Anatomy of a Deliverable Pipeline

A deliverable pipeline is a sequence of stages where each stage takes a defined input, applies a transformation, and produces a defined output that feeds the next stage. The simplest pipeline has three stages: research, draft, review. The research stage takes a brief and produces structured findings. The draft stage takes structured findings and produces a formatted document. The review stage takes the formatted document and produces a QA report with corrections.

The critical design principle is stage isolation. Each stage must be independently executable and independently testable. If the draft stage fails, you should be able to re-run it from the research output without re-running the research. If the review stage catches errors, you should be able to send corrections back to the draft stage without touching the research. Stage isolation is what makes pipelines debuggable, restartable, and composable.

1. 1. Define Stage Boundaries Each stage has one input artifact and one output artifact. The output of stage N is the input of stage N+1. If a stage requires two inputs, it is either two stages or it needs a merge step before it.
2. 2. Specify Artifacts Name and format every inter-stage artifact. "Research output" is vague. "Structured findings in extraction template format with source citations" is specific. Ambiguous artifacts produce ambiguous handoffs.
3. 3. Install Quality Gates Every stage transition has a gate — a set of criteria the output must meet before the next stage begins. Gate failures loop back to the current stage, not forward to the next one.