Website Project Setup

Initialize a cogni-website project by discovering content from existing insight-wave plugins, selecting a theme, and scaffolding the project directory.

Core Concept

A website project aggregates content from multiple insight-wave plugins into a multi-page static website:

• cogni-portfolio (required) — products, features, propositions, solutions, customer narratives
• cogni-marketing (optional) — blog posts, articles, whitepapers
• cogni-trends (optional) — trend reports with investment themes for an insights page
• cogni-research (optional) — research reports for a resources/whitepapers page

Setup discovers what content is available across these plugins, validates minimum requirements, captures company details, selects a visual theme, and creates website-project.json — the configuration that all downstream skills depend on.

If a website-project.json already exists in the working directory, redirect to the website-resume skill.

Workflow

1. Check for Existing Project

Scan the current directory and immediate children for website-project.json. If found, inform the user and invoke the website-resume skill instead.

2. Discover Content Sources

Scan the workspace for insight-wave plugin projects using recursive globs. This mirrors the discovery pattern used by cogni-marketing:marketing-setup — recursive globs ensure discovery works regardless of workspace layout.

Portfolio Discovery (required)

• Glob for **/portfolio.json (recursive — do not use narrow single-level globs)
• Read each portfolio.json to extract: company name, description, industry, language
• Check for synthesized output: output/README.md
• Check for enriched customer narratives at output/communicate/customer-narrative/:
  • Overview narrative (portfolio-overview.md)
  • Market-level narratives (market/*.md)
  • Persona-level narratives (customer/*.md)
  • These are audience-tailored prose — richer than raw entity JSON. Store their paths for downstream page generation.
• Count entities: products, features, markets, propositions, solutions, packages

Marketing Discovery (optional)

• Glob for **/marketing-project.json
• Read marketing-project.json for brand voice and content strategy
• Count content pieces by type: content/thought-leadership/*.md, content/demand-generation/*.md, content/lead-generation/*.md

Trends Discovery (optional)

• Glob for **/tips-project.json
• For each found, check if tips-trend-report.md exists alongside it (trend report completed)
• If present, also check for tips-value-model.json to count investment themes
• Trend reports contain executive-ready narrative content with inline citations — ideal for an Insights page

Research Discovery (optional)

• Glob for directories containing **/output/draft-v*.md or **/output/report.md (cogni-research output)
• Count available research reports
• Research reports serve as whitepapers/resources — high-value content for establishing authority

Present Findings

Gefundene Inhaltsquellen:

Portfolio: ../acme-cloud/
  ✓ 3 Produkte, 8 Features, 2 Märkte, 12 Propositions
  ✓ Kundendarstellungen: 1 Übersicht, 2 Markt-Narrative, 3 Persona-Narrative
  ✓ Synthese erstellt

Marketing: ../acme-marketing/
  ✓ 4 Thought-Leadership-Artikel
  ✓ 6 Demand-Generation-Beiträge
  ✓ 2 Landing Pages

Trends: ../b2b-ict-trends/
  ✓ Trend-Report mit 5 Investitionsthemen

Research: ../cloud-security-report/
  ✓ 1 Forschungsbericht

If multiple projects of the same type are found, present all and ask the user to select one. If only one exists, confirm automatically.

Validation Gates

Hard gate — if no portfolio project is found, warn the user that a portfolio is the minimum requirement. Offer to help set one up via cogni-portfolio:portfolio-setup. Do not proceed without a portfolio.

Hard gate — the discovered portfolio must have at least 1 product AND a company name/description. Without these, the website has no meaningful content to render.

Soft warnings (inform but allow proceeding):

• No propositions → product pages will lack benefit messaging
• No customer narratives → case studies page not available
• No marketing content → no blog section
• No trend report → no insights page
• No research reports → no resources page

3. Gather Company Details

Extract company information from the discovered portfolio.json:

• Company name, description, tagline
• Contact email, phone, address (ask if not in portfolio)
• Language (from portfolio.json language field)

Ask only for what's missing. Adapt all user-facing text to the portfolio language (examples below use German):

"Aus dem Portfolio übernehme ich: Acme Cloud Services — Cloud-Infrastruktur für den Mittelstand. Fehlen noch: Kontakt-E-Mail und Adresse für die Kontaktseite. Können Sie die ergänzen?"

4. Select Theme

Invoke cogni-workspace:pick-theme to let the user select a visual theme. The theme drives all colors, fonts, and styling across the website.

After theme selection, derive design variables by reading the theme.md file and generating output/design-variables.json following the convention in cogni-workspace/references/design-variables-pattern.md.

5. Configure Build Options

Ask the user about build preferences using AskUserQuestion. Only show options for content sources that were actually discovered:

• Homepage Hero: "Pencil MCP" (AI-generated hero image, ~3-5 Min.) or "CSS-only" (schneller, Farbverlauf-Hintergrund)
• Blog einbinden: Ja/Nein (only if marketing content exists)
• Fallstudien einbinden: Ja/Nein (only if customer narratives exist)
• Insights-Seite einbinden: Ja/Nein (only if trend report exists)
• Ressourcen-Seite einbinden: Ja/Nein (only if research reports exist)

6. Create Project Structure

Create the website project directory and write configuration:

mkdir -p cogni-website/{output/website/{css,pages,images},output}

Write website-project.json following the schema documented in ${CLAUDE_PLUGIN_ROOT}/libraries/EXAMPLE_WEBSITE_PLAN.md (see the project config section). Key fields:

• slug, name, language, theme_path
• company — name, tagline, description, contact details
• sources:
  • portfolio_project — path to portfolio directory (required)
  • marketing_project — path to marketing directory (null if not found)
  • trends_project — path to trends directory (null if not found)
  • research_projects — array of paths to research report directories (empty array if none)
  • enriched_portfolio_narratives — object with overview, markets (map: slug → path), personas (map: slug → path). null if no portfolio-communicate output exists.
• build_options — hero_renderer, include_blog, include_case_studies, include_insights, include_resources
• content_discovery — entity counts per source for change detection by website-resume

All boolean build options default to true when the corresponding content source exists, false when it does not. Set hero_renderer to "pencil" or "html" based on user choice.

7. Present Summary and Next Steps

Website-Projekt erstellt: {slug}

Konfiguration:
  Unternehmen: {name}
  Sprache: {language}
  Theme: {theme_name}
  Hero: {pencil|html}

Inhaltsquellen:
  Portfolio: ✓ ({N} Produkte, {M} Features, {K} Propositions)
  Marketing: {✓ N Artikel | ✗ nicht gefunden}
  Trends:    {✓ Trend-Report | ✗ nicht gefunden}
  Research:  {✓ N Berichte | ✗ nicht gefunden}

Nächster Schritt: /website-plan — Seitenstruktur planen und Inhalte zuordnen

Output Language

Read the language field from portfolio.json. Generate all user-facing text in that language. JSON field names and slugs remain in English. Default to German (de) if not specified.

Slug Convention

Derive the website slug from the company name:

• Lowercase, hyphens instead of spaces
• Replace umlauts: ü→ue, ö→oe, ä→ae, ß→ss
• Remove special characters
• Append -website suffix
• Example: "Acme Cloud Services" → acme-cloud-website