DEV Community

Roberto Luna
Roberto Luna

Posted on

Adding JSON Schema Validation to the Content‑Automation Pipeline

Adding JSON Schema Validation to the Content‑Automation Pipeline

TL;DR: I introduced a JSON schema validator (AJV) into the CI workflow to catch malformed content files before they reach the publishing scripts. The change stops broken Bluesky posts and keeps metadata.json in sync across all channels.


The Problem

Our content-automation repo generates Markdown and JSON payloads for Medium, Substack, and Bluesky every night. A single typo in a JSON file—like a missing comma or an invalid date string—breaks the downstream publishing step, causing the CI job to fail after the heavy build phase.

The most recent symptom was a Bluesky post with an empty date field, which the publishing script interpreted as undefined and threw:

Error: Invalid post payload – "date" is required (bluesky_en.json line 12)
Enter fullscreen mode Exit fullscreen mode

The offending file (content/2026/11/19/content-automation/bluesky_en.json) looked like this in the diff:

+[
+  {
+    "type": "progress",
+    "text": "Finally fixed the bug in content/2026/11/19/content-automation/bluesky_en.json that was breaking the auto‑generated post dates. Swapped 
Enter fullscreen mode Exit fullscreen mode

Notice the missing closing quote and the truncated line—this is exactly what slipped through our previous checks because we only linted Markdown files.

What I Tried First

My first instinct was to add a quick grep in the CI script to look for the string "date": "". I also tried running node -c on the JSON files, but node -c only checks syntax, not schema compliance (e.g., required fields, format). Both approaches missed the problem because the file was syntactically valid JSON (the truncation happened after the CI checkout) and the missing date was a logical error, not a syntax error.

The Implementation

1. Define a JSON schema

Create schema/content-schema.json to describe the shape of every platform payload:

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "Content Payload",
  "type": "object",
  "properties": {
    "type": { "type": "string", "enum": ["progress", "announcement", "article"] },
    "text": { "type": "string", "minLength": 1 },
    "date": { "type": "string", "format": "date-time" }
  },
  "required": ["type", "text", "date"],
  "additionalProperties": false
}
Enter fullscreen mode Exit fullscreen mode

The schema enforces that every JSON file must contain a date in ISO‑8601 format, which is exactly what the Bluesky publisher expects.

2. Add a validation script

Add scripts/validate-content.js that walks the content/**/**/*.json tree and validates each file against the schema using AJV:

// scripts/validate-content.js
const fs = require('fs');
const path = require('path');
const Ajv = require('ajv');
const ajv = new Ajv({ allErrors: true, strict: false });

const schema = JSON.parse(
  fs.readFileSync(path.resolve(__dirname, '../schema/content-schema.json'), 'utf-8')
);
const validate = ajv.compile(schema);

function* walk(dir) {
  for (const entry of fs.readdirSync(dir, { withFileTypes: true })) {
    const res = path.resolve(dir, entry.name);
    if (entry.isDirectory()) yield* walk(res);
    else if (entry.isFile() && entry.name.endsWith('.json')) yield res;
  }
}

let hasError = false;
for (const file of walk(path.resolve(__dirname, '../content'))) {
  const data = JSON.parse(fs.readFileSync(file, 'utf-8'));
  const ok = validate(data);
  if (!ok) {
    console.error(`\n❌ Validation failed for ${file}:`);
    console.error(validate.errors);
    hasError = true;
  } else {
    console.log(`✅ ${file} is valid`);
  }
}

process.exit(hasError ? 1 : 0);
Enter fullscreen mode Exit fullscreen mode

3. Wire the script into package.json

{
  "scripts": {
    "validate:content": "node scripts/validate-content.js",
    "ci": "npm run lint && npm run test && npm run validate:content && npm run build"
  }
}
Enter fullscreen mode Exit fullscreen mode

Now npm run ci will abort early if any JSON payload violates the schema.

4. Update GitHub Actions workflow

Edit .github/workflows/ci.yml to run the new step:

name: CI

on:
  push:
    branches: [ main ]
  pull_request:
    branches: [ main ]

jobs:
  build:
    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
      - name: Lint & Test
        run: npm run lint && npm run test
      - name: Validate content payloads
        run: npm run validate:content
      - name: Build
        run: npm run build
Enter fullscreen mode Exit fullscreen mode

5. Verify with a failing commit

I introduced a deliberately broken bluesky_es.json (missing the date field) and pushed it. The CI output now stops at the validation step:

❌ Validation failed for /content/2026/11/20/content-automation/bluesky_es.json:
[
  {
    instancePath: '',
    schemaPath: '#/required',
    keyword: 'required',
    params: { missingProperty: 'date' },
    message: "must have required property 'date'"
  }
]
Error: Process completed with exit code 1.
Enter fullscreen mode Exit fullscreen mode

No further steps (build, deploy) run, saving minutes of wasted compute and preventing a broken post from ever reaching Bluesky.

Key Takeaway

Schema‑driven validation early in the CI pipeline catches logical content errors (missing fields, wrong formats) that static linting can't detect. It turns a runtime publishing failure into a fast, deterministic CI failure, keeping the multi‑channel pipeline reliable.

What's Next

  • Front‑matter validation: Extend the same approach to the Markdown files by defining a YAML schema and using yaml-schema-validator.
  • Auto‑fix date formats: Add a small pre‑processor that normalizes any date strings to ISO‑8601 before validation, reducing manual edits.
  • Typed content loader: Generate

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

Repo: zaerohell/content-automation · 2026-11-21

#playadev #buildinpublic

Top comments (1)

Collapse
 
userxnom profile image
Abdullah Naseer •

The failing-commit test is the part that sells this for me — a validator nobody has seen fail is easy to distrust, and proving CI stops at the validation step before the build runs is exactly the evidence that matters. One question: does the schema also cover metadata.json itself, or just the per-platform payloads? You mention keeping metadata.json in sync across channels, and that file drifting out of shape feels like the next place the same bug could hide.