content-builder-agent

🎨 Content Builder Agent

This example demonstrates the Multi-Modal Asset Pattern. It shows how a single orchestrator can use Memory, Skills, and Subagents to produce a complete content package: a researched blog post paired with AI-generated hero images and social media snippets.

🔍 Deep Dive: The Three Primitives of an Agent

Deep Agents are built on three foundational layers:

1. Memory (AGENTS.md): The "Soul"—defines the brand voice, tone, and editorial standards.
2. Skills (skills/): The "Talent"—specialized workflows for different formats (Blog vs LinkedIn).
3. Subagents (subagents.yaml): The "Staff"—delegated workers for tasks like depth research or image generation.

By separating these, you can update a brand voice for an entire organization without changing a single line of logic in the skills.

Content Creation Pipeline

graph TD
    User([Task: Write Blog]) --> Agent[Main Content Agent]
    Agent --> Skill[Load Skill: blog-post/social-media]
    
    subgraph Research_Phase [1. Research]
        Agent --> Researcher[Task: Researcher Subagent]
        Researcher --> Web[Web Search: Tavily]
        Web --> Notes[Save to research/]
    end
    
    subgraph Draft_Phase [2. Drafting]
        Notes --> Draft[Draft Content per Skill]
        Draft --> SaveContent[Save to blogs/ or linkedin/]
    end
    
    subgraph Asset_Phase [3. Creative]
        Draft --> ImageGen[Generate Cover: Gemini]
        ImageGen --> SaveImage[Save hero.png]
    end
    
    SaveContent --> User
    SaveImage --> User

Loading

🛠️ Module Setup

Prerequisites

• Python 3.11+
• API Keys:
  • ANTHROPIC_API_KEY: For the main writer brain.
  • GOOGLE_API_KEY: For the generate_cover tool (uses Gemini).
  • TAVILY_API_KEY: For the research subagent.

Installation & Launch

cd examples/content-builder-agent
uv sync

# Run the builder for a specific topic
uv run python content_writer.py "Write a blog post about prompt engineering"

🛑 Troubleshooting & Common Pitfalls

• "Image generation failed": This example uses Gemini for images. Ensure your GOOGLE_API_KEY has access to the Imagen models.
• "Voice is too robotic": If the output doesn't match your brand, edit AGENTS.md. This file is the primary source of truth for the agent's writing style.

✅ Self-Check Challenge

• Look at skills/blog-post/SKILL.md. What specific sections of a blog post does the agent try to write first?
• Try creating a new skill folder newsletter/ with its own SKILL.md. Add a rule that every newsletter must end with a "Top 3 Links" section. Run the agent and see if it follows the new pattern!

Output

blogs/
└── prompt-engineering/
    ├── post.md       # Blog content
    └── hero.png      # Generated cover image

linkedin/
└── ai-agents/
    ├── post.md       # Post content
    └── image.png     # Generated image

research/
└── prompt-engineering.md   # Research notes

Customizing

Change the voice: Edit AGENTS.md to modify brand tone and style.

Add a content type: Create skills/<name>/SKILL.md with YAML frontmatter:

---
name: newsletter
description: Use this skill when writing email newsletters
---
# Newsletter Skill
...

Add a subagent: Add to subagents.yaml:

editor:
  description: Review and improve drafted content
  model: anthropic:claude-haiku-4-5-20251001
  system_prompt: |
    You are an editor. Review the content and suggest improvements...
  tools: []

Add a tool: Define it in content_writer.py with the @tool decorator and add to tools=[].

Security Note

This agent has filesystem access and can read, write, and delete files on your machine. Review generated content before publishing and avoid running in directories with sensitive data.

Requirements

• Python 3.11+
• ANTHROPIC_API_KEY - For the main agent
• GOOGLE_API_KEY - For image generation (uses Gemini's Imagen / "nano banana" via gemini-2.5-flash-image)
• TAVILY_API_KEY - For web search (optional, research still works without it)

---

⬅️ Back to Course Catalog