Fix Stale Submodule Refs in Specs

Four specification passages call be-best a git submodule. It is not.

Fix Stale Submodule Refs in Specs

Commit 416d145 moved the theme out of a submodule and into the repository root. The published specifications still describe the old arrangement, so they tell a reader to commit and push inside be-best and to record a submodule pointer in this repository - neither is possible.

The facts the documents should state: there is no .gitmodules file, the only remote is ivanhawkes/ivanhawkes.github.io, the theme’s 139 files are tracked directly by this repository, and Hugo selects the theme through theme: be-best with themesDir: '.' in config/_default/hugo.yaml.

• content/specification/technical/index.md, “Presentation Layer”: reword the theme paragraph so be-best is a directory of this repository, and drop the submodule-pointer commit rule.
• content/specification/requirements/index.md: the Maintainability row says “the be-best submodule” and the History and recoverability row says “the theme is a submodule with its own history”. Reword both.
• content/specification/decisions/index.md: the decision “The theme is a git submodule published on its own features branch” is now falsified. Record that the theme moved to the repository root, with the reason, so the decision history stays honest rather than just deleting the row.
• Search the rest of content/ for further submodule mentions and fold them into the same pass. Two rows of the requirements and technical specifications said the Hugo cache is keyed on “the theme commit”; the workflow hashes the be-best tree, so both now name the tree hash.

The agent instruction files (AGENTS.md, .pi/APPEND_SYSTEM.md, .pi/skills/dcp, .pi/skills/theme-dev, .pi/prompts/review.md) were already corrected on 2026-10-24; this card covers the published content only.

Acceptance Criteria

• No published page describes be-best as a git submodule.
• The technical specification’s Presentation Layer names the theme’s location as be-best/ in this repository and cites the theme / themesDir settings that select it.
• The requirements specification’s Maintainability and History and recoverability rows contain no submodule wording.
• The decision record shows the submodule decision superseded by the move to the repository root, dated, with the reason recorded.
• grep -rin submodule content/ returns only the superseded decision record, and this card, whose own path and title name the defect.
• pnpm run test passes.

The card’s own claim that “there is no .gitmodules file” is close but not exact: commit 416d145 emptied the file and left it tracked, empty. The repository declares no submodules; git ls-files -s lists no gitlink, and git submodule status is empty. Deleting the leftover file is a repository cleanup, outside this card’s published-content scope.