⏳ This skill is pending AI review.

Scores will appear once the review pipeline completes.

v1.0.0

personal-website-claude-skill

gpz12138@gpz12138⭐ 3 stars· last commit 7mo ago· 0 open issues

Build a monochrome, PhD-caliber personal homepage for an AI / ML researcher. Static single-page site, two-column layout with sticky sidebar, English-by-default with optional bilingual mode (English + whatever language the user is chatting in — Chinese, Japanese, Spanish, French, Korean, etc. — auto-detected, not hardcoded), light / dark theme, live Google Scholar citation widget (polite two-tier daily poll). Deployable to GitHub Pages with zero build step. The user provides a LinkedIn URL (minimum), a Google Scholar URL, a GitHub account to deploy to, and ideally a resume PDF — the skill does the rest. Invoke when the user asks for an academic researcher homepage, a personal site for job-search / grad-school / faculty / industry-lab applications, a Jekyll-less alternative to al-folio, or a refresh of their existing Scholar-facing homepage.

Choose how to use this skill

You do not need every option. Choose the path your AI client supports. The stable page stays the same; versioned files are immutable.

1. Native installer

This listing has no registered native installer command. Use the complete package or source fallback below, depending on what your client supports.

Do not guess an installer command or replace an existing version without reviewing the diff.

2. Complete package recommended

Download the ZIP when available. It includes SKILL.md plus the references, security notes and version metadata.

No complete ProSkills package is published for this listing yet.

3. Prompt-only

Copy the prompt above when the agent can read the stable page or when you want to adopt the workflow without installing a skill.

Need only the instruction file?

Download SKILL.md only if your client requires a single file. The complete ZIP is safer for a full installation because it preserves the references and release context.

No path installs or executes anything by itself. Your agent still needs access to the project files. Before updating, compare the installed version and review the diff.

—/10

// RATINGS

⭐GitHub Stars
⭐ 3 on GitHubGitHub ↗

New / niche

🟢ProSkills Score
—
📍

Not yet listed on ClawHub or SkillsMP

// README


Why I made this

For a young researcher or engineer today, a personal homepage is close to required. It's where a hiring manager, a recruiter, a PI, or a program chair decides — in about thirty seconds — whether to read your paper, open your side project, forward your CV, or book an interview.

When I was applying to schools, a senior I trusted mentioned — in passing — that I'd need a personal site before sending anything out. I tried a couple of existing site builders, and tried hand-writing HTML from scratch. None of it came together fast enough during an application cycle where every day mattered.

So I wrote this skill — the thing I wish had existed for that version of me. You give it three URLs; it drafts the site, deploys to GitHub Pages, reviews itself, and hands back the link. Free, MIT, open source.

Hoping it saves somebody else that week.

Who it's for

I wrote this thinking mostly about young professionals — PhD applicants, postdocs, and visiting students on the academic side; new grads going into SWE / ML / data / PM roles, return-offer candidates, and mid-career engineers changing companies on the industry side. The market is hard right now across both tracks. Cycles close earlier every year, referrals have replaced cold applications, and every group chat has someone exhausted, refreshing their inbox at midnight, comparing themselves to whoever's Twitter looks like they already made it. I'm one of those people too.

What actually moves the needle in a crowded market is being findable. When a recruiter, a hiring manager, a program chair, or a potential advisor searches your name, the first thing they land on frames you before any interview does. A clean, current homepage is a small-but-real signal: this person is active, thoughtful, and serious about their work.

I don't want cost or front-end taste to be the reason any of us skips that step. Squarespace is $144/year. al-folio takes a weekend you don't have. Design sense isn't distributed evenly, and neither is time. So I wrote the skill I wished existed when I started my own page, and I'm open-sourcing it for free, MIT, no strings.

If you're applying anywhere this cycle — grad school, full-time roles, internships, return offers, a visa-sponsored switch — use it. Ship something imperfect and iterate. Being seen matters more than being perfect, and both matter more than being invisible. You already did the hard part; this is the five-minute part.

⚡ Install (10 seconds)

Path 1 — Terminal CLI (recommended, zero session-cache issues):

claude plugin marketplace add GPZ12138/personal-website-claude-skill
claude plugin install personal-website-claude-skill

Run these in a regular terminal, not inside a running Claude Code session. This is the fastest and most reliable path — the plugin registers globally (user scope) and every future Claude Code session on this machine will see it.

Path 2 — Inside a Claude Code session:

/plugin marketplace add GPZ12138/personal-website-claude-skill

Then restart the Claude Code session (exit, then claude again) before running:

/plugin install personal-website-claude-skill

The restart is required because Claude Code's in-session plugin index is populated at startup — running /plugin install right after /plugin marketplace add in the same session can fail with "Plugin not found in any marketplace" even though the marketplace was just added successfully. This is a session-cache quirk, not a manifest problem.

Path 3 — install.sh (user-global skill, no plugin system):

curl -fsSL https://raw.githubusercontent.com/GPZ12138/personal-website-claude-skill/main/install.sh | bash

Symlinks the skill directly into ~/.claude/skills/. Works without the plugin / marketplace layer.

Path 4 — manual clone + symlink: see docs/INSTALL.md.

💬 Then say

Open Claude Code in an empty folder and type something like:

"Build me a personal homepage. I'm an ML engineer / researcher / new grad / PhD applicant — whatever fits. LinkedIn: <your-linkedin-url> GitHub: <your-github-username> Scholar: <your-scholar-url> (skip if you don't have one)"

The skill asks for anything missing, proposes a section-by-section plan, and — once you confirm — runs end-to-end (fetch → draft → deploy → review → redeploy) without interrupting. A typical run is 8–20 minutes depending on how much the reviewer agent flags.

If you'd rather not have Claude Code guess, attach a resume PDF too. It helps with dates, GPA, publication metadata.

📦 What it produces

  • Static single-page site — index.html, styles.css, script.js, assets/. Nothing else.
  • Two-column layout — sticky sidebar (photo, name, citation stats, contact grid) and a main column with About, News, Publications, Experience, Honors, Education, Skills, Contact.
  • Live Google Scholar widget — citation count, h-index, per-year chart, updated daily by a polite two-tier poll (a lightweight check that only triggers the full scrape when a number actually changed).
  • Bilingual toggle — English by default, plus a second language auto-picked from whatever language you're chatting with the skill in (Chinese, Japanese, Korean, Spanish, French, German, …). Not hardcoded to any one pairing. Off by default; opt in during clarification.
  • Light / dark theme — respects prefers-color-scheme, persists via localStorage.
  • Mobile-responsive — two-column collapses at <960px, nav hides on phones, safe-area insets on iOS, 40×40px minimum touch targets, prefers-reduced-motion honored.
  • A GitHub Actions workflow that runs the Scholar sync daily, at a non-round minute to avoid the hourly rate-limit peak.
  • **Deployed

// HOW IT'S BUILT

// FILES

2 files shipped with this skill. Inline text is the catalog copy. The upstream link is the source repository.

  • skills/personal-website-claude-skill/SKILL.md7.9 KBupstream ↗
    ---
    name: personal-website-claude-skill
    description: Build a monochrome, PhD-caliber personal homepage for an AI / ML researcher. Static single-page site, two-column layout with sticky sidebar, English-by-default with optional bilingual mode (English + whatever language the user is chatting in — Chinese, Japanese, Spanish, French, Korean, etc. — auto-detected, not hardcoded), light / dark theme, live Google Scholar citation widget (polite two-tier daily poll). Deployable to GitHub Pages with zero build step. The user provides a LinkedIn URL (minimum), a Google Scholar URL, a GitHub account to deploy to, and ideally a resume PDF — the skill does the rest. Invoke when the user asks for an academic researcher homepage, a personal site for job-search / grad-school / faculty / industry-lab applications, a Jekyll-less alternative to al-folio, or a refresh of their existing Scholar-facing homepage.
    ---
    
    # Personal Website for AI Researcher
    
    Build a monochrome academic homepage that looks like it belongs on a
    top AI lab's team page — the kind of site frontier-lab researchers,
    PhDs and grad students actually use. Static HTML + CSS + vanilla JS,
    GitHub Pages deploy, no framework, no build step.
    
    The skill's entry point is deliberately simple: **the user hands over
    a LinkedIn URL and a Scholar URL; ideally a resume PDF. You do the
    rest.** Every phase below is about turning that input into a shipped
    live site.
    
    ---
    
    ## When to invoke
    
    - *"Help me build a personal homepage, I'm an ML / AI researcher."*
    - *"Here's my LinkedIn — make me a Scholar-ish academic page."*
    - *"I need a PhD-tier site for job search / grad school / faculty apps."*
    - *"Refresh my existing GitHub Pages homepage."*
    
    ---
    
    ## Phase 1 · Clarification protocol (required before building)
    
    **Users rarely know what to submit first.** The skill's job is to
    proactively collect every required input through an explicit
    clarification step. Don't infer from partial information, don't
    guess defaults without confirming, and don't let the user pick-and-
    choose which required fields to skip. Walk them through the
    checklist.
    
    ### Declared fields
    
    The skill has a closed set of declared fields. Every field has a
    name, a type, whether it's required, and (for required-with-default
    fields) the default the skill will propose.
    
    | Field | Required? | Default | Notes |
    |---|---|---|---|
    | `language` | **required** | `English only` | Default is English-only. If the user converses with the skill in a non-English language (Chinese, Japanese, Korean, Spanish, French, etc.), propose `bilingual EN + <that language>` and confirm. The non-English language is auto-detected from the user's conversation, not assumed. User can always override to English-only even if they chat in another language. |
    | `linkedin_url` | **required** | — | Source of truth for roles, dates, bio. |
    | `scholar_url` | **required** | — | Skill extracts the `user=...` ID → `<scholar-id>` for the live widget. |
    | `github_username` | **required** | — | User-page URLs (`<github-username>.github.io`) require the repo name to match exactly — this is a GitHub rule. |
    | `confidentiality` | **required** | `strict — public numbers only` | Offer the default + ask for overrides. Typical at frontier AI labs. |
    | `resume_pdf` | strongly preferred | — | Not required, but LinkedIn alone misses dates, GPA, rank, publication metadata. Ask once. |
    | `headshot` | optional | Pull from Scholar profile photo; if absent, inline SVG initials avatar. | Accept a direct file upload or an image URL. |
    | `phone` | opt-in only | **hidden** | Do not put a phone on a public academic page unless the user explicitly asks. |
    | `address` | opt-in only | **hidden** | Default is no physical address. One or two cities is usually all that appears (e.g., 📍 `<city-a>` · `<city-b>`). |
    | `email` | derived | pulled from LinkedIn / CV | Only surface a second email if the user explicitly provides one. |
    
    ### Defaults policy
    
    - Fields with a **default** (`language`, `confidentiality`): the
      skill proposes the default and confirms — the user can accept
      with a single "ok" or override.
    - Fields with **no default** (`linkedin_url`, `scholar_url`,
      `github_username`): the skill **cannot proceed** until the
      user supplies them. Do not substitute guesses from a name.
    - **Opt-in fields** (`phone`, `address`): never include them
      unless the user brings them up. Do not ask for them.
    - **Deployment target** defaults to **GitHub Pages**. If the
      user wants a different host, they'll say so; don't ask.
    
    ### Existing-website input
    
    If the user provides a URL to their **existing personal website**
    (either as "refresh this" or "migrate this"), add that URL to the
    source list alongside LinkedIn / CV / Scholar, and **extract every
    factual item from it first** — bio paragraphs, role descriptions,
    news entries, selected publications, honors, links. An existing
    site is often the most up-to-date source.
    
    - Fetch the existing site's HTML (plain HTTP → Playwright
      fallback, per the fetch policy).
    - Extract: About text, News items, Publications list, Experience
      entries, Honors, Education, Contact.
    - Feed all of it into the union rule below — treat the existing
      site as "one more source, possibly richer than LinkedIn".
    - Do **not** copy the existing site's **design**. The skill's
      design is monochrome-academic by default; keep that.
    
    ### Design-style override
    
    If the user provides a **specific design reference** — an image,
    a URL of someone else's site, a Figma file, or an explicit style
    description ("I want it to look like al-folio / Karpathy's page /
    this-other-researcher's site") — follow the user's reference
    instead of the skill's default monochrome palette.
    
    The default *only* applies when the user has not expressed a
    design preference. When they have, their preference wins —
    *within reason*: the skill's architecture (static HTML, bilingual,
    Scholar widget, two-agent review) is orthogonal to palette /
    typography, so you can swap the aesthetic layer without rewriting
    the structural rules.
    
    If the user's requested style conflicts with a hard rule in this
    skill (e.g., they want italics or colored venue badges) — build
    it their way, but *flag the trade-off once* ("heads up: this is
    why those rules exist") so they can make an informed choice.
    
    ### Source-merging rule (union, not intersection)
    
    When LinkedIn, the CV, and anything else the user supplies
    disagree or differ in coverage:
    
    - **Take the union.** If an item appears in any source, include
      it. Do not drop items that are missing from one source — that
      source is probably just out of date.
    - On conflicts, prefer the more specific / more recent value
      (e.g., month-level date over year-only; explicit end-date over
      "present").
    - Include single-source items silently per the union rule;
      **batch** them into the end-of-work wrap-up message ("items
      I found only on Scholar: …; kept them — say 'drop X' if
      you'd rather not"). Per the execution-discipline rule, do
      **not** interrupt mid-work to ask about each one. Never
      silently drop.
    
    Full details in Phase 5, "Merging inputs: take the union".
    
    ### How to fetch the source material
    
    The skill will need to read the user's LinkedIn profile, Google
    Scholar profile, and possibly a hosted resume. Use this
    escalating fallback:
    
    1. **Try a plain HTTP fetch first** — via the `WebFetch` tool,
       `curl`, or any built-in web-fetch agent. This is the cheapest
       path and works for Scholar, GitHub, and most personal-site
       URLs.
    
    2. **If plain fetch fails** (authentication wall, JS-rendered
       page, HTTP 403 / 999 from LinkedIn, CAPTCHA, etc.): fall back
       to a **headless browser / Playwright-click** approach — the
       MCP Playwright tools, a `Playwright click` skill, or any
       browser-automation agent installed locally (e.g.,
       `npx playwright`). This is the insurance layer.
    
    3. **If no headless-browser option is available**: check the
       rest of the local environment and installed Claude skills
       for anything that ca
  • README.mdupstream ↗

    Contents are not inlined in the catalog. Open the upstream file.

Plain-text copy, no login: /skill-bundles/gpz12138-personal-website-claude-skill-skills-personal-website-claude-skill.md

// REPO STATS

3 stars
0 open issues
Last commit: 7mo ago