Astro Static SEO Site — Context Pack

A copyable context pack describing an Astro static SEO site so an AI agent can work in it correctly from the first prompt.

Paste this at the top of a task so the agent understands the project before it writes anything. It’s written to be read by humans and by your agent.

Project Background

A content-driven static site built with Astro. No database, no server runtime.
Content lives in Markdown/MDX under src/content and is rendered to static HTML
at build time. Deployed to Cloudflare Pages.

Stack

Astro 6 (static output)
MDX content collections (Content Layer API)
Tailwind CSS v4 (via @tailwindcss/vite)
Pagefind for static search
TypeScript (strict)

Directory Structure

src/content/        # MDX content, one folder per collection
src/content.config.ts
src/layouts/        # BaseLayout, ResourceLayout
src/pages/          # routes + .md/.json endpoints + llms.txt
src/components/

Coding Conventions

- Content is data: edits to copy should be MDX edits, not template changes.
- Keep components in plain Astro; add a framework island only when needed.
- Every resource page also emits a .md and .json version.
- No client-side rendering of primary content — output HTML at build time.

AI Task Boundaries

- Do not add a server adapter or database.
- Do not break the .md/.json/llms.txt endpoints.
- Validate frontmatter against the zod schema in content.config.ts.

llms.txt

This site publishes /llms.txt and a .md version of every page so agents can
read content as clean Markdown instead of scraping HTML.