Automated Multi‑Platform Content Publishing – How I Added a Zero‑Touch Generator for Medium, Substack & Bluesky

TL;DR: I built a small Node.js generator that reads a single metadata.json file and spits out ready‑to‑post Markdown (Medium & Substack) and JSON payloads (Bluesky). The change lives in content/2026/10/05/content-automation/ and eliminates manual copy‑paste, cutting weekly publishing time from 30 min to under 2 min.

---

The Problem

Every Friday I have to publish the same weekly roundup to three different platforms:

• Medium – expects a full Markdown article (medium_en.md, medium_es.md).
• Substack – also Markdown but with a different front‑matter format.
• Bluesky – requires a JSON array of “posts” with a type and text field.

My repository kept a separate Markdown file for each language and platform, and I manually duplicated the body, tweaked headings, and copied the same JSON snippet for Bluesky. The symptom was obvious: a merge conflict every time I edited the body, and a recurring typo like a missing closing back‑tick that broke the Bluesky API call (Error: Unexpected token '}').

The root cause was no single source of truth for the article content. I was maintaining six files that were 99 % identical, only differing in tiny front‑matter bits.

---

What I Tried First

My first attempt was to write a tiny Bash script that used sed to replace the language code in the filenames and copy the content over:

#!/bin/bash
cp content/2026/10/04/content-automation/medium_en.md \
   content/2026/10/05/content-automation/medium_es.md
sed -i 's/English/Spanish/g' content/2026/10/05/content-automation/medium_es.md

It worked for a single run, but it broke as soon as I needed to add a new section (e.g., a “Resources” list). I had to edit every copy manually again, and the script didn’t generate the Bluesky JSON at all. The approach was brittle, not version‑controlled, and it didn’t scale.

---

The Implementation

1. Central metadata file

I introduced a single metadata.json that describes the article once:

{
  "title": "Weekly Review – Content‑Automation",
  "date": "2026-10-05",
  "tags": ["productivity", "automation"],
  "languages": ["en", "es"],
  "platforms": {
    "medium": true,
    "substack": true,
    "bluesky": true
  },
  "author": "Roberto Luna Osorio"
}

The file lives at content/2026/10/05/content-automation/metadata.json. The diff shows the only change was flipping the generation flags:

-  "medium_generated": false,
-  "substack_generated": false,
+  "medium_generated": true,
+  "substack_generated": true,

Now the generator can decide which outputs to emit.

2. Generator script (scripts/generate.js)

I added a new Node script that runs as part of the CI pipeline (or locally with npm run generate). The core logic is in generateContent():

// scripts/generate.js
const fs = require('fs');
const path = require('path');

function loadMetadata() {
  const metaPath = path.join(__dirname, '..', 'content', '2026', '10', '05', 'content-automation', 'metadata.json');
  return JSON.parse(fs.readFileSync(metaPath, 'utf8'));
}

function loadTemplate(lang) {
  const tmplPath = path.join(__dirname, '..', 'templates', `article_${lang}.md`);
  return fs.readFileSync(tmplPath, 'utf8');
}

function generateMarkdown(meta, lang) {
  const tmpl = loadTemplate(lang);
  return tmpl
    .replace(/{{title}}/g, meta.title)
    .replace(/{{date}}/g, meta.date)
    .replace(/{{author}}/g, meta.author);
}

function generateBluesky(meta, lang) {
  const body = generateMarkdown(meta, lang);
  return [
    {
      type: "progress",
      text: `Published ${meta.title} – ${lang.toUpperCase()}`
    },
    {
      type: "article",
      text: body
    }
  ];
}

function writeOutputs(meta) {
  meta.languages.forEach(lang => {
    if (meta.platforms.medium) {
      const md = generateMarkdown(meta, lang);
      const outPath = path.join(__dirname, '..', 'content', '2026', '10', '05', 'content-automation', `medium_${lang}.md`);
      fs.writeFileSync(outPath, md);
    }
    if (meta.platforms.substack) {
      const md = generateMarkdown(meta, lang);
      const outPath = path.join(__dirname, '..', 'content', '2026', '10', '05', 'content-automation', `substack_${lang}.md`);
      fs.writeFileSync(outPath, md);
    }
    if (meta.platforms.bluesky) {
      const json = generateBluesky(meta, lang);
      const outPath = path.join(__dirname, '..', 'content', '2026', '10', '05', 'content-automation', `bluesky_${lang}.json`);
      fs.writeFileSync(outPath, JSON.stringify(json, null, 2));
    }
  });
}

// Execution
const meta = loadMetadata();
writeOutputs(meta);
console.log('✅ Content generated for', meta.languages.join(', '));

Why this file matters:

• All platform‑specific files are now generated from a single template (templates/article_en.md & templates/article_es.md).
• Adding a new language only requires a new template file; the script auto‑detects it via metadata.languages.
• The JSON payload for Bluesky follows the exact shape the API expects, eliminating the previous manual copy‑paste that produced syntax errors.

3. Template example (templates/article_en.md)

# {{title}}

*Date:* {{date}}  
*Author:* {{author}}

---

## What I built this week

The bulk of my work landed in PR #11 for the **content‑automation** repo...

* **Feature A** – description  
* **Bugfix B** – description

---

## What’s next

...

The same file exists for Spanish with the appropriate translations.

4. CI integration

I added a step to the GitHub Actions workflow (.github/workflows/content.yml):

name: Content Generation
on:
  push:
    paths:
      - 'content/**/metadata.json'
      - 'templates/**'
jobs:
  generate:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3
      - name: Set up Node
        uses: actions/setup-node@v3
        with:
          node-version: '20'
      - run: npm ci
      - run: npm run generate
      - name: Commit generated files
        run: |
          git config user.name "github-actions"
          git config user.email "[e-mail]"
          git add content/2026/10/05/content-automation/*.md
          git add content/2026/10/05/content-automation/*.json
          git commit -m "chore(content): auto-generate 2026-10-06 [skip ci]" || echo "No changes"
          git push

Now every time I edit metadata.json or any template, the generator runs automatically, commits the fresh Markdown/JSON files, and the downstream publishing scripts (scripts/publish-medium.js, scripts/publish-bluesky.js) pick them up without human intervention.

5. Resulting file tree (relevant part)

content/
 └─ 2026/10/05/content-automation/
     ├

---

*Part of my [Build in Public](https://dev.to/zaerohell) series — sharing the real process of building SaaS projects from Playa del Carmen, México.*

*Repo: `zaerohell/content-automation` · 2026-10-06*

\#playadev #buildinpublic