you have a few hundred skills, make a page for each skill, each author, each agent and each topic, each with its own title and description.

**How to get users:** every time you add a feature, post a screen-recorded video on X once, so the authors you've listed have a reason to repost it since you featured their skill. Show up in daily.dev and in LinkedIn posts from peers. Chinese, Japanese and Traditional Chinese tech blogs will write about it on their own. UI Skills' traffic tier in Taiwan, Vietnam and Thailand is a step higher than in Europe and the US, which lines up in timing with those write-ups.

**When to do it:** before the first "curated by a respected name" directory appears in your niche. The skill format took off at the end of 2025 and it's been less than a year: broad directories are full, and niche directories are popping up one by one. If you start six months from now, you'll most likely be building a clone for someone else.

**What would show you got it wrong:** (1) UI Skills' sponsor slots get sold in early 2027 and explicit paid job listings appear in its job board, which shows narrow directories can really bring in money, so the rating should go up. (2) Its ranking keeps climbing into the top 100k while GitHub stars don't follow, which shows SEO plus agent return visits can build traffic on their own more than expected, so people without a big name also have a chance. (3) On the flip side, if the skill marketplaces built into Claude Code and Cursor build curation into the product and third-party directory traffic drops along with it, then this whole category of site isn't worth following.

[Website ↗](https://ui-skills.com)

**Research notes: competitor table, acquisition evidence, full research report**

Competitors

| Name | Where | Launched | Scale | How it differs from this one |
|---|---|---|---|---|
| skills.sh(Vercel) | Web | 2026-01 | Tranco 63,822; vercel-labs/skills 38k stars | Overall ranking and installer, no curation; ui-skills' install command borrows its npx skills add |
| SkillsMP | Web | End of 2025 (unverified) | Tranco 108,033 and declining | Full aggregation of GitHub skills, no editorial curation |
| agentskills.io | Web | 2025-12 (unverified) | Tranco 60,220 | Site promoting the open standard, not niche-specific |
| LobeHub Skills / explainx / tessl / agenticskills | Web | 2026 | lobehub Tranco 46k, explainx 161k, tessl 277k | General marketplace that also lists ui-skills; no design-scene backing, no niche curation |
| Anthropic frontend-design(anthropics/skills) | Big-company features | 2025-10 | anthropics/skills, 188k stars | Single skill; ui-skills treats it as a listed entry |
| impeccable(pbakaus) | Open source | 2025-11 | 105k stars; impeccable.style Tranco 253,873 | Only promotes its own design skill set; ui-skills is a multi-author curated aggregation |
| emilkowalski/skills(emilkowal.ski/skill) | Open source | 2026-03 | 54k stars; Tranco 504,463 | Personal skill collection; listed by ui-skills |
| ui-ux-pro-max / taste-skill | Open source | 2025-11 / 2026-02 | 146k / 111k stars | A single design skill, not a directory |
| getdesign.md(VoltAgent awesome-design-md) | Web | 2026 | Tranco 447,251 | Only collects DESIGN.md; ui-skills also added a Design.md section in September, a direct overlap |
| OpenDesign(open-design.ai) | Web | 2026 | Tranco 68,059; 106k stars | Design agent workspace product; skill pages are a side feature |
| aiuxplayground.com / lazyskills.sh | Web | 2026 | Not visible in Tranco's top 1M (unverified) | Same format niche directory, but no author influence and no CLI / MCP |

Where users come from

- **Verified** **Author's X account + reposts in the design engineering scene** : @Ibelick 11.6k followers; 2026-01-30 registry tweet 27.9K views, 06-16 CLI tweet 64.4K views / 739 likes; an audience already built by motion-primitives (6.9k stars) and prompt-kit

- **Verified** **SEO / programmatic pages** : 300 expansion commits from 8 to October: 08-17 Playbook, 09-17 agents / CLI / MCP / Design.md pages, 09-18 Jobs (with SEO), 10-09 "feat(seo): guides, agent setup pages, unique /skills metadata", 10-30 design system directory; sitemap has 933 URLs (791 skill pages); Tranco crawling began at the end of September; Brave search for "UI skills for AI agents" ranks #1

- **Verified** **CLI / MCP / llms.txt (agent return visits)** : npm ui-skills monthly downloads went from 975 in May to 11,042 in October; the CLI at runtime requests www.ui-skills.com

- **Verified** **Third-party write-ups and listings** : daily.dev 06-16, LinkedIn 06-11 / 09-25, Taiwanese Tony.Wu blog 07-29, txtmix 07-16, Japanese Akishimodo 07-26, dev.to 01-27; listed on lobehub / tessl / context7; HN 3 points, 5 points

- Inferred **Paid ads / Product Hunt** : neither our ad library nor our PH library shows anything

## Why ui-skills.com (UI Skills) broke out

> Data as of 2026-11-07. Sources: our traffic library (Tranco / CrUX), github library (repo_day / repo_profile), GitHub API (commit history), npm downloads API, ui-skills.com pages (read via Jina), Brave / Exa search, original X posts (read via Jina).

### One line

UI Skills is a design-engineering directory of AI agent skills run by Julien Thibeaut, a French design engineer (@Ibelick, 11.6k followers on X, author of motion-primitives and prompt-kit). It lists UI, motion and accessibility skills from people like anthropics, emilkowalski, shadcn, addyosmani and pbakaus, and gives each one an `npx skills add` install command. Around it he has built his own CLI (`npx ui-skills start`), an MCP server, and llms.txt and registry.txt files. It started as a small skill repo 2026-01 and, by 2026-08 October, the author had expanded the site into a directory plus content plus community SEO site, and web traffic stepped up with it.

### 1. The data: where the growth came from

**Site ranking** (traffic library rank_day)

- Tranco: 2026-09-24 first entered the top 1M (873,798), 10-01 616,852,10-15 429,056,11-01 255,251,11-05 245,031. The ranking kept moving up and never fell back.

- CrUX (Chrome real-user data, monthly tiers): 2026-05 already in the top 500k in tw / vn / in; from 06, 20 multiple countries in the top 500k, and Thailand in the top 100k; 09 globally went from "top 1M" to "top 500k", with tw / vn / th entering the top 100k. **In other words, real traffic was there from 5–June, not something that only showed up in September. Tranco's climb came in 8–October, after the site expansion, when real traffic stepped up again, plus Tranco's 1–2 week lag.**

- Self-reported (sponsor page, live from 2026-11-02): the site claims "240k+ page views, 88k+ unique visitors", and the GitHub repo "14 days, 60k+ clones". The page doesn't state the period. Going by the sponsorship price of "$3k / 30 days", it's probably the last 30 days, unverified.

**GitHub** (github library + GitHub API)

- ibelick/ui-skills: repo created in 2026-01-07, 2026-11-07 10,722 stars, 513 forks; recently gaining +30–65 stars a day (between 10-28 and 11-07 it stayed in this range every day), 10-10 9,564 → 11-07 10,722.

- Stars are only mid-to-low tier among skill repos: the big ones in the same category are mattpocock/skills 316k, anthropics/skills 188k, ui-ux-pro-max-skill 146k, pbakaus/impeccable 105k, emilkowalski/skills 54k. **But ui-skills.com's site ranking (245k) has already caught up with impeccable.style (254k, matching a 100k-star repo), and is well ahead of emilkowal.ski (504k). Its traffic comes mainly from the website itself, not from the repo.**

**CLI** (npm downloads API, package `ui-skills`)

- 2026-01-15 release; monthly downloads went from 975 in May to 5,385 in June (6-16 CLI tweet), then 07–September about 9.3–9.7k a month, and 11,042 in October. At runtime the CLI hits www.ui-skills.com (the address is hard-coded in the package, along with the /api/github-stars endpoint). So part of the "traffic" is agents / CLIs calling the site. Tranco has DNS-based data sources, so these requests may also contribute to the ranking (inferred, unverified).

**Where the traffic lands** : web pages plus CLI / MCP only, no app.

### 2. Competitors: who's doing the same thing

| Product | Form | Launched | Scale | How it differs from ui-skills |
|---|---|---|---|---|
| skills.sh(Vercel,vercel-labs/skills) | Web + `npx skills` CLI | 2026-01 | Tranco 63,822; repo 38k stars | Overall skill ranking and installer, no curation, no niche split. ui-skills' install command itself borrows `npx skills add`, so the two are upstream and downstream of each other. |
| SkillsMP(skillsmp.com) | Full-category skill aggregator | End of 2025 to early 2026 (unverified) | Tranco 108,033, declining since October | Crawls GitHub at scale, no editorial curation |
| agentskills.io | Agent Skills open standard / directory | 2025-12 (unverified) | Tranco 60,220 | Promotes the standard, not niche-specific |
| LobeHub Skills, explainx.ai, tessl.io, agenticskills.io, llmbase.ai | General skill marketplace / registry | 2026 | lobehub 46k, explainx 161k, tessl 277k | All list ibelick/ui-skills, but all are full-category |
| Anthropic official frontend-design (anthropics/skills, claude.com/marketplace) | Big-company features | 2025-10 | anthropics/skills, 188k stars | It's just one skill, not a directory; ui-skills lists it as an entry |
| impeccable.style(pbakaus/impeccable) | Single flagship design skill + site | 2025-11 | Tranco 253,873; 105k stars | Only pushes its own skill set; ui-skills is a curated aggregation of many |
| emilkowal.ski/skill(emilkowalski/skills) | Personal motion / design skill collection | 2026-03 | Tranco 504,463; 54k stars | Personal portfolio style; ui-skills lists it |
| ui-ux-pro-max (uupm.cc), taste-skill (tasteskill.dev) | Single design skill | 2025-11 / 2026-02 | 146k / 111k stars | Same: a single product, not a directory |
| getdesign.md(VoltAgent awesome-design-md) | DESIGN.md collection | 2026 | Tranco 447,251 | Only collects DESIGN.md; ui-skills 9-17 also added a Design.md section, direct overlap |
| open-design.ai(nexu-io/open-design) | Design agent workbench + skill pages | 2026 | Tranco 68,059; 106k stars | It's a product (open-source alternative to Claude Design); the skill pages are a side feature |
| aiuxplayground.com, lazyskills.sh | UX / frontend design skill directory | 2026-04 / 2026-09 before-and-after (by search result dates) | Not in our Tranco view | Same format vertical directory, but no author influence and no CLI / MCP |

Bottom line: general-purpose skill directories are already crowded, and skills.sh, backed by Vercel, is the biggest. **A directory limited to UI / design engineering, curated by well-known designers in the scene: ui-skills is one of the earliest and most complete.** Small directories of the same kind (aiuxplayground, lazyskills) have appeared but haven't reached Tranco scale.

### 3. Growth: where users come from

1. **The author posts on X + design engineering circles repost (confirmed)** : @Ibelick has 11.6k followers. Every feature update goes out as a video post: 2026-01-22 "ui-skills now includes multiple skills to polish agent-generated UI"; 01-30 "added a registry… to curate the best ones for agentic UI" (search snippet shows 27.9K views); 06-16 CLI release (64.4K views, 739 likes); 07-15 improve-ui; 10-13 component-canvas. His earlier motion-primitives (6.9k stars) and prompt-kit (3.1k stars) already built an audience in the React / design engineering community.

2. **SEO / programmatic pages (confirmed, and the main cause of the 8–October inflection)** : 505 GitHub commits, 102 in August, 109 in September, 88 in October, concentrated on expanding the site:
    - 08-17 Playbook (48 pages), CLI page, "update SEO";
    - 09-14 Collections (19 pages); 09-17 shipped agents (29 pages in one day, covering Claude Code / Cursor / Codex / OpenCode / OpenClaw), CLI, MCP, collections, Design.md pages;
    - 09-18 Jobs board launched, along with "optimize jobs page SEO"; 09-22 another "SEO" commit;
    - 10-09 "feat(seo): add guides, agent setup pages, and unique /skills metadata"; 10-15 expanded topics and collections; 10-16 login + favorites; 10-30 design systems directory (13 pages, importing a full public design systems list).
    - The sitemap now has 933 URLs, of which 791 are skill detail pages. Brave ranks it #1 (/agents) for "UI skills for AI agents" and #6 (/agents/cursor) for "frontend design skill cursor"; broader terms (like best design skills for claude code) aren't in the top 10 yet. Tranco began crawling in late September, which lines up in time with this wave of expansion.

3. **Agent-native distribution (confirmed)** : `npx ui-skills start` lets agents pick skills by task; MCP, llms.txt, and registry.txt let agents fetch directly. npm monthly downloads ~10k. Every user who installs it has their agent repeatedly visiting the site. Skill detail pages offer `npx skills add https://github.com/...`, which connects to the Vercel skills.sh ecosystem.

4. **Picked up / rewritten by others (confirmed)** : daily.dev (06-16), LinkedIn posts (06-11 "A directory of UI Skills for design engineers just dropped"; 09-25 "The best design resource I found this year isn't for designers"), Taiwanese Tony.Wu blog (07-29), Chinese txtmix (07-16), Japanese Akishimodo tech blog (07-26), Chinese aibrev SDLC Playbook (09-30), dev.to tutorial (01-27), two small HN posts (01-11 3 points, 07-15 5 points, neither took off). It is also listed on lobehub, tessl, explainx, context7, and SourceForge mirrors. **CrUX tiers for Taiwan, Vietnam, and Thailand are one level higher than in Europe and the US, which matches the rewrites on Traditional Chinese and Japanese blogs** (correlation, causation not verified).

5. **Listed subjects bring their own traffic (inferred)** : the directory includes pages for big-name authors like shadcn, emilkowalski, anthropics, and addyosmani (author pages like /skills/shadcn and /skills/emilkowalski), so when people search for their skills, the directory pages catch some of that traffic.

6. **Not found** : paid ads (not in our ad library), an official Product Hunt launch (not in the PH library; only a PH mirror page, unverified), TikTok, or Xiaohongshu.

### 4. Monetization (just starting)

- Hiring: /jobs lists a posting for $199 / month (09-18 live); 11-07 board shows Deepgram Product Designer II and Cohere Design Engineer, and it's unclear whether these are paid postings or scraped.

- Sponsorship: /sponsor (live as of 11-02), "site + README" at $3k / 30 days, up to 3 sponsors at a time, with a card fixed in each page's corner. The sponsor contact email is hello@interfaceoffice.com (the author's own studio, Interface Office). The current slot is taken by the author's own product, Mesurer.

- Newsletter email collection; logged-in users can favorite skills.

- Revenue isn't public (not found on TrustMRR or in the author's tweets).

### 5. Differentiation

1. **Vertical + curated** : rather than aggregating everything, it only takes design engineering skills, and what gets in is work from well-known authors in the field; it then organizes them into collections by theme like motion, accessibility, and systems. This answers a specific pain point: AI-generated UI has a cheap "AI slop" look that needs dedicated design skills to fix.

2. **The author is the brand** : in design engineering circles, Ibelick's recommendation carries weight (several tweets with 28k–64k views). General-purpose directories (skillsmp, lobehub) can't offer that kind of endorsement.

3. **An entry point for agents** : the CLI automatically routes to the right skill based on intent (ui-skills-root), plus MCP and llms.txt. Users don't need to come back and browse the site; the agent fetches what it needs. Most general directories only have web pages.

4. **Growing from a directory into a vertical portal** : beyond skills, it added Playbook, Design.md, component libraries, design systems, icons, and jobs. These programmatic pages capture long-tail search and give design engineers a reason to return often (job hunting, looking up design systems).

Are these differentiators why it took off? Yes, in two stages. Points 1 and 2 got it its first users and 2026k stars in the first half of 10 (CrUX showed substantial traffic by 5 June). What pushed Tranco from 870k to 245k was point 4: the expansion and SEO from 8 through October, plus the ongoing return visits driven by the CLI / MCP in point 3. The star growth on the repo is modest (30–60 per day), while site traffic has doubled, which suggests the growth didn't come from GitHub virality but from the site's own search traffic and repeat visits.

### VI. How visible it is

In niche circles it's being talked about: design engineering / frontend X circles, LinkedIn design posts, and Chinese, Japanese, and Taiwanese tech blogs have all covered it, but HN didn't take off and no major media coverage has been seen. Revenue hasn't been disclosed. 11-02 The sponsorship page self-reports traffic (240k PV / 88k UV), the first time public numbers have been given.

### VII. Copycats

- Direct clones: there are forks on GitHub that copy the whole site or README (niklasp/ui-skills and zxzvsdcj/ui-skills with word-for-word identical descriptions), and SourceForge has a mirror.

- Same-format vertical directories: aiuxplayground.com (UX skill directory), lazyskills.sh (frontend design skill guide page), getdesign.md (DESIGN.md collection). Which came first hasn't been verified one by one.

- The agenticskills, skillselion, and agnxi mentioned in the card are general-purpose directories and don't count as direct copies.

### 8. What others can learn from it

- **"Pick a vertical + collect others' skills + write your own routing skill / CLI"** : the content cost is nearly zero (MIT-licensed, linking back to the original authors), and the value lies in curation and organization. Verticals like legal, marketing, data analytics, and healthcare can all do the same, provided the builder has credibility in that community.

- **Agent-native distribution** : `npx <name> start`, MCP, llms.txt, registry.txt. This makes agents repeat visitors and may also boost ranking signals.

- **Growing from a directory into a portal** : use skills to gather people first, then add jobs ($199 / month), sponsorships ($3k / 30 days, only 3 slots open), a newsletter, and login with favorites. This is a classic monetization path for vertical communities: the narrower the audience, the more valuable jobs and sponsorships become.

- **Programmatic SEO pages** : one page for each skill / author / agent / topic / design system, each with its own metadata (the 10-09 commit batch).

### 9. What I'm still unsure about

- Which period the self-reported 240k PV / 88k UV refers to hasn't been verified.

- It's impossible to separate how much of Tranco's climb comes from people browsing the site versus CLI / MCP / agent requests (DNS signals).

- Why CrUX tiers for Taiwan, Vietnam, and Thailand are unusually high hasn't been verified (there are Traditional Chinese / Japanese blog rewrites, and perhaps Asian vibe-coding communities are also spreading it).

- Whether jobs and sponsorships have real paying customers is unknown. The current sponsor slot is the author's own Mesurer.

- The 8–October commits include Cursor agent branches (cursor/…), so the author most likely used agents to batch-expand the site. This approach is also quick for others to copy.

This article was written by AI from public data and the data we collect daily. Numbers are sourced, and inferences are labeled.