Skip to main content

Documentation Bootstrap Loop

warning

Full bootstrap is exceptional. The repository provides a .codex/skills/docs-loop-bootstrap entry, but does not currently track the historical 21-story PRD/prompt described by older versions of this page. Use a targeted documentation update for bounded maintenance.

Purpose​

A bootstrap loop creates or reconstructs a broad documentation surface from source. It is appropriate only when the site or a major domain lacks a usable baseline. It should not overwrite a functioning site merely because individual pages are stale.

Required preparation​

  1. Read the repo-local docs-loop-bootstrap skill completely and verify its linked templates/scripts exist.
  2. Define included audiences, domains, source roots, prohibited content, navigation ownership, generated contracts, and acceptance checks.
  3. Inventory existing pages and preserve useful history, redirects, and intentional hand-authored guidance.
  4. Establish source precedence: implementation/configuration, tests, freshly checked generated contracts, accepted/current architecture, then prose.
  5. Run in an isolated worktree with explicit commit and cleanup policy.

Expected phases​

  • discovery and page/source mapping;
  • extraction of verifiable facts;
  • information architecture and audience filtering;
  • page writing with source/path checks;
  • contract, route, symbol, and link validation;
  • affected Docusaurus variant builds;
  • final gap/evidence report.

Do not create placeholder pages just to satisfy navigation. Mark proposals and historical evidence explicitly, and do not publish internal secrets or tenant details to external variants.

Tracked loop state​

The only current task template committed under .agents/tasks/templates is the documentation-update template. A full bootstrap run must generate/review its own artifacts through the applicable skills; do not invent missing paths from old examples.

See Documentation update loop and Docs development.