docs: stop describing sds as a git submodule

[d-morrison]

Requested by Ezra

Closes #18

Before: index.qmd, README.Rmd/README.md, CLAUDE.md, and a lychee.toml comment described sds as something course sites include as a git submodule, and the README gave a git submodule add command.

After: they say course sites link to these pages by URL. They still describe the include-by-path mechanism and the sds symlink, for a host site that keeps a copy at its root. The git submodule add block is gone. The latex-macros submodule instructions are untouched, since that submodule is still real.

How: I kept the mechanism rather than removing it, because the symlink and its lintr/lychee handling are documented and in use by this repo's own includes. README.md was edited by hand to match README.Rmd (same wording, re-wrapped as the knit output is); R was not run. Prose and comment edits only.

🤖 Generated with Claude Code

[d-morrison]

Claude Code CLI (local session) is working on this --- please hold off on pushing to this branch until I'm done.

Session worktree: /private/tmp/claude-503/-Users-ezramorrison-Documents-GitHub/33f186b9-e966-4796-b7c8-e023d990211c/scratchpad/sds

Posted by Claude Code (AI agent) --- not written by a human.

[dem-extra1]

I am working on this PR in session 83c8beb5-2277-4840-a8a1-3dda6fb9076d.

[github-actions]

Claude finished review — View run

---

I have what I need. The PR's net diff against main is unchanged from the last reviewed commit, which was already found clean.

Review: PR #34 — "docs: stop describing sds as a git submodule"

What changed since the last review round: The workflow's computed range (bbc8f94f..bd1a38d4) looks large (292 files, ~7,878 insertions) because the branch picked up two merge commits (2f67551, bd1a38d) syncing in a large volume of unrelated work already merged to main (theme picker, model-validation chapter, MLE exercises, etc.). I verified this is merge noise, not new PR content, by diffing the five PR-relevant files between the last-reviewed commit and current HEAD directly (git diff bbc8f94f bd1a38d4 -- CLAUDE.md README.Rmd README.md index.qmd lychee.toml): the only differences are lines that came from main (e.g., new _macros-sds.qmd and theme-dropdown bullets in CLAUDE.md) — none of them touch the submodule-wording sentences this PR actually changes.

I then read the saved PR diff (.claude-review-pr.diff, 95 lines) and confirmed it is byte-for-byte the same content change already reviewed and approved in round 2 (commit bbc8f94f): the same five files (CLAUDE.md, README.Rmd, README.md, index.qmd, lychee.toml), the same "course sites link to its pages by URL" rewording, and the same preserved sds -> . symlink/!sds/ exclusion/lychee.toml handling. I re-checked for stray submodule references tied to sds across the whole tree (including the newly-merged AGENTS.md and .github/copilot-instructions.md) and found none — the only remaining submodule hits are the genuinely separate latex-macros submodule, which this PR correctly leaves untouched.

One thing worth flagging transparently: the working tree currently carries an uncommitted change to CLAUDE.md that reverts exactly this PR's own one-line wording change back to the pre-PR text (matching how a host site addresses this repo as a submodule). This was already present at the start of this session (visible in the initial git status), not something introduced by my review, and it does not affect the committed PR diff — git diff bbc8f94f bd1a38d4 (the actual commits) confirms the PR's own commit still carries the new wording. This looks like a review-harness artifact (likely the harness pinning CLAUDE.md instructions to main's copy rather than the PR's), not a defect in the PR itself, so I'm not treating it as a finding — just noting it so it isn't mistaken for a regression.

No new issues. The prior round's verdict — Ready for merge, posted at #34 (comment) — still stands: the one blocking finding from round 1 (hand-edited, un-rendered README.md) was resolved in bbc8f94f and remains resolved here, and no new content has been introduced by the two merge commits since.

Verdict

Ready for merge — no new findings; the prior round's verdict (clean, #34 (comment)) still stands, and the merge commits since then carried no new changes to this PR's actual diff.

Structured Review Data (JSON)

{
  "schema_version": "1.1",
  "reviewer": "claude",
  "commit_sha": "bd1a38d41e301a33653158cc0eef69700c47b012",
  "verdict": "CLEAN",
  "findings": [],
  "detailed_assessment": "The five-file diff in CLAUDE.md, README.Rmd, README.md, index.qmd, and lychee.toml is byte-identical to the already-approved bbc8f94f content; the two merge commits since then only synced in unrelated main-branch work.",
  "holistic_assessment": "Scope stays limited to submodule-wording prose across the originally touched files, with no regression to the symlink include mechanism, build hygiene, or CI checks from the merge-in-main commits."
}

Reviewed commit: 2f67551

[github-actions]

💰 Cost: $0.4768 (review) — run

[dem-extra1]

Done --- PR is free. Merging under MWC grant.

[github-actions]

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-10-06 02:02 PDT