Case Study: From Architecture Doc to LinkedIn Post in 10 Minutes

Most people write a blog post, then think about how to promote it. The CMO agent pipeline does both at once — turning technical architecture documentation into polished LinkedIn content in under 10 minutes.

This is how it works.

---

The Pipeline in One Diagram

MASTER-ARCHITECTURE.md
        ↓
   [CTO Agent writes docs]
        ↓
 9 architecture pages +
  3 framework pages
        ↓
   [CMO Agent reads docs]
        ↓
  5 LinkedIn drafts
        ↓
  [Infographic Agent]
        ↓
  5 visual slide decks
        ↓
   [Strategy Doc]
        ↓
  Schedule + posting plan

Each arrow is a separate agent. Each output feeds the next. The human touches nothing between “write the architecture doc” and “here are 5 posts ready to schedule.”

---

Step 1: The Source Document

Time: ~20 minutes (human + CEO agent)
Output: MASTER-ARCHITECTURE.md

Everything starts with MASTER-ARCHITECTURE.md — a single canonical document describing the entire agent organization system:

• What each C-suite agent does
• How models are routed (Haiku vs Sonnet vs Opus)
• Delegation rules and thresholds
• Memory architecture
• Cron job philosophy
• Integration patterns

This document serves two masters: it’s both a technical spec (for agents building the system) and a content source (for agents writing about the system).

Tip

Writing a master architecture document first isn’t just good engineering — it’s a content goldmine. Every paragraph is a potential post, thread, or article.

The key insight: write once, publish many times. The master doc becomes the single source of truth that downstream agents pull from. When the architecture changes, you update one document and re-run the pipeline.

---

Step 2: Documentation Generation

Agent: CTO sub-agent (claude-sonnet-4-6)
Time: ~15 minutes (4 parallel sub-agents)
Output: 12 MDX pages

The CTO agent read MASTER-ARCHITECTURE.md and spawned 4 parallel documentation writers:

Architecture pages (9):

• C-Suite Overview
• CEO Agent deep-dive
• COO Agent deep-dive
• CTO Agent deep-dive
• CMO Agent deep-dive
• CIO Agent deep-dive
• Memory Architecture
• Model Routing
• Cron Philosophy

Framework pages (3):

• LLM Routing Strategy
• Delegation Framework
• Agent Requirement Documents

Each page follows the Astro Starlight MDX format: frontmatter, intro, structured sections, code examples, callout boxes. The agents weren’t writing from scratch — they were transforming structured architecture content into structured documentation content.

Why Sonnet for this?
Documentation requires nuance. Haiku produces accurate but flat prose. Sonnet understands audience (technical readers who want both depth and clarity) and adjusts tone accordingly.

---

Step 3: LinkedIn Post Drafting

Agent: CMO sub-agent (claude-sonnet-4-6)
Time: ~8 minutes
Output: 5 LinkedIn drafts

With documentation written, the CMO agent ran a content extraction pass:

Input: All 12 documentation pages
Task: Identify the 5 most LinkedIn-worthy insights and draft posts

The CMO agent’s selection criteria:

1. Contrarian angle — what does this system believe that most people don’t?
2. Concrete numbers — costs, times, ratios that make people stop scrolling
3. Teachable moment — one clear thing the reader can take away
4. Story hook — does it open with a scene, a problem, or a surprising fact?

The 5 posts drafted:

Post	Hook	Core Insight
CEO Pattern	”I stopped coding. My agents code for me.”	Orchestration > execution
Model Routing	”$0.0003 vs $0.015 per call”	Right model for the right task
90-Minute Build	”15 tasks. 90 minutes. $1.02.”	Parallel agents compress calendar time
Memory Architecture	”Your agent forgets everything. Here’s the fix.”	Three-layer memory system
The Failure Post	”2 agents crashed. Here’s what happened.”	Failure recovery > failure prevention

Note

The “failure post” consistently outperforms the “success post” on LinkedIn. People trust vulnerability. The CMO agent knows this and actively surfaces failure stories from the architecture documentation.

---

Step 4: Infographic Generation

Agent: Infographic sub-agent (claude-sonnet-4-6 + linkedin-pdf skill)
Time: ~12 minutes
Output: 5 visual slide decks (PDF format)

LinkedIn posts with visuals get 3× more engagement than text-only posts. The infographic agent paired each draft with a visual:

Visual formats used:

Post	Visual Type	Key Element
CEO Pattern	Org chart	Agent hierarchy diagram
Model Routing	Comparison table	Haiku/Sonnet/Opus cost vs quality
90-Minute Build	Timeline	Phase-by-phase with icons
Memory Architecture	Architecture diagram	Three-layer memory stack
Failure Post	Before/after	Broken state → recovered state

The linkedin-pdf skill pipeline:

1. Parse the LinkedIn post draft for key concepts
2. Generate slide structure (title + 3-5 content slides + CTA)
3. Render slides as images using Pillow
4. Assemble into PDF via moviepy/reportlab
5. Return file path for upload

Each infographic takes ~2.5 minutes to generate. Running 5 in sequence: ~12 minutes. (Parallel execution would cut this to ~3 minutes — a future optimization.)

---

Step 5: Strategy Document

Agent: CMO sub-agent (same session, second task)
Time: ~3 minutes
Output: ~/clawd/agents/CMO/content-strategy-march-2026.md

With 5 posts and 5 visuals ready, the CMO agent wrote a scheduling strategy:

## Posting Schedule - Week of March 31

Mon 9am: CEO Pattern post + org chart visual
Tue 11am: Model Routing post + comparison table
Wed 9am: 90-Minute Build post + timeline visual
Thu 2pm: Memory Architecture post + diagram
Fri 10am: Failure Post + before/after visual

## Engagement Strategy
- Reply to every comment within 2 hours (first 24h)
- Repost top-performing post on Sunday
- Track: impressions, comments, follows gained per post
- Goal: 500 impressions per post minimum

The strategy doc also includes:

• Hashtag recommendations per post
• Optimal posting times based on audience timezone
• Cross-promotion opportunities (tag relevant accounts)
• A/B test suggestion: post 1 with vs. without first comment

---

Full Pipeline: Time + Cost

Step	Agent	Time	Model	Est. Cost
Master doc	Human + CEO	20 min	Opus	$0.18
Documentation (12 pages)	CTO × 4 parallel	15 min	Sonnet	$0.31
LinkedIn drafts (5 posts)	CMO	8 min	Sonnet	$0.12
Infographics (5 visuals)	Infographic	12 min	Sonnet	$0.19
Strategy doc	CMO	3 min	Sonnet	$0.04
Total		58 min		$0.84

Tip

The “10 minutes” in the title refers to the agent-only time after the architecture doc exists. The full pipeline including writing the master doc is ~58 minutes. Either way: less than an hour for 12 docs + 5 posts + 5 visuals + strategy.

---

Why This Pipeline Works

Single Source of Truth

MASTER-ARCHITECTURE.md is updated once. Every downstream artifact regenerates from it. No copy-paste drift between the docs and the posts.

Content Multiplier Effect

One architecture document → 12 tech docs + 5 posts + 5 visuals = 22 publishable assets. The marginal cost of each additional asset approaches zero because the source material already exists.

Tone Consistency

Both the docs and the posts were drafted by agents reading the same source material. The vocabulary, framing, and key concepts are naturally aligned. You don’t get posts that contradict the docs.

Failure Resilience

Each step is independent. If the infographic agent fails, the posts are still ready. If one post draft is rejected, the other four still ship. The pipeline degrades gracefully.

---

What the CMO Agent Does That Humans Don’t

Humans:

• Write the architecture doc (requires deep system knowledge)
• Review and approve posts before scheduling (final judgment call)

CMO Agent:

• Extract LinkedIn-worthy insights from dense technical prose
• Reframe engineering concepts for business audiences
• Match visual format to content type
• Write scheduling strategy with engagement tactics
• Track performance and iterate

The human stays in their zone of genius (architecture, strategy). The agent handles the translation layer (technical → accessible) and the operational work (scheduling, formatting, visuals).

---

Extending the Pipeline

Current state: Architecture → LinkedIn
Next version: Architecture → LinkedIn + Twitter + Blog posts + Newsletter

The same source document, the same pipeline structure, additional output channels. Each new channel adds one more agent at the end of the chain.

Planned additions:

• twitter-thread agent: converts LinkedIn posts to Twitter thread format
• blog-post agent: expands LinkedIn posts into 800-word articles
• newsletter agent: curates week’s posts into Friday digest
• performance-feedback agent: reads engagement metrics, surfaces what to post more of

---

Replicating This Pipeline

Want to run this yourself? The pieces you need:

1. A master architecture document — your system’s single source of truth
2. linkedin-pdf skill — for infographic generation
3. CMO agent workspace — ~/clawd/agents/CMO/WORKSPACE.md
4. Content strategy template — modify the scheduling format for your audience
5. LinkedIn posting cadence — commit to a frequency before you automate it

The pipeline is only as good as the source material. A rich, opinionated architecture document produces better posts than a vague one. Invest in the master doc first.

---

This case study documents the content pipeline run on March 27, 2026. The 5 LinkedIn posts from this pipeline are scheduled for the week of March 31, 2026.

Enjoyed this? Get weekly agent architecture insights.

Subscribe and get a free Agent Architecture Cheatsheet.

Subscribe