Automating Multi‑Platform Content Publishing with a Date‑Driven Markdown Generator

TL;DR: I built a Node.js generator that reads a single source of truth (JSON + Markdown templates) and emits platform‑specific Markdown files with correct headers, front‑matter, and token‑aware changelogs. The change eliminates manual copy‑paste errors and keeps every channel (Dev.to, Medium, Substack, Bluesky) in sync.

---

The Problem

Our weekly newsletter pipeline lives in the content-automation repo. Every Monday we need to push the same article to four platforms, each with its own markdown quirks:

• Dev.to expects a front‑matter block (title, tags, published_at).
• Medium uses a simple markdown body but no front‑matter.
• Substack needs a “Subject:” line at the top.
• Bluesky consumes a JSON payload (title, content, tags).

Previously we kept a copy of the article in each folder:

content/2026/09/06/VS/medium_en.md
content/2026/09/06/VS/substack_en.md
content/2026/09/06/VS/bluesky_en.json
…

When a change was required (e.g., fixing a typo or updating the changelog header), we edited every file manually. This caused two concrete bugs:

1. Header drift – the changelog header in VS/changelog.md was ## 2026‑09‑06 VS in some files and ## [2026‑09‑06] VS in others, breaking our automated indexer.
2. Missing metadata – the metadata.json file was updated, but the generated devto markdown still pointed to the old URL, leading to broken links on the platform.

The symptom showed up in CI as a diff mismatch:

ERROR: content/2026/09/06/VS/changelog.md is out of sync with source template.

What I Tried First

My first attempt was a quick shell script that used sed to replace placeholders:

#!/usr/bin/env bash
DATE=$(date +%Y-%m-%d)
sed "s/{{DATE}}/$DATE/g" templates/medium.md > content/$DATE/VS/medium_en.md

It worked for simple token substitution, but quickly fell apart:

• Line‑ending issues – sed on macOS vs Linux produced different newline characters, breaking the CI diff.
• Complex structures – The Bluesky JSON payload required nested quoting, which sed cannot handle reliably.
• Scalability – Adding a new platform meant writing another sed line, increasing maintenance cost.

After a weekend of debugging, the script produced malformed JSON for Bluesky:

SyntaxError: Unexpected token '}' in JSON at position 102

The Implementation

I rewrote the generator in Node.js (v20) using Handlebars for templating and a tiny orchestration layer. The architecture looks like this:

content-automation/
├─ src/
│  ├─ generate.js          # entry point
│  ├─ templates/
│  │  ├─ devto.hbs
│  │  ├─ medium.hbs
│  │  ├─ substack.hbs
│  │  └─ bluesky.hbs
│  └─ utils/
│     └─ date.js
├─ data/
│  └─ 2026-09-06.json      # source of truth for the article
└─ content/
   └─ 2026/09/06/VS/...

1. Source‑of‑Truth JSON

data/2026-09-06.json holds everything needed for the week:

{
  "title": "Migrating 30 Pages from Inline Styles to Design Tokens",
  "slug": "inline-styles-to-design-tokens",
  "date": "2026-09-06",
  "tags": ["nextjs", "design-tokens", "refactor"],
  "author": "Roberto Luna Osorio",
  "summary": "How we replaced inline CSS with centralized tokens in a monorepo.",
  "content": "## Introduction\n\n... (markdown body) ..."
}

2. Handlebars Templates

devto.hbs

---
title: "{{title}}"
published_at: "{{date}}T09:00:00Z"
tags: {{json tags}}
canonical_url: "https://dev.to/zaerohell/{{slug}}"
---

{{content}}

bluesky.hbs

{
  "title": "{{title}}",
  "content": "{{{markdownEscape content}}}",
  "tags": {{json tags}},
  "date": "{{date}}"
}

Notice the custom helpers json (pretty‑prints arrays) and markdownEscape (escapes newlines for JSON strings).

3. Generator Logic (src/generate.js)

import fs from 'fs';
import path from 'path';
import Handlebars from 'handlebars';
import { format } from 'date-fns';
import { markdownEscape } from './utils/markdown.js';

// Register helpers
Handlebars.registerHelper('json', (context) => JSON.stringify(context));
Handlebars.registerHelper('markdownEscape', markdownEscape);

const DATA_DIR = path.resolve('data');
const TEMPLATE_DIR = path.resolve('src/templates');
const OUT_ROOT = path.resolve('content');

async function loadJSON(date) {
  const file = path.join(DATA_DIR, `${date}.json`);
  return JSON.parse(await fs.promises.readFile(file, 'utf8'));
}

async function render(templateName, data) {
  const tplPath = path.join(TEMPLATE_DIR, `${templateName}.hbs`);
  const source = await fs.promises.readFile(tplPath, 'utf8');
  const template = Handlebars.compile(source);
  return template(data);
}

async function writeFile(outPath, content) {
  await fs.promises.mkdir(path.dirname(outPath), { recursive: true });
  await fs.promises.writeFile(outPath, content, 'utf8');
}

async function main() {
  const date = process.argv[2] ?? format(new Date(), 'yyyy-MM-dd');
  const data = await loadJSON(date);
  const platforms = ['devto', 'medium', 'substack', 'bluesky'];

  for (const platform of platforms) {
    const rendered = await render(platform, data);
    const ext = platform === 'bluesky' ? 'json' : 'md';
    const outPath = path.join(
      OUT_ROOT,
      date.replace(/-/g, '/'),
      'VS',
      `${platform}.${ext}`
    );
    await writeFile(outPath, rendered);
    console.log(`✅ ${platform} generated → ${outPath}`);
  }
}

main().catch((err) => {
  console.error('❌ Generation failed:', err);
  process.exit(1);
});

4. Updating the Changelog Header

The original problem with the header drift was solved by moving the header generation into the same pipeline. src/templates/changelog.hbs:

## [{{date}}] VS

### Added
{{#each added}}
- **{{this}}**
{{/each}}

### Fixed
{{#each fixed}}
- {{this}}
{{/each}}

The data/2026-09-06.json now contains an

---

Part of my Build in Public series — sharing the real process of building SaaS projects from Playa del Carmen, México.

Repo: zaerohell/content-automation · 2026-09-07

#playadev #buildinpublic