Pipeline Pattern

The pipeline pattern is a sequential chain where each agent's output becomes the next agent's input. Research feeds into drafting. Drafting feeds into review. Review feeds into formatting. Each step transforms the data, and the final output is the product of every stage in the chain. This is the most natural pattern for content creation, report generation, and any workflow with a logical order of operations.

1. 1. Research The first agent gathers raw information — web searches, document analysis, data collection. Its output is structured data: facts, sources, and key findings in a format the next agent can consume.
2. 2. Draft The second agent receives the research output and writes the first draft. It transforms raw data into narrative, applying the appropriate tone, structure, and format specified in its role definition.
3. 3. Review The third agent evaluates the draft against quality criteria. It returns a score and specific feedback. If the draft passes, it moves forward. If not, it goes back to the drafter for revision.
4. 4. Publish The final agent handles formatting, delivery, or deployment. It takes the approved content and puts it where it needs to go — a file, an API, a message channel, a database.

The parent agent's role in a pipeline is critical. It does not just pass outputs blindly from one stage to the next. Between each stage, the parent reviews the output, extracts the relevant information, and formats it as input for the next agent. This intermediate processing step is where the parent adds value — it ensures each agent receives exactly the context it needs, nothing more and nothing less.