Add CI check for duplicate crossref ids

[d-morrison]

Requested by Ezra · project thread

Before: nothing stopped a second definition of an existing crossref id (win#79).
After: .github/workflows/check-duplicate-crossref-ids.yml runs .github/scripts/check-duplicate-crossref-ids.py, which fails when a crossref id is defined twice across index.qmd and chapters/**/*.qmd.
It reads attribute-block ids and #| label: ids in executable cells, and ignores other fences, inline code and HTML comments.
No existing check in Morrison-Lab/gha covers this.
How: the script fails on current main (it reports #def-ip-weights) and passes on the branch of the chapter 12 fix, so this check will be red until that PR merges.

Not closing an issue; this is a guard for #79.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Gh7L9vJhwrpvpBhzYMdjPK

---

Generated by Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[d-morrison]

OpenCode adversarial review (deepseek-v4-pro, then gpt-5.6-luna) found parser gaps (non-cell fences, inline code and multi-backtick spans, multi-line comments, missing prf/sol prefixes, annotations without file and line).
All were fixed in 5d49e07.
The final re-review by luna of the pre-fix diff gave only the two gaps now fixed; kimi-k3 and deepseek-v4-pro stalled on the last pass, so there is no final "VERDICT: clean" line.
Remaining deferred: prefixes are hardcoded rather than read from _quarto.yml.
The check job is red on this PR by design until the chapter 12 rename (#114) merges; on that branch the script passes.

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

---

Generated by Claude Code

[d-morrison]

/review

---

Generated by Claude Code

[d-morrison]

/review

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

---

Generated by Claude Code

[github-actions]

👀 Claude Review Dispatched — run reviewing PR #115 at 5d49e07.

[github-actions]

Claude finished review — View run

---

This PR's head commit is still 5d49e0783479cd6a29b74c902d57facef0d2995d — identical to the commit reviewed in the prior round (no new commits have landed; the only new PR activity is another /review trigger comment). I re-ran check-duplicate-crossref-ids.py against the current tree and confirmed it still reports exactly the same expected pre-existing duplicate (#def-ip-weights in chapters/02-randomized-experiments.qmd:433 and chapters/12-ip-weighting-marginal-structural-models.qmd:149), red-by-design pending PR #114, consistent with the prior review's findings.

Review: PR #115 — Add CI check for duplicate crossref ids

No code has changed since the previous review round. My prior review (posted 2026-10-09T10:02:35Z, commit 5d49e0783479cd6a29b74c902d57facef0d2995d) verified the parser's fence-nesting, inline-code-stripping, HTML-comment, and multi-backtick handling via synthetic fixtures and the real chapters/ tree, found no bugs, security issues, or lab-manual/CLAUDE.md violations, and confirmed the workflow YAML (permissions, timeout, triggers) matches sibling conventions. That verdict still stands — this confirming pass re-ran the script and reproduced the identical expected output, so there is nothing new to add.

The three observations from that review remain non-blocking optional follow-ups: no automated tests for the new script (consistent with the existing convention for this whole family of .github/scripts/ checks), the local-vs-Morrison-Lab/gha placement question (already resolved as acceptable given the local-script convention), and the hardcoded PREFIXES list vs. reading _quarto.yml (already disclosed by the author as a deferred trade-off, and still matches current _quarto.yml reality).

No d-morrison/macros math changes are introduced by this diff, so the macro-opportunity check does not apply, consistent with the prior round.

Verdict

Ready for merge — no new findings; the prior round's "Ready for merge" verdict stands unchanged since no commits have landed since that review.

Structured Review Data (JSON)

{
  "schema_version": "1.1",
  "reviewer": "claude",
  "commit_sha": "5d49e0783479cd6a29b74c902d57facef0d2995d",
  "verdict": "CLEAN",
  "findings": [],
  "detailed_assessment": "Re-ran check-duplicate-crossref-ids.py against the unchanged chapters/ tree and reproduced the identical pre-existing def-ip-weights duplicate reported in the prior review round.",
  "holistic_assessment": "No new commits landed since the last review, so scope, integration, and validation posture against issue 79 and PR 114 remain exactly as previously assessed."
}

Reviewed commit: 5d49e07

[github-actions]

💰 Cost: $0.1825 (review) — run