Automating Content and Blog Post Research

Content creation is one of the highest-leverage areas for the Cosine CLI. By building a dedicated folder for your content work, you create an agent that knows your brand voice, your past work, and your content strategy — and can produce consistent output without being re-briefed every time.

Setting Up a Content Agent Folder

1. Create a folder for your content work:

   mkdir ~/blog-posts
   cd ~/blog-posts
   cos init

2. Edit the agent.md to define the agent’s persona and workflow. For example:

   You are an expert content strategist and copywriter for Cosine — an AI software engineering agent.
   
   Voice and tone:
   - Confident, clear, and direct
   - No emojis or filler phrases
   - Punchy paragraphs — max 3 sentences each
   - Write for a technical audience but keep it accessible
   
   Workflow:
   - Before drafting anything, save research notes as `.md` files in this folder
   - Check existing posts in this folder before starting to maintain consistency
   - Always reference our product vision when describing Cosine's capabilities

Researching Competitors and Writing Posts

With the folder set up, you can give the agent a broad content task and let it run:

“Research what our competitors (v0, Cursor, Claude Code) say about AWS Marketplace. Look at our own offering. Then write a blog post comparing them, emphasising our strengths.”

Switch to Auto mode (Shift+Tab) and let it work. Come back when it’s done — you don’t need to watch it.

Using Swarm Mode for Multiple Posts

When you want to produce several posts at once, use Swarm Mode:

1. First use Plan Mode to let the agent outline the topics and approach.
2. Then accept the plan with “Yes, accept and enable Swarm mode”.
3. The orchestrator spawns subagents — one per blog post — and they all write in parallel.
4. Come back to multiple completed drafts.

For example: “Research and write 10 blog posts covering the top themes from our recent customer calls.” With Swarm Mode, that’s a single prompt — not 10 separate sessions.

Maintaining Consistency Over Time

When you return to this folder weeks later, the agent still has:

• All your previous posts as examples to mimic.
• Your agent.md rules defining voice and tone.
• Research files saved from previous sessions.

You can say: “Write a new post about [topic]. Look at our last 5 posts for style guidance.” The agent will read those posts and match the tone automatically.

Feeding In External Research

If you want the agent to incorporate content from external sources — say, a Notion page with your product positioning — you can either:

• Use the Notion MCP to pull the page directly (see Setting Up MCPs and MCP Examples).
• Ask the agent to visit a specific URL and summarise it into the folder.

“Go to [URL] and save a summary of the key points to competitor-research.md in this folder.”

Then reference that file in future prompts.

Key Takeaways

• A dedicated content folder with a well-defined agent.md is the foundation of a scalable content workflow.
• The agent learns your voice over time by referencing past posts in the folder.
• Use Swarm Mode to produce multiple pieces of content in parallel.
• Feed in external research via MCPs or direct URL summarisation.

---

Next: Building a Summarisation Agent