[d-morrison]

Requested by Ezra · project thread

Closes #805.

Why

Ezra asked (2026-11-12) whether mds should show a cross-repo tree of how mds builds on mcs. mds's graph already records 40 mcs definitions as prerequisites (391 edges), but the Dependency Tree page shows only mds's own divs. Morrison-Lab/mcs#166 (merged) made mcs publish a graph of its own, so the two can now be joined.

What changed

dependency-tree.qmd gets a second section with a tree that joins the published mcs and mds graphs. It uses the same #dependency-graph-all widget as the lds all-notes tree, which mds's dependency-graph.js already supports. No script or builder changes.

lychee.toml gets one exact-match exclusion, for the data-json attribute. lychee reads that comma-separated list as one URL and reports a 404, although both files it names return 200. The exclusion matches only that exact string, so a real broken link to either file still fails the check. It's here because this PR fails the link check without it.

Verification

• quarto render dependency-tree.qmd --to html succeeds.
• I ran the script on the rendered page under jsdom (no browser), feeding it mcs's local build of dependency-graph.json. The joined tree drew 1,552 boxes: 1,389 from mds, 162 from mcs, and 1 sds node that mds links to. Each of the 162 mcs boxes links to its mcs page, for example https://morrison-lab.github.io/mcs/expressions-algorithms.html#def-algorithm. The existing site tree is unchanged at 1,389 boxes.
• The mcs site is live (checked 2026-11-12, 16:00 PST): mcs/dependency-tree.html, mcs/dependency-graph.json and mds/dependency-graph.json each return 200.

---

Generated by Claude Code

[d-morrison]

The review's not-clean verdict (and so require-clean-verdict) is about one thing: the new links point at the mcs site, and mcs hasn't finished publishing yet. Morrison-Lab/mcs#166 merged at 23:45 UTC, and its Quarto Publish run is still in progress. That run deploys mcs/dependency-tree.html and mcs/dependency-graph.json.

I'm holding the next push (it also takes the review's suggestion to drop the repeated hover and click instructions) until both URLs return 200. That push will re-run the review and the link check.

---

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

---

Generated by Claude Code

[d-morrison]

Cross-harness review (agy Flash) on e3d7132: VERDICT: CLEAN

The new section links valid mcs anchors, the data-json div mirrors the existing graph div, and the anchored lychee exclude regex is valid and matches only the comma-joined value. Flash asked whether the new # heading should nest lower; I checked dependency-tree.qmd on main, whose existing section is also # How to read the tree {.unnumbered}, so the new heading matches. Reviewed from the diff; I did not render the page.

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

[github-actions]

Claude finished review — View run

---

I fetched PR 806 live just now. It is open, not a draft, and not merged (mergedAt is null). Its head is a86d2a4af3c79df1e4d0bd5b6bb41318d6050703, the commit I reviewed, so the "Ready for merge" verdict applies to the current head.

The prior clean review of e3d7132 is at #806 (comment). It still stands. The PR's own diff is unchanged since then, and the only commit added is a merge of main that matches git's automatic merge.

Verdict

Ready for merge

Structured Review Data (JSON)

{
  "schema_version": "1.1",
  "reviewer": "claude",
  "commit_sha": "a86d2a4af3c79df1e4d0bd5b6bb41318d6050703",
  "verdict": "CLEAN",
  "findings": [],
  "detailed_assessment": "The dependency-tree.qmd section and the anchored lychee.toml exclude are unchanged; the mcs dependency-graph.json and pages all return HTTP 200 now.",
  "holistic_assessment": "The only change since the last clean review is a main merge matching git's automatic result, so scope, link-check and integration concerns are unchanged and resolved."
}

Reviewed commit: a86d2a4

[github-actions]

💰 Cost: $0.2358 (review) — run

[dem-extra1]

Non-Claude reads of this PR's diff at head a86d2a4 (OpenCode Go endpoint; the paid models are at their weekly limit, so the free ids):

• iceman-contributor-free: CLEAN.
• beam-free: FINDINGS. It says the link to https://morrison-lab.github.io/mcs/proof-writing.html#def-theorem uses a definition anchor for a theorem and is likely broken. Rebutted: the link is to the definition of the word "theorem" in mcs, so a #def- anchor is correct, and the published page has id="def-theorem" (checked with curl just now).

I also merged main into this branch (server-side update, no conflicts), because the merge gate reported it 120 commits behind with changed workflows. That also fixes the preview's mds-only tree, which counted every umbrella-page div twice before #748 (#812).

---

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

---

Generated by Claude Code

[github-actions]

PR Preview Action v1.8.2
🚀 Preview available at: https://Morrison-Lab.github.io/mds/pr-preview/pr-806/

_{Built to branch gh-pages at 2026-11-13 06:09 UTC.}

[d-morrison]

Annotation dispositions for head a86d2a4:

I batch-read the annotations of all 50 passing runs on this head (53 annotations in 4 groups). There are no warning or failure annotations; all 53 are notices:

• ubuntu-latest migration notice, on 50 runs: a GitHub-wide runner-image notice, not tied to this diff.
• SUBMODULES_TOKEN is empty; cloning submodules anonymously, 1 run: routine for a repo whose submodules are public; the checkout succeeded.
• Skipping element-level diff highlighting: ... exceed the 500-element cap, 2 runs (the two pages with more than 500 elements): the preview diff tool's size cap, not a check result.

Nothing here needs a fix.

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

[d-morrison]

Cross-harness review (agy Flash), carried to a86d2a4: VERDICT: CLEAN

The CLEAN verdict on e3d7132 (review) carries to this head: gh pr diff for both heads is identical apart from the index lines (0 differing lines), so the only change is the merge of main.

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

[dem-extra1]

Annotation dispositions for head a86d2a4

All 53 annotations on this head's 50 check runs are notices; none is a warning or failure.

• 50 × "The ubuntu-latest label will migrate to Ubuntu 26": runner-image notice, not about this PR.
• 1 × "SUBMODULES_TOKEN is empty; cloning submodules anonymously": the only submodule (latex-macros) is public, and checkout succeeded.
• 2 × "Skipping element-level diff highlighting: ... exceed the 500-element cap": the preview's change highlighter skipped two large pages; it affects only the preview's highlighting, not the build.

---

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

---

Generated by Claude Code

[dem-extra1]

Post-merge non-Claude re-read of this PR's diff (the same diff as at head a86d2a4), sent to the OpenCode Go endpoint's free ids at 2026-11-12 22:17 PST:

• beam-free: CLEAN ("the added section and join-tree markup look consistent ... the lychee exclude regex matches the comma-separated data-json value as intended").
• iceman-contributor-free: CLEAN.

This PR was merged after beam-free's earlier FINDINGS read had been rebutted, without a clean re-read first. That broke the lab rule that a rebutted FINDINGS read is not a clean read. The re-read above is clean, so nothing needs reverting.

---

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

---

Generated by Claude Code