Every repository that uses the AI-SDLC has two layers of docs, and every part of its code has a page. A reader, a person or an agent, then finds a true page for each part instead of guessing.
- A technical layer that says how the repository is built. It has one page for each unit: a .NET project, a Bicep building block, or a package.
- A functional layer that says what the repository gives its users, in plain words.
- A gate that keeps it complete.
Check-DocCoverage.ps1fails when a layer is missing or a unit has no page, and names the page to write. - A first draft of each missing page. The
adopt-docsskill collects the facts of the unit from the code into a facts file, and an agent drafts the page from it. You check each claim before it is merged.
- A repository has just joined the AI-SDLC: run
/adopt-docsonce, in its own pull request. - A change adds a project or a building block:
write-docsadds its page in the same change. - A change touches an area that has no page yet, such as deployments. Then
write-docswrites a page for the whole area in both layers. No gate checks this;wrap-updoes. - A page you wrote fails the reading level or another doc gate:
write-docsrewrites that page.
- A draft is a proposal from an agent. Nothing reaches a page until a person has checked it against the code.
- A page that covers a unit under another name says so in its front matter (
unit: <path>). - A unit that nobody uses can be left out, with the reason in the pull request.
- What the AI-SDLC gives a team - where complete docs fit in the way of working.
- core - the gate and the drafting tool live in the core.
- Glossary - what a unit and a layer are.
- Index - every page of this layer.