Get Started
In this section
Back to reference

14. Content-as-Code Pipelines

aDNA Specification v2.5 Section 14 of 20 Full text

Scan: Folder-based workflows where a file’s directory location IS its processing state — pipeline structure, stage AGENTS.md, pipeline index.

Decisions: D14

14.1 Paradigm

Content-as-code is a universal paradigm for folder-based workflows: a file’s directory location IS its processing state. Moving a file between stage directories advances it through the workflow.

This paradigm applies wherever content flows through defined stages — research ingestion, document review, approval workflows, deployment pipelines.

stateDiagram-v2
    direction LR
    [*] --> inbox: New content
    inbox --> processing: Agent picks up
    processing --> review: Processing complete
    review --> done: Approved
    review --> processing: Revision needed
    done --> [*]

    note right of inbox: AGENTS.md defines<br/>acceptance criteria
    note right of processing: AGENTS.md defines<br/>processing steps
    note right of review: AGENTS.md defines<br/>review checklist

14.2 Pipeline Structure

Pipelines live in how/pipelines/{pipeline_name}/:

how/pipelines/{pipeline_name}/
├── AGENTS.md           # Pipeline overview, stage transitions
├── inbox/              # Stage 1
│   └── AGENTS.md       # Processing instructions for this stage
├── processing/         # Stage 2
│   └── AGENTS.md
├── review/             # Stage 3
│   └── AGENTS.md
└── done/               # Stage 4
    └── AGENTS.md

Each stage folder MUST have an AGENTS.md with processing instructions specific to that stage.

Stage names and count are pipeline-specific. The file’s location is its state — no separate status tracking is needed.

14.3 Pipeline Index

The how/pipelines/ directory SHOULD have an AGENTS.md documenting all pipelines, their purposes, and their stage flows.