content-production

Content Production Workflow

This is mandatory. No content ships without review.

Core Rule

Every content task follows: Produce → Review → Revise → Deliver

Never deliver a first draft directly. Always spawn a review agent.

---

Workflow

Step 1: Produce

Write the content using the appropriate skill:

• SEO blog articles → seo-blog-writer/SKILL.md (fully automated: keyword research → competitive analysis → write → QA → delivery package). Mandatory dependency: heavy-task-protocol/SKILL.md — load it first for checkpoint/resume safety.
• Marketing copy → your copywriting skill
• Comparison pages → your competitor-alternatives skill
• Social posts → your social-content skill
• Email sequences → your email-sequence skill
• Landing pages → your page-cro skill + copywriting skill
• Release notes → your release-notes skill

Step 2: Auto-Review

Immediately after producing, spawn a review sub-agent with the appropriate review profile below. Use a cheaper model (sonnet) for review to save cost.

Step 3: Revise

Apply review feedback. If critical issues found (broken links, factual errors, brand violations), fix and re-review.

Step 4: Deliver

Only after review passes, deliver to the user.

Step 5: Post-Publish Verification (separate session after publishing)

This step runs in a NEW session after the article is live. Do not chain it into the same writing session — context would be too large and external dependencies (seomator, PSI API) add failure risk.

Trigger: user confirms the article is published and provides the live URL.

python3 seo-geo-qa/scripts/post_publish_check.py https://your-domain.com/blog/your-slug

# With Core Web Vitals (if PSI API key is configured)
python3 seo-geo-qa/scripts/post_publish_check.py https://your-domain.com/blog/your-slug --psi-key YOUR_KEY

Verdict:

• PASS — no action needed
• WARN — review seomator findings, fix high-priority items
• FAIL — fix critical issues (noindex, broken canonical, missing H1) before the page gets indexed

---

SEO Blog Pipeline (Full Lifecycle)

Session 1 — Write + QA (heavy task, checkpoint protocol required):

Load heavy-task-protocol → Load seo-blog-writer
→ Phase 0: Context discovery + confirmation gate
→ Phase 1–4: Research, write, QA, iterate
→ Decision-Grade Quality Gate (before Phase 5)
→ Write draft + QA report to disk
→ Spawn content-qa review agent
→ Revise until PASS
→ Deliver

Session 2 — Publish (human-driven):

Take draft from disk → publish to CMS/repo

Session 3 — Post-publish (separate, lightweight):

Run post_publish_check.py on live URL → report verdict

---

Review Profiles

SEO Blog Article

Reference docs:

• content-qa/SKILL.md — QA checklist framework
• seo-geo-qa/SKILL.md — technical SEO checks and link verification
• brand/voice.md — brand voice, red lines, tone guidelines (provide via Brand Context input)
• product/ — product facts (pricing, features, platforms)

Checklist:

• All external links verified (no 404/403)
• All internal links point to real pages on your domain
• Word count meets target (2500-3500 for comparison articles)
• Title includes primary keyword
• Meta description under 160 chars, includes keyword
• H2/H3 structure is logical
• Comparison table present and accurate
• Product info is accurate (check product/ context)
• Competitor info is current (pricing, features)
• No brand voice violations (check brand/ context)
• FAQ section present (for schema markup)
• Internal links to other blog posts included
• CTA present
• No fluff paragraphs
• Decision-Grade Quality Gate passed (exclusion boundaries, ranking fallback, decision engine, convergence summary)

Social Post (X / LinkedIn)

Reference docs:

• Your social content skill
• brand/voice.md — brand voice guidelines

Checklist:

• Within character/format limits for platform
• Stays in product's topic lane
• No hashtag spam (0-1 max for X)
• Matches brand voice
• Clear point or take
• Engagement hook or CTA present

Email Sequence

Reference docs:

• Your email-sequence skill
• brand/voice.md

Checklist:

• Subject line under 50 characters
• No spam trigger words in subject
• Unsubscribe link present
• Each email has ONE primary CTA
• Sequence logic makes sense
• All links work

Ad Copy

Reference docs:

• Your paid-ads skill
• brand/voice.md

Checklist:

• Character limits met for platform
• CTA clear
• Value prop in first line
• No competitor names in ad (policy risk)
• Matches landing page messaging

Landing Page

Reference docs:

• Your page-cro skill
• Your schema-markup skill
• brand/voice.md

Checklist:

• Above-fold: headline + value prop + CTA visible without scrolling
• Single primary CTA
• Social proof present
• H1 contains target keyword
• Meta description optimized
• Schema markup suggested

---

Review Agent Spawn Template

sessions_spawn:
  task: |
    You are a content QA reviewer. Read content-qa/SKILL.md for your full checklist and protocol.

    Review the content at [FILE_PATH].

    Content Type: [seo-blog | ad-copy | social-post | email-sequence | landing-page]
    Target Keyword: [keyword] (SEO blog only)
    Brand Context: [path to brand/voice.md or inline brand guidelines]
    Round: 1

    Follow the full checklist for this content type in content-qa/SKILL.md.
    Use the automated QA runner (content-qa/scripts/run_qa.sh) for Step 0.
    Calculate SEO Quality Score (0-100) for blog articles.

    Output format: as defined in content-qa/SKILL.md (PASS ✅ / FAIL ❌).
  mode: run
  model: sonnet

---

When to Skip Review

Never. But you can use a lighter review for:

• Internal notes/memos (no review needed)
• Casual conversational replies

Everything that goes public gets reviewed.