本文へ移動
cccskills
無料GitHub で公開

vellum-editorial

Scaffold editorial documentation sites with the Vellum design system — OKLCH palette, Bodoni Moda / Source Serif 4 / Inconsolata typography, warm or dark space theme, 35 semantic components (state cards, parity matrices, entity dossiers, stat stacks, utterance blocks, editorial notes, entity flows, fact blocks, sub-navigation, source lists, debugger panels, ring visualizations, phase timelines, glossaries, subject-intro, stat-bar, hero-image with lightbox, gallery/masonry grid, audio player with Web Audio API). Static HTML, zero build step. Use when building competitive intelligence pages, personal archives, internal documentation, roadmaps, or multi-page static sites that need a designed, non-generic aesthetic.

インストール方法を見る

含まれるファイル(14)

  • SKILL.md10.8 KB
  • assets/_audio-player.js4.5 KB
  • assets/_auth.js2.0 KB
  • assets/_components.css28.7 KB
  • assets/_shared.css26.2 KB
  • assets/template-one-pager.html10.8 KB
  • assets/template-subpage.html3.6 KB
  • assets/template.html3.9 KB
  • README.md2.2 KB
  • references/advanced-patterns.md11.1 KB
  • references/component-catalog.md29.4 KB
  • references/design-tokens.md12.3 KB
  • references/page-architecture.md14.9 KB
  • scripts/init_project.py22.2 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Build editorial documentation sites with the Vellum design system. Two theme variants (warm parchment, dark space), OKLCH color science, three-font typographic hierarchy, 35 semantic components. Static HTML with zero build step — deploys to Cloudflare Pages, GitHub Pages, or any static host.

Creative Direction

Two theme variants, each with a distinct character:

Warm (default): "Vellum manuscript" — warm cream backgrounds with a subtle SVG crosshatch texture, deep ink-brown text, copper accents. Authoritative, like a well-made journal.

Dark: "Space archive" — deep blue-black backgrounds with CSS starfield (box-shadow stars + nebula gradients), bright gold copper accents, cool blue-gray text. Atmospheric, like a command center or digital archaeology site.

Both themes share the same typography, component library, and layout system. The palette transformation is purely in :root custom properties — no structural CSS changes needed.

If the minoan-frontend-design skill is available, read it for broader aesthetic principles. Vellum encodes those principles into a concrete design system. Pages built with Vellum should score 16+/20 on /design-audit and 28+/40 on /design-critique.

Quick Start

Scaffold a new project:

python3 ~/.claude/skills/vellum-editorial/scripts/init_project.py <project-name> \
  [--style editorial|instrument] [--template landing|one-pager] \
  [--pages N] [--theme warm|dark] [--auth] [--no-auth] \
  [--password <string>] [--deploy-ready] [--no-deploy-ready] \
  [--audio] [--no-audio]

This generates a deploy-ready site with CSS, starter HTML, OG/Twitter social meta tags, a light/dark theme toggle, and optionally a client-side auth gate and ambient audio player. By default, --deploy-ready generates _headers (CF Pages security headers) and robots.txt (crawl prevention). The --theme dark flag transforms the palette to a dark space aesthetic with CSS starfield background. Open index.html in a browser — fonts load from Google Fonts, icons from Phosphor Icons CDN.

Styles control the visual aesthetic:

  • editorial (default) — Bodoni Moda / Source Serif 4 typography, SVG crosshatch texture. Scholarly manuscript character.
  • instrument — Manrope typography, grid-dot texture. Precision instrument panel character for technical product overviews.

Templates control the content scaffolding:

  • landing (default) — Simple 3-section page with overview, features, comparison.
  • one-pager — 8-section technical overview with TOC, state-card grids, configuration matrix, phase timeline, and deployment section.

Design System

Three fonts: Bodoni Moda (display/headlines, italic), Source Serif 4 (body text), Inconsolata (monospace/labels/badges). All loaded from Google Fonts.

OKLCH palette with 50+ custom properties organized as base / bg / border triples for each semantic color. The core palette — --ink, --copper, --bg, --bg-warm, --bg-card, --border, --text-body, --text-label, --text-muted, --ghost — defines the vellum character. Navigation state (--nav-active / --nav-active-bg / --nav-active-border), severity tiers (--dealbreaker, --high, --moderate), platform colors (--subq, --claude, --cursor, --codex), and status indicators (--planned, --building, --shipped, --blocked) extend it. Content organized by era or category uses additional --era-* color triples.

Full token reference: references/design-tokens.md.

Component Library

35 components across 7 groups: layout (breadcrumb, page-nav, header, section, footer), cards (state-card, hook-card, cap-panel, further-card, quote-card, req-card), data display (matrix, matrix-note, glossary, ring, debugger-panel), indicators (status-dot, status-chip, tier-badge, chip), editorial (toc, pullquote, savings-callout, supersession, advantage-row, phase-timeline), intelligence (dossier, stat-stack, utterance, editorial-note, entity-flow, fact-block, sub-navigation, source-list), and media (subject-intro, stat-bar, hero-image with lightbox, gallery/masonry-grid, audio-player). Each supports multiple color variants following the semantic palette.

Copy-paste HTML snippets for every component: references/component-catalog.md.

Page Architecture

Six page patterns:

Landing page: breadcrumb → page-nav → header (icon, tag, h1, subtitle) → main (numbered sections with ghost section-num) → footer

Deep-dive page: breadcrumb (with back link) → page-nav → header (logo, tag, h1, subtitle) → main (TOC → numbered sections) → footer

Section hub: breadcrumb → page-nav → sub-nav → header (badge) → main (key findings + further-card grid) → footer

Profile dossier: breadcrumb → page-nav → sub-nav → header → subject-intro or editorial-note → primary dossier → secondary dossiers → sources → footer

Feature comparison: breadcrumb → page-nav → sub-nav → header → TOC → comparison panels with dimension grids → matrix → sources → footer

Analysis/implications: breadcrumb → page-nav → sub-nav → header → TOC → state-card grid (gaps) → advantage-rows → recommendations → sources → footer

Every page includes: skip link (<a href="#main" class="skip-link">), _shared.css (core system), _components.css (components), Google Fonts preconnect, Phosphor Icons CDN. Auth gate (_auth.js) loads before body content when enabled. Audio player (_audio-player.js) loads at end of body when enabled.

Full page structure patterns: references/page-architecture.md.

Customization

Adapt for a different client by editing :root custom properties in _shared.css:

  1. Theme selection (scaffold time): Use --theme dark to start with a dark palette, or accept the default warm theme.
  2. Color swap (~5 min): Change --copper hue, adjust --ink/--bg warmth, update platform colors to match client brands.
  3. Semantic rename (~30 min): Rename --subq → --brand-primary, update corresponding CSS class variants.
  4. Era/category colors: Add --era-* triples (base / bg / border) for content organized by time period, category, or entity type. See design-tokens.md for the pattern and lightness guidelines.
  5. Structural additions: Add new component variants following the base / bg / border triple pattern and BEM-like naming.

The 50+ CSS custom properties are the entire theming surface. No build step, no preprocessor.

Auth Gate

Optional client-side courtesy gate using FNV-1a hashing with sessionStorage. Not real security — a visibility gate for non-public docs. Set a password at scaffold time with --password <string> — the init script computes the FNV-1a hash automatically and writes the correct constant into _auth.js. Each project gets a unique sessionStorage key (vellum_auth_{project-name}) to avoid cross-site collisions.

Diagrams (Optional)

Vellum pages can include Mermaid diagrams rendered as inline SVGs via the beautiful-mermaid skill. No JavaScript dependency — diagrams are static SVG inlined at build time.

Generate a vellum-themed SVG:

printf 'graph TD\n  A --> B --> C' | \
  node ~/.claude/skills/beautiful-mermaid/scripts/mermaid.mjs \
  -f svg -t vellum --transparent -o diagram.svg -

Inline into a diagram panel:

<figure class="diagram-panel">
  <p class="diagram-panel__label">System Architecture</p>
  <!-- Paste SVG output here (remove width/height, keep viewBox) -->
  <figcaption class="diagram-panel__caption">Request flow from API gateway to storage</figcaption>
</figure>

The vellum theme uses cream/ink/copper colors matching the design system. The --transparent flag lets the panel's --bg-card background show through. Supported diagram types: flowchart, state, sequence, class, ER. See references/component-catalog.md for component variants (--copper, --bleed).

QA Workflow

After building pages, run these passes in order:

  1. /design-audit — Technical checks (a11y, performance, responsive, theming, anti-patterns)
  2. /design-critique — UX review (Nielsen's heuristics, cognitive load, persona red flags)
  3. /design-polish — Final pass (alignment, spacing, interaction states, transitions)

Deploy to Cloudflare Pages

After building and QA:

  1. Create project (first time only):

    wrangler pages project create <project-name> --production-branch main
    
  2. Deploy:

    wrangler pages deploy <project-dir> --project-name <project-name>
    
  3. Verify headers:

    curl -sI https://<project-name>.pages.dev/ | grep -iE 'x-robots|x-frame|content-security'
    

The scaffolded _headers and robots.txt handle security and crawl prevention automatically. Use --no-deploy-ready to skip these files for local-only development.

Social Preview (OG Image)

Every scaffolded page includes OG and Twitter Card meta tags pointing to og-image.png. Generate a 1200x630 OG image with ImageMagick:

magick -size 1200x630 xc:'#111111' \
  -gravity West -font Inconsolata-Bold -pointsize 52 \
  -fill '#C4956A' -annotate +140+0 'Project Title' \
  -pointsize 22 -fill '#8A8A8A' -annotate +140+50 'Subtitle line' \
  og-image.png

Or use /badge-generator for a medallion-style icon, then composite onto the dark background. Test the preview by pasting the deployed URL into Slack or the Twitter Card validator.

Anti-Patterns

Never use border-left accent stripes on cards — use ::before top-bar (like state-card). Never use gradient text (background-clip: text). Never use system-ui or default fonts. Always include skip-link and prefers-reduced-motion. Break up sequences of more than 6 state-cards with a pullquote or section break.

References

FileWhen to consult
references/design-tokens.mdBuilding new components or customizing the palette
references/component-catalog.mdNeed HTML snippets for any component
references/page-architecture.mdSetting up page structure, nav, TOC, or glossary
references/advanced-patterns.mdGhost watermarks, asymmetric grids, fluid typography, opacity hierarchy, logo handling

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Search academic papers, build literature reviews, and synthesize research findings — combines Exa MCP (research_paper category, arxiv filtering) with arxiv-mcp-server for paper discovery, download, and deep analysis. Triggers on academic paper, literature review, research synthesis, arxiv, find papers, scholarly search.

日本語の概要は準備中です。原文の説明を表示しています。

tdimino/claude-code-minoan412026年9月28日 更新

Automates browser interactions for web testing, form filling, screenshots, and data extraction. Use when the user needs to navigate websites, interact with web pages, fill forms, take screenshots, test web applications, or extract information from web pages.

日本語の概要は準備中です。原文の説明を表示しています。

tdimino/claude-code-minoan412026年9月28日 更新

This skill should be used when creating, auditing, or maintaining AGENTS.md files and Codex CLI configuration for any project—including initializing AGENTS.md for cross-agent compatibility (Codex, Cursor, Copilot, Devin, Jules, Amp, Gemini CLI), generating config.toml or .rules files, scaffolding .agents/skills/, converting CLAUDE.md to AGENTS.md, or auditing existing agent configs for bloat and staleness. Complementary to codex-orchestrator (which executes subagents; this skill creates the config files they consume).

日本語の概要は準備中です。原文の説明を表示しています。

tdimino/claude-code-minoan412026年9月28日 更新

Academic research skill for Biblical Hebrew, Semitic linguistics, cuneiform studies, and comparative Ancient Near Eastern research. Provides Sefaria API for Hebrew Bible, CDLI/ORACC for cuneiform databases, and web discovery via Omnisearch, Exa, Firecrawl, and Obscura for finding scholarly sources across JSTOR, Perseus, Persée, Google Scholar, and academia.edu. Triggers on Hebrew quotes, cuneiform, Sefaria, ANE research, Minoan, search for scholarship, find papers, literature review, scholarly search, academic search, Genesis/Tehom, Ugaritic, Talmudic sources, extract from PDF, OCR academic.

日本語の概要は準備中です。原文の説明を表示しています。

tdimino/claude-code-minoan412026年9月28日 更新

Build comprehensive ARCHITECTURE.md files following matklad's canonical guidelines — bird's-eye views, ASCII/Mermaid diagrams, codemaps, invariants, and layer boundaries. Triggers on document the architecture, create ARCHITECTURE.md, map this codebase, architectural overview.

日本語の概要は準備中です。原文の説明を表示しています。

tdimino/claude-code-minoan412026年9月28日 更新

Generate physically-based atmospheric scattering shaders — sky domes, planetary atmospheres, LUT-optimized pipelines, depth-aware post-processing. Four modes — sky-dome, atmosphere-post, planet, lut. Triggers on atmospheric scattering, sky shader, sunset rendering, planet atmosphere, Rayleigh scattering, Mie scattering, volumetric sky, sky dome, atmosphere post-processing, aerial perspective, transmittance LUT, sky rendering, realistic sky, planetary rendering.

日本語の概要は準備中です。原文の説明を表示しています。

tdimino/claude-code-minoan412026年9月28日 更新

tdimino のスキルをすべて見る

このスキルの問題を報告する