publish-article

name	publish-article
description	Safe single-article deployment to production. File-by-file only (never wildcards/recursion). Pre-deployment checklist, exact file targeting, post-deployment verification, rollback procedure.

Publish Article Skill

Overview

Publish Article handles the safe, controlled deployment of a single approved article to production. This is the moment of truth: the article leaves the newsroom and goes live to readers.

Core principle: File-by-file deployment only. Never recursive, never wildcard operations. Never deploy anything except the exact article being published.

This skill ensures:

• No accidental overwrites of other articles
• No deployment of incomplete/unapproved content
• Atomic operations (article either fully deployed or fully rolled back)
• Verification before/after deployment
• Rollback capability if something goes wrong post-publish

---

Supported Deployment Platforms

Platform	Protocol	Configuration	Example
Google Cloud Storage (GCS)	gsutil, API	Bucket name, folder structure	gs://news-bucket/articles/2025-04-01-story.html
Amazon S3	aws-cli, API	Bucket name, access key	s3://news-bucket/articles/2025-04-01-story.html
Netlify	git push, API, CLI	Deploy key, site ID	Connected to Git repo; deploy on push
Vercel	git push, API, CLI	Deploy token, project	Connected to Git repo; auto-deploy
WordPress	REST API, WP-CLI	Admin URL, auth token	POST to /wp-json/wp/v2/posts
Static Site (SFTP)	SFTP, SCP	Server host, credentials, path	sftp://[e-mail]/public_html/articles/
GitHub Pages	git push	Repo name, branch	Commit to gh-pages branch; auto-publish

---

Pre-Deployment Checklist

Before the article is deployed, every item must be approved and completed:

Editorial Gate

• Article approved by final-editor-review (editorial review completed; legal defensibility confirmed)
• Copy review passed (readability, accessibility, SEO metadata verified)
• Fact-check sign-off obtained (all claims verified; sources documented)
• No outstanding comments from editor (all feedback addressed or explicitly rejected)

Technical Gate

• All images present and optimized (<200KB each, 1200×628px minimum for social)
• All links tested and working (no 404s, no typos in URLs)
• Video/audio files uploaded (not embedded from temporary URLs; URLs point to permanent archive)
• Metadata complete:
  • Meta title (50–60 chars)
  • Meta description (155–160 chars)
  • Canonical URL (if reposting)
  • Open Graph tags (og:image, og:title, og:description, og:url)
  • Structured data (NewsArticle JSON-LD)

Compliance Gate

• Attribution complete (byline, publication date, updated date if relevant)
• Corrections/errata process documented (link to corrections page, if applicable)
• Paywall/access level set (free, subscriber-only, metered)
• Comments policy applied (allow/disallow; moderation queue yes/no)
• External links checked (no dead links to external sources)

Backup & Rollback Gate

• Article backed up locally (version control snapshot taken)
• Rollback procedure documented (specific commands/steps for reverting)
• Deployment approver identified (who will execute the deploy? who can approve rollback?)

---

Pre-Deployment File Integrity Check

Before deployment, verify the article file matches expectations:

HTML/Markup Validation

# Check for broken tags (example: validator.w3.org CLI)
# OR manual check in browser DevTools: Console tab for errors

# File size check (articles shouldn't exceed 2MB)
ls -lh article.html
# Expected output: -rw-r--r-- 1 user group 245K Apr 1 09:15 article.html

# Character encoding check (must be UTF-8)
file -b article.html
# Expected: HTML document, ASCII text

# Line count sanity check
wc -l article.html
# Expected: reasonable number (50–1000 depending on length)

Image Integrity Check

# All images present
grep -o 'src="[^"]*"' article.html | sort | uniq

# Image file sizes (each should be <200KB)
for img in images/*.jpg images/*.png; do ls -lh "$img"; done

Metadata Validation

# Check meta tags present
grep -E '(og:title|og:description|og:image|og:url)' article.html

# Check canonical URL
grep 'canonical' article.html

# Check structured data
grep 'application/ld+json' article.html

---

Deployment Execution

Step 1: Choose Your Platform

Your newsroom has configured deployment for one (or more) of these platforms. Identify which platform hosts the published article:

GCS (Google Cloud Storage):

# Verify article file exists locally
ls -l article.html

# Set project (if not already set)
gcloud config set project my-newsroom-project

# Deploy SINGLE article (not directory)
gsutil cp article.html gs://news-bucket/articles/2025-04-01-story.html

# Verify deployment (check file exists)
gsutil ls gs://news-bucket/articles/2025-04-01-story.html

S3 (Amazon Web Services):

# Verify credentials configured
aws sts get-caller-identity

# Deploy SINGLE article
aws s3 cp article.html s3://news-bucket/articles/2025-04-01-story.html

# Verify deployment
aws s3 ls s3://news-bucket/articles/2025-04-01-story.html

Static Site (SFTP):

# Connect and deploy SINGLE article (NOT directory)
sftp [e-mail] << EOF
cd public_html/articles/
put article.html 2025-04-01-story.html
bye
EOF

# Verify (optional: SSH in and check)
ssh [e-mail] "ls -lh /home/news/public_html/articles/2025-04-01-story.html"

Netlify (Git-based):

# Article is already in Git repo at correct path
git add articles/2025-04-01-story.html
git commit -m "Publish article: [headline]"
git push origin main

# Netlify auto-deploys; check deploy log:
# https://app.netlify.com/sites/[site-name]/deploys

WordPress (REST API):

# Create post via REST API
curl -X POST https://newsroom.org/wp-json/wp/v2/posts \
  -H "Authorization: Bearer $WP_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "title": "Article Title",
    "content": "<p>Article content here...</p>",
    "status": "publish",
    "featured_media": 123
  }'

# Expected response: HTTP 201 Created + post ID

Step 2: Deployment Safety Checks (CRITICAL)

BEFORE you execute deployment, answer these questions:

• Am I deploying exactly ONE article? (If deploying >1 file, STOP and verify with editor)
• Am I using the exact file path specified? (Copy-paste from pre-deployment checklist; no guessing)
• Am I NOT using wildcards or directories? (e.g., *.html or articles/ NOT allowed; only 2025-04-01-story.html)
• Have I tested the deployment command on a test file first? (If possible, deploy test file to test bucket; verify it works)
• Do I have a rollback command ready? (Documented and tested before deployment)

Red flag: If ANY of the above is "no," STOP. Re-read this section.

Step 3: Execute Deployment

# Log the deployment action (for audit trail)
echo "Deploying article: 2025-04-01-story.html | By: [e-mail] | Time: $(date)" >> deploy.log

# EXECUTE DEPLOYMENT (exact command from pre-deployment checklist)
# Example: GCS
gsutil cp article.html gs://news-bucket/articles/2025-04-01-story.html

# Capture exit code
DEPLOY_STATUS=$?

if [ $DEPLOY_STATUS -eq 0 ]; then
  echo "Deployment successful"
else
  echo "Deployment FAILED (exit code: $DEPLOY_STATUS)"
  exit 1
fi

---

Post-Deployment Verification

Immediately after deployment, verify the article is live and correct:

Step 1: Verify File Deployed

# GCS
gsutil ls -l gs://news-bucket/articles/2025-04-01-story.html

# S3
aws s3 ls s3://news-bucket/articles/ | grep 2025-04-01-story.html

# Static site (SFTP)
ssh [e-mail] "ls -lh /home/news/public_html/articles/2025-04-01-story.html"

Expected output: File exists, size matches original, timestamp is current (within last minute).

Step 2: Verify Article is Accessible

Test from two locations (to verify CDN propagation if applicable):

# Get the article via HTTP (standard browser request)
curl -s -o /dev/null -w "%{http_code}" https://newsroom.org/articles/2025-04-01-story.html

# Expected: HTTP 200 (success), NOT 404 (not found) or 503 (error)

# (Optional) Check full response
curl -s https://newsroom.org/articles/2025-04-01-story.html | head -20

Step 3: Verify Content Integrity

# Download article from published URL
curl -s https://newsroom.org/articles/2025-04-01-story.html > published.html

# Compare to original (should be identical or very similar)
diff article.html published.html

# If no differences: success
# If differences: investigate (may be auto-formatting by platform)

Step 4: Verify Metadata

Open the article in a browser and check:

• Headline displays correctly (no encoding errors, special characters)
• Images load (all images visible, no broken image icons)
• Links work (click a few; confirm they open expected pages)
• Byline present (author name, publication date visible)
• Social preview correct (right-click; inspect Open Graph tags; should show correct og:image, og:title)

Step 5: Verify Social Metadata

# Use platform API to check Open Graph tags
# Facebook Sharing Debugger
# https://developers.facebook.com/tools/debug/

# Twitter Card Validator
# https://cards-dev.twitter.com/validator

# OR use curl to check raw tags
curl -s https://newsroom.org/articles/2025-04-01-story.html | grep -E 'og:(title|description|image|url)'

Step 6: Search Engine Indexing (optional, but check after a few hours)

# Google Search Console (manual check; not API-able)
# Go to: https://search.google.com/search-console/
# Inspect URL: https://newsroom.org/articles/2025-04-01-story.html
# Request indexing if not yet crawled

# (Advanced) Manually trigger crawler
curl -X POST 'https://www.google.com/ping?sitemap=https://newsroom.org/sitemap.xml'

---

Rollback Procedure

If post-deployment verification fails, roll back immediately:

Automatic Rollback (if using version control)

# For Git-based deployment (Netlify, Vercel, GitHub Pages)
# Find the commit that deployed the article
git log --oneline | head -5

# Revert the commit
git revert [commit-hash]
git push origin main

# Netlify/Vercel will auto-redeploy previous version
# Check status: https://app.netlify.com/sites/[site]/deploys

Manual Rollback (GCS/S3)

# Option A: Delete the article (removes it entirely)
gsutil rm gs://news-bucket/articles/2025-04-01-story.html

# Option B: Restore from backup (if versioning enabled)
# GCS
gsutil cp gs://news-bucket/articles-backup/2025-04-01-story.html.bak \
  gs://news-bucket/articles/2025-04-01-story.html

# S3
aws s3 cp s3://news-bucket/articles-backup/2025-04-01-story.html.bak \
  s3://news-bucket/articles/2025-04-01-story.html

Manual Rollback (Static Site / SFTP)

# Option A: Delete from live server
ssh [e-mail] "rm /home/news/public_html/articles/2025-04-01-story.html"

# Option B: Restore from backup (pre-deployment snapshot)
scp articles-backup/2025-04-01-story.html.bak [e-mail]:/home/news/public_html/articles/2025-04-01-story.html

Manual Rollback (WordPress)

# Get the post ID
POST_ID=$(curl -s 'https://newsroom.org/wp-json/wp/v2/posts?search=Article%20Title&_fields=id' \
  -H "Authorization: Bearer $WP_TOKEN" | jq '.[0].id')

# Delete the post
curl -X DELETE https://newsroom.org/wp-json/wp/v2/posts/$POST_ID \
  -H "Authorization: Bearer $WP_TOKEN" \
  -d '{"force": true}'

Document the Rollback

rollback:
  date: "2025-04-01 10:15 AM"
  reason: "Images failed to load; metadata incomplete"
  by: "[e-mail]"
  command: "gsutil rm gs://news-bucket/articles/2025-04-01-story.html"
  status: "success"
  next_steps: "Re-deploy after fixing image paths"

---

Common Mistakes & Red Flags

Deployment Mistakes

Mistake	Why It's Bad	Fix
Deploying directory instead of file (gsutil -m cp articles/* gs://bucket/)	May overwrite other articles; creates chaos	Use explicit filename: gsutil cp articles/2025-04-01-story.html gs://bucket/articles/...
Using wildcard (aws s3 cp *.html s3://bucket/)	Uploads everything in current directory; disaster	Deploy one file at a time; verify path before executing
Deploying unapproved version (old draft instead of final)	Wrong content goes live; requires emergency rollback	Triple-check file path and version before deploy command
No backup before deploying (no rollback possible)	If something goes wrong, can't recover	Always maintain a pre-deployment backup (version control or manual copy)
Deploying incomplete article (missing images, metadata)	Broken page goes live; low user experience	Run pre-deployment checklist; verify all images uploaded
Forgetting to update sitemap (article not discoverable in search)	Search engines don't index; no organic traffic	After deployment, update sitemap.xml; ping Google

Verification Mistakes

Mistake	Why It's Bad	Fix
Not testing from live URL (only checking local file)	May be cached differently on live server; may not work	Always test via actual published URL; use multiple browsers
Assuming deployment succeeded without checking	Article may have failed silently; readers see 404	Always verify HTTP 200 + content check after deployment
Only checking one aspect (images but not metadata)	Metadata missing; social cards don't display	Use full post-deployment checklist (5 steps)
Checking immediately after deploy (before CDN propagates)	May see old cached version; false success	Wait 30 seconds; check from 2+ locations; check multiple times
Not testing on mobile	Mobile users see broken layout; don't discover problem	Test via mobile browser or responsive view in DevTools

Rollback Mistakes

Mistake	Why It's Bad	Fix
Trying to "fix" the live article instead of rolling back	May make it worse; wastes time; better to start over	Roll back immediately; fix offline; redeploy fresh
Deleting article without checking if it's still being read	Early readers miss article; confusion; embarrassment	Rollback ASAP; add note to homepage ("Article temporarily removed for updates")
Rollback command typed wrong (deletes wrong article)	Deletes innocent article; chaos; reputation damage	Always test rollback command on test data first
No rollback procedure documented	When disaster happens, don't know what to do	BEFORE deployment, document and test rollback steps

---

Critical Safety Rules

Rule 1: File-by-File Only

NEVER use:

# WRONG: Recursive deploy
gsutil -m cp articles/* gs://bucket/articles/

# WRONG: Wildcard deploy
aws s3 cp *.html s3://bucket/

# WRONG: Directory deploy
sftp news@server << EOF
  cd public_html
  put -r articles/
EOF

ALWAYS use:

# RIGHT: Single file
gsutil cp articles/2025-04-01-story.html gs://bucket/articles/2025-04-01-story.html

# RIGHT: Explicit path
aws s3 cp articles/2025-04-01-story.html s3://bucket/articles/2025-04-01-story.html

Rule 2: Atomic Deployment (All or Nothing)

Either the article is fully deployed and verified, or it's rolled back. No partial/broken states.

# Deploy
gsutil cp article.html gs://bucket/articles/2025-04-01-story.html

# Verify
curl -s https://newsroom.org/articles/2025-04-01-story.html | grep -q "Article Title"

# If verification fails, rollback immediately
if [ $? -ne 0 ]; then
  echo "Verification failed. Rolling back..."
  gsutil rm gs://bucket/articles/2025-04-01-story.html
  exit 1
fi

echo "Deployment and verification successful"

Rule 3: Editorial Approval Before Deployment

DO NOT DEPLOY until all gates are passed:

• Editorial approved (final-editor-review complete)
• Copy review approved
• Fact-check approved
• Pre-deployment checklist signed off

---

Deployment Workflow Checklist

Pre-Deployment (30 min before go-live)

• Gather approvals (editorial, copy, fact-check sign-off)
• Run pre-deployment checklist (all items marked complete)
• Verify file integrity (HTML valid, images present, metadata complete)
• Prepare rollback procedure (document exact commands)
• Brief deployment approver (who will authorize? how to reach?)

Deployment (go-live moment)

• Answer safety questions (single file? exact path? no wildcards?)
• Execute deployment command (copy-paste exact command from checklist)
• Document deployment action (log who/what/when)
• Capture exit code (0 = success, non-zero = failure)

Post-Deployment (30 min after go-live)

• Verify file deployed (ls/check confirms file exists)
• Verify HTTP 200 (article loads in browser)
• Content integrity check (diff original vs. published)
• Metadata verification (og:tags, canonical, structured data correct)
• Social preview check (og:image displays correctly)
• Manual spot-check (open in browser; check images, links, byline)

If Something Goes Wrong

• Stop immediately (don't try to patch)
• Roll back (execute pre-prepared rollback command)
• Document issue (what failed? why?)
• Fix offline (correct the problem locally)
• Re-deploy (full deployment + verification cycle again)

---

Escrow & Collaboration

If article is collaborative (multiple reporters, editors):

1. One person owns deployment (prevent multiple deployments of same article)
2. Deployment approver is editor (signs off on checklist)
3. Deployment executor is tech/manager (actually runs commands)
4. Rollback authority is editor (decides to roll back if issues arise)

Communication protocol:

Reporter Alice: "Article ready for deployment checklist"
Editor Bob: "Approved. Running checklist... ✓ All items complete. Deploying at 9:00 AM."
Tech Manager Carol: "Confirmed. Standing by for verification."
[Deployment at 9:00 AM]
Carol: "Article deployed. Running verification... ✓ All checks pass. Live!"
Bob: "Confirmed. Article is live. Notifying social team."

---

Escalation

If deployment fails or article has issues post-publication:

1. Editors: Decide whether to roll back or push forward
2. Tech: Executes rollback or deployment fix
3. Communications: Update homepage/social if article is pulled
4. Legal (if needed): Review article for liability if corrections posted