Install the Pagewell skill

Before you start

Use this guide when you want a coding agent to generate and maintain SEO, GEO, docs, free tool, landing, comparison, or private ABM pages as normal repo files.

You need:

• a coding-agent harness that supports skills
• a terminal that can run bunx
• an existing website repo where the agent can read and edit files
• time to review the generated context before creating pages

Pagewell works best after the target repo has two durable context files: PAGEWELL.md for product facts, claim rules, route taxonomy, site conventions, page patterns, generation rules, and QA rules; and DESIGN.md for visual identity and design constraints.

Steps

1. Install Pagewell with the skills CLI:

   bunx skills add ReScienceLab/pagewell --skill pagewell

2. For Pi specifically, target the Pi agent install:

   bunx skills add ReScienceLab/pagewell --skill pagewell -a pi

3. If Pagewell is already installed globally, update it instead:

   bunx skills update pagewell -g -y

4. Restart the agent session so the skill metadata is loaded.

5. Ask the agent to discover the repo before generating pages:

   Use Pagewell to discover this repo and create PAGEWELL.md.
   Import product facts, supported and prohibited claims, route patterns, site implementation details, page/component patterns, and QA rules.
   Keep DESIGN.md separate as the visual source of truth.

6. Generate a page only after discovery is complete:

   Use Pagewell to generate /use-cases/generate-seo-pages-as-code.
   Page type: use-case.
   Audience: solo founders and developers using coding agents.
   CTA: Open Pagewell.
   Run QA and update discovery files afterward.

Verify it worked

The next agent session should list or use the Pagewell skill when the task involves SEO/GEO pages, docs, free tools, discovery files, or page QA. A healthy repo setup should include a current PAGEWELL.md, a separate DESIGN.md, and generated pages that match the site’s existing route and component conventions.

What Pagewell should change

A good Pagewell run should make repo-native changes, not produce a detached copy document. In an Astro site, that usually means content collection entries, route-aware metadata, related links, schema props, and discovery-file updates through the site’s existing utilities. In another framework, the same principle applies: use the current route, layout, component, and build conventions before adding new abstractions.

Pagewell should also preserve claim discipline. If a product fact is not in PAGEWELL.md, a source file, the page brief, or a cited source, the generated page should not state it as fact. That is especially important for pricing, customer counts, rankings, benchmarks, uptime, security, and revenue claims.

Common issues

• If the skill does not appear, restart the agent and list installed skills.
• If PAGEWELL.md is missing, run the Pagewell discover-site workflow before page generation.
• If a page draft invents product metrics, customers, prices, rankings, or integrations, reject the draft and update the prohibited claims in PAGEWELL.md.
• If a generated route does not appear in sitemap or llms.txt, run Pagewell discovery-file updates for indexable pages.

Related pages

• Pagewell product page
• Generate SEO pages as code
• Pages as code
• Generative engine optimization

Sources

• Official Pagewell repository
• GitHub Docs: About repositories
• Last checked: 2026-05-09