Changelog — 2026-04-25_01

Summary

This changeset adds a new Strategies section to the site where markdown-authored strategy documents render as full blog-style pages with inline citations, callout boxes, code blocks, and a sources list. It also introduces the site's first LFM (Lossless Flavored Markdown) integration for rich content rendering, and fixes a pronoun error in a team member's bio.

What stakeholders see:

• New /strategies index page listing strategy articles as preview cards

• Each article has its own page (/strategies/[slug]) with professional formatting, numbered source citations, and styled callout boxes

New content collection: Strategies

Content config

• Added strategies collection to src/content.config.ts

• Loader: glob({ pattern: '**/*.md', base: './src/content/strategies' })

• Schema fields:

  • title (string, required)

  • description (string, optional)

  • author (string, optional)

  • date (coerced date, optional) — uses z.coerce.date() so string dates in frontmatter parse correctly

  • image (string, optional) — for card preview images

  • tags (string array, optional)

Content directory

• src/content/strategies/ — already contained one file: why-markdown-needs-to-grow-up.md

• No new content files were added; the existing file is now served by the new collection and pages

New pages

src/pages/strategies/index.astro

• Route: /strategies

• Loads all entries from the strategies collection via getCollection('strategies')

• Renders each entry as a preview card in a responsive CSS grid (grid-template-columns: repeat(auto-fill, minmax(320px, 1fr)))

• Card displays: title, description, author, date, tags, optional cover image

• Uses BaseThemeLayout wrapper consistent with other site pages

• Hero section with gradient background using site CSS custom properties (--color-primary-950, --color-primary-900)

• Empty state message when no strategies exist

src/pages/strategies/[slug].astro

• Route: /strategies/:slug

• Uses getStaticPaths to generate one page per strategy entry

• Parses markdown body through parseContent() (see below), receiving an enriched MDAST tree and structured citation array

• Renders article with:

  • Back link to /strategies

  • Header: title, description, author, date, tags

  • Markdown body via <AstroMarkdown node={tree} />

  • Sources section via <Sources citations={citations} /> (only rendered when citations exist)

• Prose styling via scoped :global() selectors for headings, paragraphs, lists, links, blockquotes, code, and horizontal rules

LFM integration — Markdown rendering pipeline

Dependencies added

• @lossless-group/lfm — already present at ^0.1.3 in package.json

• mdast-util-to-string@^4.0.0 — added (used by AstroMarkdown.astro for heading ID generation)

Markdown components copied from packages/lfm-astro

Per the copy-and-adapt pattern documented in CLAUDE.md, the following components were copied from packages/lfm-astro/components/ into src/components/markdown/:

Import paths: AstroMarkdown.astro imports CodeBlock, Callout, and MarkdownImage from ./ (same directory). No path changes were needed since all five components live together in src/components/markdown/.

Key rendering behaviors

• footnoteReference nodes render as superscript [n] using node.data.citationIndex (sequential number), with an anchor link to #source-{hex}

• footnoteDefinition nodes are suppressed (&& null) — their content is rendered by the Sources component instead

• containerDirective nodes with name === "callout" route to the Callout component

• containerDirective nodes with name === "image" route to the MarkdownImage component

• Unrecognized node types fall through to a catch-all that renders their children

src/lib/parse-content.ts — Reusable markdown processing utility

This is the core addition that makes LFM rendering work across the site. It wraps parseMarkdown() from @lossless-group/lfm and polyfills two features that are missing from the published v0.1.3 but present in the unpublished v0.2.x workspace source.

Why this exists

The published @lossless-group/lfm@0.1.3 on GitHub Packages includes remark-gfm, remark-directive, and remark-callouts, but:

1. remark-citations does not exist in 0.1.3 — it was added in 0.2.x (not yet published). Without it, footnoteReference nodes retain their raw hex identifiers (a1b2c3) instead of sequential numbers (1).

2. remark-callouts has a regex bug in 0.1.3 — the CALLOUT_REGEX uses $ without the multiline flag, so when remark-parse merges the > [!type] Title line and the body text into a single text node with \n, the regex fails to match. Callouts render as plain blockquotes with raw [!tip] text.

API

import { parseContent } from '../lib/parse-content';
const { tree, citations } = await parseContent(markdownString);

Returns:

• tree — enriched MDAST root node, ready to pass to <AstroMarkdown node={tree} />

• citations — ordered array of Citation objects for <Sources citations={citations} />

Callout polyfill (fixCallouts)

• Runs after parseMarkdown() returns the tree

• Walks the tree looking for blockquote nodes whose first text child starts with [!type]

• Matches against the first line only (splits on \n before matching), avoiding the multiline regex bug

• Transforms matching blockquotes into containerDirective nodes with:

  • name: 'callout'

  • attributes: { type, title }

  • children: remainder of the paragraph text + subsequent blockquote children

  • data.hProperties: { class: 'callout callout-{type}' }

• If remarkCallouts already transformed the blockquote (future LFM version), there's nothing left to match, so the polyfill is a no-op

Citation polyfill

• Checks tree.data.citations.ordered — if populated (LFM >= 0.2), returns it directly

• Otherwise, replicates the remark-citations logic:

  1. Collect: Walks tree depth-first collecting footnoteReference identifiers in document order

  2. Index: Assigns sequential 1-based indices by order of first appearance

  3. Enrich: Sets node.data.citationIndex and node.data.citationHex on each footnoteReference node

  4. Parse definitions: Extracts text from footnoteDefinition children, parses [Title](URL) and Published: YYYY-MM-DD patterns into structured Citation objects with fields: identifier, hex, index, title, url, source (hostname), publishedDate, raw, parsed

  5. Remove definitions: Filters footnoteDefinition nodes out of the tree (they're rendered via Sources component)

  6. Return: Sorted citation array

Pattern notes for other sites

This utility is designed to be portable. To implement LFM rendering in another site:

1. Copy src/lib/parse-content.ts and src/components/markdown/*.astro from this site (or from packages/lfm-astro/)

2. Install @lossless-group/lfm and mdast-util-to-string

3. In any page: const { tree, citations } = await parseContent(entry.body)

4. Render: <AstroMarkdown node={tree} /> and <Sources citations={citations} />

5. When @lossless-group/lfm is published at 0.2.x+ with native remark-citations, the polyfills in parseContent will detect the existing output and skip themselves — no code changes needed

When to remove the polyfills

Once @lossless-group/lfm is published at a version that includes:

• remark-citations with citationIndex enrichment

• remark-callouts with a multiline-aware regex (or first-line splitting)

...the polyfill code in parseContent becomes dead weight. At that point:

• Update package.json to require the new LFM version

• Simplify parseContent to just const tree = await parseMarkdown(md); return { tree, citations: tree.data.citations.ordered ?? [] };

• Or remove parse-content.ts entirely and call parseMarkdown directly in pages

Build status

• pnpm build passes successfully

• New pages generated:

  • /strategies/index.html

  • /strategies/why-markdown-needs-to-grow-up/index.html

• Pre-existing Unsplash 403 warnings during prerender are unrelated (FactCard--w-Image component, missing/expired UNSPLASH_ACCESS_KEY)

Files changed (tracked)

• package.json — added mdast-util-to-string@^4.0.0 dependency

• src/content.config.ts — added strategies collection definition

• src/content/people/team.json — bio text correction

• src/content/people/team-extended.json — bio text correction

Files added (untracked)

• src/lib/parse-content.ts — reusable LFM parsing + citation/callout polyfill

• src/components/markdown/AstroMarkdown.astro — recursive MDAST renderer (copied from packages/lfm-astro)

• src/components/markdown/Callout.astro — callout box component (copied from packages/lfm-astro)

• src/components/markdown/CodeBlock.astro — code block component (copied from packages/lfm-astro)

• src/components/markdown/MarkdownImage.astro — image directive component (copied from packages/lfm-astro)

• src/components/markdown/Sources.astro — citation list component (copied from packages/lfm-astro)

• src/pages/strategies/index.astro — strategies listing page

• src/pages/strategies/[slug].astro — individual strategy article page

Known issues / follow-ups

• LFM version: Site is on @lossless-group/lfm@^0.1.3. The parseContent polyfill covers the gap, but publishing LFM 0.2.x would let all sites drop the polyfill.

• Callout regex bug: Exists in both the published 0.1.3 and the current workspace source (packages/lfm/src/plugins/remark-callouts.ts line 15). The CALLOUT_REGEX should either use the m (multiline) flag or split on \n before matching. This should be fixed in the LFM package itself.

• Strategy content: Only one article exists (why-markdown-needs-to-grow-up.md). The infrastructure is ready for more.

• No .npmrc in site root: The site relies on the monorepo root .npmrc for @lossless-group registry resolution. For independent deployment, a site-level .npmrc will be needed.