Typed templates for ADRs, RFCs, and posts

ADR-0001 settled how posts are compiled. This RFC settles how they render.

Three kinds of writing share the collection: Architecture Decision Records, Requests for Comments, and freeform posts. ADRs and RFCs benefit from a predictable skeleton — a reader should find the Decision of any ADR in the same place every time. But forcing a blog post into "Context / Decision / Consequences" would be absurd. The system needs structure where structure helps and freedom where it doesn't.

Each post declares its type in frontmatter. The [slug] route compiles the MDX once, then dispatches to a template by type:

switch (meta.type) {
  case "adr":
    return <AdrTemplate meta={meta}>{content}</AdrTemplate>;
  case "rfc":
    return <RfcTemplate meta={meta}>{content}</RfcTemplate>;
  default:
    return <BlogTemplate meta={meta}>{content}</BlogTemplate>;
}

Each template renders a shared CaseStudyShell (back-link, title, date, tags, repo link, rail terminus) plus a type-specific header — a status badge and decider/author block for ADRs and RFCs.

Structure inside the body comes from section components passed into the MDX scope, so authors write semantic tags instead of repeating markup:

Every section renders through one LabeledSection primitive that draws the mono blueprint label, so the look stays consistent and the labels can't drift.

• One universal template. Simplest to build, but it can't show an ADR's status or an RFC's number without a pile of conditionals — the complexity just moves into the template.
• Convention-based headings. Detect ## Context and style it. Fragile: a typo in a heading silently loses the styling, with no compile-time signal.
• Per-post bespoke layouts. Maximum control, zero reuse. Rejected immediately.

The section-component approach keeps authoring declarative while giving the compiler a chance to catch a missing section.

Authors must learn the section-component vocabulary for ADRs and RFCs rather than writing pure markdown. The mitigation is that the three seed posts in this repo double as copy-paste templates.

• Should a missing required section (an ADR with no <Decision>) be a build error rather than a silent omission?
• Do RFCs need a status-history table once posts start superseding one another?