Content Workflow · ReduceOps Marketing Agent

Article Pipeline

From topic idea to live post — research the SERP, draft and score with 7 agents, then ship to Hugo in one command.

34articles in calendar
7auto-triggered agents
≥70quality score to publish
5scoring dimensions
01
1Research
Pick, Dig, Plan
Check the content calendar, pull SERP data, and write a structured article plan before a word of prose is drafted.
📋
/priorities
Reads CONTENT_CALENDAR.md and returns the next unwritten article — topic, target keyword, pillar, and SEO rationale.
🔍
/research-serp [keyword]
Analyzes top SERP results for the target keyword — identifies competitor angles, content gaps, and ranking signals to beat.
📡
/research [topic]
Compiles a research brief with social signals, trending angles, and competitor weaknesses. Saved to content-pipeline/research/ as a dated markdown file.
article plan generated
🗺️
Article Plan
Structured outline saved as article-plan-[topic]-[date].md covering H2/H3 structure, keyword placement targets, CTA placement, and internal link anchors from the 39-article linking map.
Research Outputs
briefcompetitive brief
serpSERP analysis
socialsocial signals
planarticle outline
02
2Write
Draft, Scrub, Score
Seven agents polish the raw draft — removing AI fingerprints, optimizing SEO, and generating internal link suggestions until the composite score clears 70.
Agent Pipeline (auto-triggered by /article)
1
content-analyzer Quality and structure analysis — flags issues before other agents run
→ SCAN
2
editor Humanizes text, aligns with ReduceOps brand voice — direct, technical, no-BS
→ EDIT
3
/scrub Strips invisible Unicode (zero-width spaces, BOM, soft hyphens) and replaces AI em-dash patterns
→ CLEAN
4
seo-optimizer Checks keyword in H1, meta title 50-60 chars, meta description 150-160 chars, keyword in first 100 words
→ SEO
5
meta-creator Generates meta title, meta description, OG tags, and Twitter card data
→ META
6
internal-linker Recommends internal links from the 39-article linking network in internal-links-map.md
→ LINK
7
keyword-mapper Checks keyword density and distribution — target primary keyword at 1-2% density
→ MAP
quality gate
5-Dimension Scoring (≥70 composite required)
H
Humanity (30%) AI phrase density <5/1000 words, contractions present, active voice dominant
S
Specificity (25%) Numbers, dollar amounts, percentages, and dates — not "many" or "significant"
B
Structure Balance (20%) 50-75% prose — not all bullets, not wall-of-text
E
SEO (15%) Meta title, meta description, keyword in H1 and first 100 words, ≥2,000 words
R
Readability (10%) Flesch 60-70, grade 8-10, paragraphs ≤4 sentences, varied sentence rhythm
✓ Score ≥70
Draft moves to human review, then publish
✕ Score <70
Top 5 priority fixes returned, revise and re-score
03
3Publish
Image, Ship, Deploy
Generate a featured image via Replicate, transform frontmatter for Hugo, commit to git, and Cloudflare Pages auto-deploys within seconds.
Pre-publish Steps
🎨
/generate-image [topic]
Calls Replicate MCP → SeedReam 4 model — outputs featured image to landing/content/images/
👁️
Human Review
Read draft, approve or request changes — the one required human checkpoint
approved for publish
/publish [file] Steps
🔄
Transform Frontmatter
Converts pipeline metadata to Hugo front matter (title, date, tags, featured image path, slug)
📁
Copy to Hugo
Writes to landing/hugo/content/blog/[slug].md — Hugo picks it up on next build
💾
Archive to Published
Moves source file to content-pipeline/published/ with a datestamped filename
🚀
Git Commit → Deploy
Commits to master → Cloudflare Pages auto-deploys the Hugo build (cd landing/hugo && hugo --minify)
Guardrails
≥2,000words minimum
1 humanreview required
⚡ No auto-publish— human approval gates every post