Practical Examples: Blog Posts, Courses, Product Descriptions
Session 9.7 · ~5 min read
Different Content, Different Chains
The Research-Writer-Editor chain from Session 9.3 is a general-purpose starting point. But a blog post has different requirements than a course session, which has different requirements than a product description. Each content type benefits from a chain customized to its specific demands.
This session provides complete agent chain specifications for four content types. These are not rigid templates. They are starting points you modify for your needs.
Chain 1: Blog Post (Thought Leadership)
Thought leadership content needs a strong opinion backed by evidence. The chain adds an Opinion Agent between Research and Writing.
Agent"] --> B["Opinion
Agent"] B --> C["Writing
Agent"] C --> D["Editing
Agent"] style A fill:#222221,stroke:#c8a882,color:#ede9e3 style B fill:#222221,stroke:#c47a5a,color:#ede9e3 style C fill:#222221,stroke:#6b8f71,color:#ede9e3 style D fill:#222221,stroke:#8a8478,color:#ede9e3
| Agent | System Prompt Focus | Key Difference from Generic Chain |
|---|---|---|
| Research | Find data that supports and contradicts the thesis | Explicitly searches for counterarguments |
| Opinion | Given the research, formulate a specific, defensible position | New agent: bridges research and writing with a stated opinion |
| Writer | Write an argumentative piece with the opinion as the spine | Instructed to take a side, not present "balanced" coverage |
| Editor | Check argument strength, not just artifacts | Additional dimension: "Does the argument hold under scrutiny?" |
The Opinion Agent solves a common problem: AI defaults to both-sides-ism. By separating opinion formation from writing, you ensure the draft has a spine. The opinion is informed by research, stated before writing begins, and maintained throughout.
Chain 2: Course Content (Educational)
Course content needs progressive structure: each session builds on the previous one. The chain adds a Curriculum Agent that maintains continuity across sessions.
Agent"] --> B["Research
Agent"] B --> C["Writing
Agent"] C --> D["Pedagogy
Editor"] style A fill:#222221,stroke:#c8a882,color:#ede9e3 style B fill:#222221,stroke:#6b8f71,color:#ede9e3 style C fill:#222221,stroke:#8a8478,color:#ede9e3 style D fill:#222221,stroke:#c47a5a,color:#ede9e3
| Agent | System Prompt Focus | Key Difference |
|---|---|---|
| Curriculum | Given the module outline and previous session summaries, define what this session must teach and what it can assume | Maintains learning progression across sessions |
| Research | Find examples, case studies, and data relevant to the lesson objective | Optimized for teaching materials, not general research |
| Writer | Write in an instructional voice with worked examples and practical exercises | Voice fingerprint adjusted for teaching tone |
| Pedagogy Editor | Check: Does it teach the stated objective? Is the difficulty appropriate? Are exercises actionable? | Replaces general editor with education-specific reviewer |
Chain 3: Product Descriptions (E-commerce)
Product descriptions need accuracy above all. The chain adds a Compliance Agent that verifies every product claim.
Agent"] --> B["Writing
Agent"] B --> C["Compliance
Agent"] C --> D["SEO
Agent"] style A fill:#222221,stroke:#c8a882,color:#ede9e3 style B fill:#222221,stroke:#6b8f71,color:#ede9e3 style C fill:#222221,stroke:#c47a5a,color:#ede9e3 style D fill:#222221,stroke:#8a8478,color:#ede9e3
| Agent | System Prompt Focus | Key Difference |
|---|---|---|
| Data | Extract product specs, features, and claims from the product data sheet | Replaces research with structured data extraction |
| Writer | Write persuasive but accurate descriptions using only verified specs | Constrained to product data; no claims beyond what is verified |
| Compliance | Verify every claim against the product data sheet; flag any embellishment | Specific to product accuracy, not general quality |
| SEO | Optimize title, meta description, and body for target keywords without altering claims | Final-stage optimization that does not compromise accuracy |
Chain 4: Technical Report
Technical reports need precision and proper citation. The chain adds a Citation Agent.
| Agent | Role | Handoff |
|---|---|---|
| Research Agent | Gather academic and technical sources | JSON brief with full citation data (authors, year, DOI) |
| Writing Agent | Draft with inline citation markers ([1], [2], etc.) | Markdown with numbered reference markers |
| Citation Agent | Verify each citation, format reference list, check for hallucinated sources | Draft with verified reference section |
| Technical Editor | Review for technical accuracy and argument rigor | Annotated draft with domain-specific feedback |
The correct number of agents in a chain is the number of distinct roles your content type requires. More agents means more API cost and more handoff complexity. Fewer agents means each agent is less focused. Find the minimum number of agents that covers all critical functions for your content type.
Further Reading
- Multi-Agent Content Creation System, Sight AI
- Agentic AI for Content Workflows, Global Publicist
- Building Autonomous Systems with Agentic AI, DigitalOcean
- Building a Multi-Agent Report Workflow, Arunkumar Ravichandran (Medium)
Assignment
Choose the content type most relevant to your work. Using the specifications in this session as a starting point, customize a complete agent chain:
- Number of agents and their roles
- Full system prompt for each agent
- Handoff format between each agent (data contract)
- One quality check per handoff
- Error handling and retry strategy
Test the chain end-to-end on a real piece of content. Compare the output to your previous three-agent chain. Does the customized chain produce better results for your specific content type?