How this works
Agile Systems Portal is not a static document. It is a managed content repository — produced by a small editorial team, governed by explicit rules, and kept alive through a structured pipeline. This page explains how content is created, reviewed, and published.
The editorial team
Three roles produce and govern content.
The Product Owner sets direction. He decides what enters the pipeline, approves analysis briefs before drafting begins, and reviews content before it is published. No content reaches the site without his approval.
The Content Analyst and Drafter (an AI assistant operating under a defined editorial role) reads raw source material, produces structured analysis briefs, and drafts content within the rules of the system. It does not draft freely — it drafts against content type schemas, design rules, and a controlled vocabulary.
The Repository Engineer (a second AI assistant) handles file operations, frontmatter updates, schema validation, and pipeline scripts. It does not make editorial decisions. It executes them.
The human is accountable for direction and quality. The AI roles accelerate production within that accountability structure.
The content pipeline
Every piece of content moves through five stages. Status is tracked in the file’s metadata — it is the single source of truth for where something is.
Funnel — Raw material lands here. Workshop outputs, methodology guides, reference documents, notes. Nothing is written yet. The funnel is an intake area, not a drafting space.
Analyzing — The source material is read and a structured analysis brief is produced. The brief defines what the content is about, what type it belongs to, what level of the organization it addresses, and what it relates to in the existing repository. Drafting does not begin until the brief is approved.
Drafting — Content is produced against the Content Design Rules, the content type schema, and the controlled vocabulary. The rules govern the output — not the drafter’s judgment.
Review — Draft content is reviewed for quality, consistency with existing content, and adherence to the rules. Rejection returns content to Analyzing — not Drafting.
Published — Content is approved and merged to the main branch. The site renders automatically.
The rules layer
Three governing artifacts define what ASP content is and is not allowed to be.
Content Rules define what is in scope: practitioner-level content, neutral in tone, grounded in evidence, free of vendor language and proprietary framework content. English only, no exceptions.
Design Rules define how content is structured: what each content type looks like, what sections are required, what patterns are forbidden, how cross-references are made.
Controlled vocabulary defines the terms that are canonical in this repository and how they are used. Terminology drift between pages is a content defect.
These rules are what make AI-assisted drafting viable at scale. A language model without rules produces fluent, inconsistent content. A language model inside a rule system produces content that is structurally coherent and terminologically stable.
The repository and deployment
Content files are Markdown with structured YAML frontmatter. The schema carries type, level, area, status, version, and a list of related content. The repository is the content — not a backup of it.
The site currently renders through Nextra. The same content structure will feed a Confluence package and other front-end systems without restructuring. The rendering layer can change. The content schema is designed not to.
Content backlog items are tracked as GitHub Issues — lightweight by design, and straightforward to integrate with Jira for teams that need it.
This content is never finished
ASP content is a living body of knowledge. Any published page can be re-entered into the pipeline when new source material arrives, when a related page changes its framing, or when a review identifies drift. Published means current and approved — not final.
Designed for a small team — built to scale
The content operating model is intentionally lightweight. A product owner, two AI-assisted roles, and a git repository are enough to run the full pipeline. The same model — funnel intake, analysis briefs, governed drafting, explicit review gates, metadata-tracked status — works for a larger editorial team or a larger organization. The roles scale. The rules stay the same.