Manage-Refs Skill

You are routing reference-handling work for a medical manuscript. The user is somewhere in the lifecycle — drafting, building a circulation DOCX, swapping CSL after a journal rejection, fixing a cross-reference defect surfaced by QC, or wiring up live Zotero field codes for a co-author Word workflow. Pick the right tool from the decision table; do not invent a parallel pipeline.

Why This Skill Exists

Reference handling spans every late-stage skill: /write-paper builds the first DOCX, /revise rebuilds it after each reviewer round, /peer-review emits a critique that quotes references back, /sync-submission packages the final tarball, /find-journal informs CSL swaps on rejection cascade, and /verify-refs audits the bibliography. Until 2026-05-01 these scripts lived under skills/write-paper/scripts/, which made /revise and /sync-submission silently depend on a sibling skill — a layering inversion that broke when /write-paper was loaded into a non-research project. Moving the lifecycle tools here turns reference handling into a first-class concern with one decision tree, one set of CSL files, and one provenance file (NOTICE.md) for the vendored Zotero CWYW writer.

Validated 2026-05-01 against a 21-reference meta-analysis manuscript (a meta-analysis project's submission) for both pandoc-citeproc and Zotero-CWYW paths.

Anti-Hallucination Guarantees

1. Citekey discipline (Phase 0): every in-text citation must be

[@bibkey] resolvable in refs.bib. scripts/check_citation_keys.py is a hard gate — UNDEFINED keys exit non-zero and block the build.

`[@NEW:topic]` placeholder convention: while drafting, /write-paper may emit [@NEW:topic_slug] markers for citations the author still needs to source. check_citation_keys.py classifies these as NEW_PLACEHOLDER (not UNDEFINED) and exits 0 — the build is allowed to proceed during drafting. Phase 7.6 (DOCX render) is a hard gate: zero NEW_PLACEHOLDER entries must remain. Resolve each by adding the citation to Zotero (then /lit-sync refreshes refs.bib) and replacing the placeholder with the real [@bibkey]. Never let a [@NEW:...] reach a rendered DOCX.

1. No hand-typed References list — references are always rendered by

pandoc citeproc + journal CSL or by the Zotero Word plugin (CWYW). See ~/.claude/rules/manuscript-references.md.

1. Zotero metadata is never invented — inject_zotero_cwyw.py fetches

item data live from http://localhost:23119. Any HTTP failure aborts with a non-zero exit so partial bibliographies never reach the user.

1. Marker conversion is mapping-driven — md_marker_convert.py will

never guess a Zotero key for a number; unmapped markers stay as [N] and are reported on stderr.

1. Cross-reference QC is a submission gate — scripts/check_xref.py

--strict exits 1 on any MISSING_DOCX / MISSING_BODY / MISMATCH, blocking pipelines that try to ship a DOCX whose Table/Figure citations don't match captions.

1. Audit boundary: this skill writes; bibliographic correctness against

PubMed/CrossRef stays in /verify-refs. Always invoke /verify-refs after a render before signing off — one read-only audit, one writer.

Decision Tree

Workflows

A. Pandoc citeproc (default for solo authors and final submissions)

User provides manuscript.md with [@bibkey] citations + refs.bib.

1. Gate: python "${CLAUDE_SKILL_DIR}/scripts/check_citation_keys.py" manuscript.md refs.bib

— exits non-zero on UNDEFINED keys. Fix and re-run.

1. Render:

   "${CLAUDE_SKILL_DIR}/scripts/render_pandoc.sh" \
     -j european-radiology \
     -i manuscript.md \
     -b refs.bib \
     -o manuscript_final.docx

Bundled CSLs (in citation_styles/): european-radiology, radiology, american-journal-of-roentgenology, cardiovascular-and-interventional-radiology, korean-journal-of-radiology, vancouver, vancouver-superscript, springer-basic-brackets, springer-vancouver-brackets. Use radiology for RYAI; use vancouver for JVIR (no dedicated CSL).

1. QC:

   python3 "${CLAUDE_SKILL_DIR}/scripts/check_xref.py" \
     --md manuscript.md --docx manuscript_final.docx \
     --out qc/xref_audit.json --strict

Treat submission_safe: false as a halt. Route fixes by symptom — see the table in references/check_xref_symptoms.md.

1. Audit hand-off: invoke /verify-refs for the PubMed/CrossRef audit

before sign-off.

B. Zotero CWYW (co-author Word workflow)

User has a markdown SSOT and wants reviewers to edit citations directly in Word. Each reference must already exist as a Zotero item; the user supplies a [N] → ZoteroKey mapping.

1. Convert markers:

   python3 "${CLAUDE_SKILL_DIR}/scripts/md_marker_convert.py" \
     --input manuscript.md --output manuscript_keys.md \
     --map ref_map.json --to-keys

Optionally stage with --active-ns 1,2,3,4,19 for a sample build first (validated on an active meta-analysis project: 5-ref sample reduces Word Refresh blast radius when debugging).

1. Render to .docx with pandoc (workflow A) so the body has plain text

[@key] markers, OR pre-build a .docx some other way that still contains plain [@key] text.

1. Inject CWYW:

   python3 "${CLAUDE_SKILL_DIR}/scripts/inject_zotero_cwyw.py" \
     --input manuscript_keys.docx --output manuscript_cwyw.docx \
     --user-id 16613550 --keys-from keys.txt

The script fetches Zotero metadata via the local connector (port 23119); any HTTP failure aborts with non-zero exit.

1. First-build instruction (REQUIRED — see Known Limitation #1): open

the output in Word → Zotero tab → Add/Edit Bibliography once. After that, Refresh keeps citations and bibliography in sync as authors edit.

1. Surgical patches are unsafe: for ref additions in later rounds, edit

the markdown SSOT and rebuild the whole .docx instead of regex-patching the post-CWYW file. Zotero's rendered [N] superscripts can collide with plain [N] markers and corrupt the field codes.

C. Cascade rejection re-render (find-journal hand-off)

User got rejected from journal A and /find-journal recommended journal B.

1. Confirm the new CSL exists in citation_styles/ (or fetch from

https://citationstyles.org/styles and drop in).

1. Re-run render_pandoc.sh -j <new-csl> against the same manuscript.md +

refs.bib.

1. Re-run check_xref.py --strict.
2. Re-run /verify-refs if any new references were added during the

inter-journal revision.

D. Cross-reference QC only

User shipped a manuscript and a reviewer flagged a Table/Figure mismatch.

1. Run check_xref.py --strict on the current manuscript.md + .docx.
2. Inspect qc/xref_audit.json. Body caption is the SSOT — fix manuscript.md

and rebuild, never patch the .docx by hand.

1. See references/check_xref_symptoms.md for the

MISSING_BODY / MISSING_DOCX / MISMATCH triage table.

1. For journals that accept figures and tables as separate attachment files

(the default in European Radiology, Radiology, AJR, JVIR, KJR, and most medical journals), pass --allow-separate-attachments. MISSING_DOCX rows are then recorded as WARN rather than FAIL; MISSING_BODY and MISMATCH remain P0 because they indicate SSOT drift, not attachment policy.

D'. v_(N+1) docx regeneration check (build-time companion)

When building v_(N+1) from a frozen v_N, the v_(N+1) docx MUST differ from v_N docx by content — a byte-identical copy is a silent seed-copy that will revert markdown edits at peer review. check_xref.py carries two flags for the build-time companion to the submission-time gate in scripts/verify_package_integrity.py --assert-vN-docx-changed:

python3 "${CLAUDE_SKILL_DIR}/scripts/check_xref.py" \
    --md manuscript_v2.md \
    --docx manuscript_v2.docx \
    --vN-docx-md5 manuscript_v1.docx \
    --vN-md manuscript_v1.md \
    --strict

• --vN-docx-md5 alone: MD5 identity check. Identical bytes = FAIL.
• --vN-docx-md5 + --vN-md: additionally extracts the markdown-only diff

between v_N and v_(N+1) and verifies each ≥40-char diff line appears verbatim (whitespace-normalized, case-insensitive) in the new docx body XML. Missing diff lines = body did not pick up the markdown edits.

Output records the result under vN_docx_check in qc/xref_audit.json. Either failure mode causes a non-zero exit even without --strict.

E. Master pre-submission gate (recommended end-to-end chain)

The single entry point that combines workflows A and D plus /verify-refs into one aborting chain. Use this immediately before submission or before circulating a v_N package to senior co-authors.

bash "${CLAUDE_SKILL_DIR}/scripts/pre_submission_gate.sh" \
    --md manuscript/manuscript.md \
    --bib manuscript/_src/refs.bib \
    --docx submission/<journal>/manuscript.docx \
    --allow-separate-attachments    # omit if the journal accepts inline figures/tables

Stage order (first failure aborts):

1. check_citation_keys.py manuscript.md refs.bib — UNDEFINED / UNUSED keys
2. verify_refs.py refs.bib --strict — PubMed / CrossRef per-entry verification
3. render_pandoc.sh -j <csl> -i ... -b ... -o ... — invoked only when --docx is omitted
4. check_xref.py --md ... --docx ... --strict [--allow-separate-attachments]

On success the chain writes qc/pre_submission_gate.json (plus the per-stage artifacts qc/reference_audit.json and qc/xref_audit.json) with submission_safe: true. On any failure the JSON records the failing stage and exit code, and the script exits non-zero — do not submit until the failing stage passes.

Critical: the gate does not reimplement any check. It calls the existing scripts as subprocesses. If you find yourself wanting to add a check, add it to the underlying script (the gate then picks it up automatically).

F. BibTeX author-format corruption (rendered-name check)

Entries written as author = {Surname AB and Surname2 CD} (family + initials, no comma) make BibTeX treat the last token as the family name, rendering "AB S, CD S2". Always store author = {Family, Full Given}. Concatenated initials even with a comma (Family, AB) still collapse to a single initial under CSL initialize-with, so use the full forename from PubMed efetch.

/verify-refs compares bib content against PubMed but does not see the rendered output; grep the rendered docx and the bib separately:

unzip -p out.docx word/document.xml | sed 's/<[^>]*>//g' | grep -oE "[A-Z]{2} [A-Z], [A-Z]{2} [A-Z]"   # corruption signature in output
grep -nE 'author\s*=\s*\{[A-Z][a-z]+ [A-Z]{1,3}( |\})' refs.bib                                          # no-comma source entries

Quality Gates

This skill defines three submission gates and one user approval gate:

• Gate 1 (citekey integrity): check_citation_keys.py exits non-zero on

UNDEFINED keys. The pipeline halts; the user reviews and fixes.

• Gate 2 (cross-reference integrity): check_xref.py --strict exits 1 on

any MISSING_DOCX / MISSING_BODY / MISMATCH row. The user reviews qc/xref_audit.json and resolves before proceeding.

• Gate 3 (audit hand-off): before sign-off, the user must run

/verify-refs and confirm submission_safe: true in qc/reference_audit.json. This skill never marks the bibliography audited on its own.

• User approval gate (CWYW first build): the user must perform Word →

Zotero → Add/Edit Bibliography manually after the first inject_zotero_cwyw.py build. The skill cannot automate this and warns on stderr that it is required.

Provenance

scripts/_vendor_citation_writer.py is vendored from alisoroushmd/zotero-mcp @ ed5dfb71, MIT licensed. See NOTICE.md and LICENSE.zotero-mcp.

Related

• ~/.claude/rules/manuscript-references.md — global rule (decision tree

this skill implements)

• ~/.claude/rules/agent-skill-routing.md — skill router (this skill is the

reference-handling row)

• ~/.claude/rules/zotero-workflow.md — BBT auto-export, MCP setup
• /verify-refs — read-only audit (PubMed / CrossRef + first-author

cross-check)

• /lit-sync — Zotero ↔ Obsidian sync, refs.bib provider
• /write-paper Phase 7.6 — calls this skill (one-line delegation)
• /revise, /sync-submission, /find-journal — call this skill on

rebuild / re-render / cascade

Known Limitations

1. First-build empty BIBL field (CWYW): inject_zotero_cwyw.py writes a

stub ADDIN ZOTERO_BIBL field; Word's Zotero Refresh treats an empty stub as user-customized and refuses to populate it. User must run Add/Edit Bibliography once. Subsequent Refresh works as expected. Validated on Word for Mac, an active meta-analysis project.

1. Webpage / non-journal item types: handled by the patched

zotero_to_csl_json that fetches Zotero's native CSL-JSON; do not bypass this patch.

1. Surgical post-build regex patches are unsafe — see Workflow B step 5.
2. Local Zotero required for CWYW — port 23119 must be reachable; no

web-API fallback yet (would need ZOTERO_API_KEY). On failure the script aborts with non-zero exit so partial builds never ship.