AI Content Creation for B2B Marketing: Build a Blog Writing Agent With Claude Cowork

Create the project and skill directory

Open Claude Desktop and go to the Cowork tab. Create a new project called blog-agent. Point it at a folder on your machine, something like ~/Documents/blog-agent/. Inside that folder, create a subdirectory called skills/. This is where your three skill files will live.

While you are in settings, also connect the Buffer MCP if you want the agent to schedule posts directly. Go to Settings, Connectors, Add custom connector, and paste https://mcp.buffer.com/mcp. Log into your Buffer account to authorise. The connector lets Claude read your Buffer queue and schedule new content, same as the official Buffer API.

Verify it works: ask Claude Cowork to list the skill files in your project. If it returns an empty list, the directory structure is correct and Claude can see it. If it returns nothing or an error, check the folder path in your Cowork project settings.

Create the research skill file

Create a file called skills/research.md in your project folder. This skill tells Claude how to research before writing a single word. Here is the template we use with clients at Supernodes:

# Research Skill

## Purpose
Before writing any blog post, gather evidence from the web.

## Research rules
1. Search for three statistics from named sources published in 2025 or 2026.
2. Find two competitor articles on the same topic and note what they miss.
3. Identify one unique angle not covered by competitors.
4. Collect the source URL, publication date, and exact wording for every claim.

## Output format
Return your findings in this structure:
- Primary stat (source, date, exact quote)
- Secondary stats (source, date, exact quote)
- Competitor gap: what the top results miss
- Unique angle: what this post adds that others do not

This file takes about 15 minutes to write. Once it exists, Claude loads it automatically before every blog task. The research step, which normally takes 45 to 60 minutes of human browsing, happens in about 30 seconds. According to practitioners who have built similar pipelines, the research quality from a structured skill like this is comparable to a junior content researcher on a first pass, and it catches sources a human might miss because it searches more broadly in less time.

Write the brand voice and structure rules

This is the skill that matters most. Without it, Claude writes generic blog posts. With it, every post follows the same tone, structure, and rules. Create skills/brand-voice.md:

# Brand Voice Skill

## Tone
Professional but direct. No jargon. Company voice ("we"), not personal ("I").
Keep paragraphs to 3 to 5 sentences. Vary sentence length.

## Audience
B2B marketing leaders at companies with 10 to 200 employees.
They know marketing but are evaluating AI tools. Speak to their outcomes, not the technology.

## Structure
Every blog post follows this structure:
1. Opening paragraph: arresting stat or scenario, then the solution promise
2. "What you need" section: tools and cost, before any setup steps
3. Numbered steps with exact paths, commands, and verification checks
4. Results section: time saved, output comparison
5. FAQ section with 5 to 6 questions

## Banned phrases
"Leverage", "robust", "seamless", "game-changer", "unlock", "delve into",
"in today's fast-paced", "it is not X, it is Y"

## Capitalisation and formatting
Normal capitalisation rules. Proper nouns capitalised. Sentence case headings.
No em dashes. Use commas and full stops instead.

According to the complete Claude Cowork guide from BuildFastWithAI, brand voice skills are the single biggest determinant of output quality. Users who skip this step get generic content. Users who spend 30 minutes on a voice file get output that matches their existing content more consistently than multiple freelance writers would. The file is plain markdown. If you can write a Notion page, you can write a Claude Skill.

Create the editing and quality-control skill

The editing step is where most AI content pipelines fall apart. The draft looks good on first read, but the stats are unsourced, the tone drifts, and a few sentences sound unmistakably AI-generated. An editing skill catches this before a human ever looks at it. Create skills/edit.md:

# Editing Skill

## Purpose
Review every blog draft before it reaches a human editor.

## Checks (run in order)
1. Fact check: every statistic must have a named source. Flag any stat without one.
2. TOV check: scan for banned phrases from the brand voice skill. Flag every violation.
3. AI-tell check: flag sentences that start with "It is not X, it is Y",
   colon constructions introducing clauses, and consecutive sentences with identical structure.
4. Link check: every external source name should be a hyperlink. Flag any that are not.
5. Readability: flag any sentence over 30 words.

## Action
For each flagged issue, rewrite the sentence and present both versions.
The human editor picks which version to keep.

One of our clients started running this editing step as a gate before the human review pass and found that the editor's time per post dropped from roughly 20 minutes to about 8 minutes. The agent caught the obvious issues, so the human only had to make strategic decisions. According to NeuraPulse's 2026 analysis of Claude blog automation, fully automated pipelines with an editing gate reduce total production time from 8 hours to roughly 45 minutes per post, with the human only touching the final polish.

Run the full research, write, and edit pipeline

Now the three skills are in place. Open Claude Cowork and give it this prompt:

"Research and write a blog post about [your topic].
Load the research skill, then the brand voice skill, then the editing skill.
Run them in that order. Research first, write second, edit third.
Return the final draft with the editing annotations."

Claude will search for statistics, competitor articles, and a unique angle. Then it writes the post following your brand voice skill. Then it runs the editing skill against its own output and flags issues. The whole pipeline runs in about two to three minutes for a standard 1,500-word post. What you get back is a draft that has been researched, written to your brand voice, and self-edited. A human still needs to review it, but the heavy lifting is done.

If you want the agent to publish directly, add a fourth instruction: "Schedule the final version to Buffer for next Tuesday at 9am." Using the Buffer MCP connector, Claude pushes the post to your queue with the correct timing. The same connector lets you edit scheduled posts by saying "Change the opening of the LinkedIn version to this instead."

Set up a recurring content cadence

Once the pipeline works for one post, the obvious next step is to make it recurring. In Claude Cowork, type:

/schedule Every Monday at 10am, check my blog topics file for new entries.
If there is a new topic, run the research, write, and edit pipeline using the three skills.
Save the draft to the drafts folder and notify me.

Claude asks you to confirm the schedule parameters. Once confirmed, it runs autonomously every Monday. By Monday afternoon, you have a researched and edited draft waiting in your drafts folder, and your only job is the strategic review. Teams using this pattern report going from two posts per month to three to five per week, according to practitioners on Reddit's ClaudeAI community who have built similar setups.

For teams already running content repurposing with the Claude Cowork and Buffer MCP workflow, this blog writing agent slots in before the repurposing step. The agent writes the blog post. The repurposing pipeline turns it into ten social pieces. You go from topic to published blog plus full social distribution with roughly 15 minutes of human involvement.