• Communicate with the user in their preferred language.
• Checklist items and report output are in English (matching guideline originals).
• Medical terminology is always in English.

• Checklists (bundled, open license):

  ${CLAUDE_SKILL_DIR}/references/checklists/

  • STROBE.md

    -- observational studies (CC BY)

  • STROBE_MR.md

    -- Mendelian randomization studies, STROBE-MR 2021 (base STROBE + MR extension; CC BY, Davey Smith et al. BMJ 2021)

  • STARD.md

    -- diagnostic accuracy studies (CC BY 4.0)

  • STARD_AI.md

    -- AI diagnostic accuracy studies (CC BY, Sounderajah et al. Nat Med 2025)

  • TRIPOD.md

    -- prediction models, classic 2015 version (no open licence — © ACP; Moons et al. Ann Intern Med 2015)

  • TRIPOD_AI.md

    -- prediction models with AI/ML (CC BY 4.0, Collins et al. BMJ 2024)

  • TRIPOD_LLM.md

    -- studies using large language models, TRIPOD-LLM 2025 (educational summary, Gallifant et al. Nat Med 2025)

  • PGS_RS.md

    -- polygenic (risk) score prediction studies, PGS-RS / PRS-RS 2021 (educational summary, Wand et al. Nature 2021)

  • CHEERS_2022.md

    -- health economic evaluations (cost-effectiveness / cost-utility / cost-benefit / budget-impact), CHEERS 2022 (CC BY 4.0, Husereau et al. BMJ 2022)

  • RECORD.md

    -- observational studies using routinely-collected health data (claims / EHR / registries / health-checkup DBs, linked or not), RECORD 2015 (base STROBE + RECORD extension; CC BY 4.0, Benchimol et al. PLoS Med 2015; RECORD-PE for drug studies)

  • CROSS.md

    -- survey / questionnaire studies (KAP, physician/patient, cross-sectional, e-surveys), CROSS 2021 (in-house faithful summary of item intents, Sharma et al. JGIM 2021) + CHERRIES (CC BY, Eysenbach JMIR 2004) for internet surveys

  • PRISMA_ScR.md

    -- scoping reviews (map the breadth/nature of evidence, clarify concepts, identify gaps; PCC framing, charting, optional appraisal), PRISMA-ScR 2018 (in-house faithful summary of item intents, Tricco et al. Ann Intern Med 2018; DOI 10.7326/M18-0850)

  • SRQR.md

    -- qualitative research, all approaches (ethnography / grounded theory / phenomenology / case study / narrative), SRQR 2014, 21 items (in-house faithful summary of item intents, O'Brien et al. Acad Med 2014; DOI 10.1097/ACM.0000000000000388)

  • COREQ.md

    -- qualitative research, interviews & focus groups specifically, COREQ 2007, 32 items in 3 domains (research team & reflexivity / study design / analysis & findings) (in-house faithful summary of item intents, Tong et al. Int J Qual Health Care 2007; DOI 10.1093/intqhc/mzm042)

  • REMARK.md

    -- prognostic tumor-marker / biomarker studies (single or multiple markers; e.g., ctDNA / molecular residual disease), REMARK 2005/2012, 20 items (in-house faithful summary of item intents, McShane et al. Br J Cancer 2005 + Altman et al. PLoS Med 2012)

  • TARGET.md

    -- observational studies emulating a target trial (causal / comparative-effectiveness questions on routinely-collected / registry / EHR data), TARGET 2025, 21 items (in-house faithful summary of item intents, Cashin/Hansford/Hernán et al. JAMA 2025; pairs with the /design-study target-trial-emulation module)

  • PRISMA_2020.md

    -- systematic reviews (CC BY)

  • PRISMA_2020_Abstracts.md

    -- the abstract of a systematic review / meta-analysis, 12 items (CC BY, Page et al. BMJ 2021). A separate instrument from the 27-item checklist, not a subset: item 2 of the main checklist defers to it. Score it with its own denominator.

  • ARRIVE_2.md

    -- animal studies (CC0)

  • PRISMA_DTA.md

    -- DTA systematic reviews (no open licence — © AMA; McInnes et al. JAMA 2018)

  • QUADAS3.md

    -- diagnostic accuracy risk of bias, current recommended version (no open licence -- (c) ACP; Whiting et al. Ann Intern Med 2026)

  • QUADAS2.md

    -- diagnostic accuracy risk of bias (no open licence — © ACP; Whiting et al. Ann Intern Med 2011)

  • RoB2.md

    -- RCT risk of bias (CC BY, Sterne et al. BMJ 2019)

  • ROBINS_I.md

    -- non-randomised studies risk of bias (CC BY-NC 3.0 — non-commercial; Sterne et al. BMJ 2016)

  • PROBAST.md

    -- prediction model risk of bias (no open licence — © ACP; Wolff et al. Ann Intern Med 2019)

  • NOS.md

    -- observational study quality (public domain, Ottawa Hospital)

  • CONSORT.md

    -- randomised controlled trials, CONSORT 2025 (CC BY 4.0, Hopewell et al. BMJ 2025)

  • CONSORT_AI.md

    -- AI clinical-trial reports, CONSORT-AI 2020 (CC BY 4.0, Liu et al. Nat Med 2020)

  • CARE.md

    -- case reports, CARE 2013 (no confirmed open licence — Elsevier TDM only; Gagnier et al. J Clin Epidemiol 2014)

  • SPIRIT.md

    -- clinical trial protocols, SPIRIT 2025 (CC BY 4.0, Chan et al. BMJ 2025)

  • SPIRIT_AI.md

    -- AI clinical-trial protocols, SPIRIT-AI 2020 (CC BY 4.0, Cruz Rivera et al. Nat Med 2020)

  • CLAIM_2024.md

    -- AI/ML in clinical imaging, CLAIM 2024 Update (RSNA open access, Tejani et al. Radiol Artif Intell 2024)

  • DECIDE_AI.md

    -- early-stage clinical evaluation of AI decision-support systems, DECIDE-AI 2022 (educational summary, CC BY-NC, Vasey et al. Nat Med 2022)

  • MI_CLEAR_LLM.md

    -- LLM accuracy studies in healthcare (CC BY-NC 4.0, Park et al. KJR 2024; 2025 update)

  • SQUIRE_2.md

    -- quality improvement in healthcare/education (no open licence — Crossref returns none; Ogrinc et al. BMJ Qual Saf 2016)

  • CLEAR.md

    -- radiomics studies (CC BY 4.0, Kocak et al. Insights Imaging 2023)

  • MOOSE.md

    -- meta-analysis of observational studies (Stroup et al. JAMA 2000)

  • GRRAS.md

    -- reliability and agreement studies (Kottner et al. J Clin Epidemiol 2011)

  • QUADAS_C.md

    -- comparative DTA risk of bias, extension to QUADAS-2 (no open licence — © ACP; Yang et al. Ann Intern Med 2021)

  • ROBINS_E.md

    -- non-randomised exposure studies risk of bias (CC BY-NC-ND 4.0, Higgins et al. Environ Int 2024)

  • ROBIS.md

    -- risk of bias in systematic reviews (Whiting et al. J Clin Epidemiol 2016)

  • ROB_ME.md

    -- risk of bias due to missing evidence in meta-analysis (no open licence — BMJ TDM policy only; Page et al. BMJ 2023)

  • PROBAST_AI.md

    -- prediction model risk of bias, updated for AI/ML (Moons et al. BMJ 2025)

  • COSMIN_RoB.md

    -- reliability/measurement error risk of bias (Mokkink et al. BMC Med Res Methodol 2020)

  • RoB_NMA.md

    -- risk of bias in network meta-analysis (Lunny et al. 2024)

  • AMSTAR2.md

    -- quality of systematic reviews (Shea et al. BMJ 2017)

  • PRISMA_P.md

    -- systematic review protocols (Shamseer et al. BMJ 2015)

  • SWiM.md

    -- synthesis without meta-analysis reporting (Campbell et al. BMJ 2020)

  • GATHER.md

    -- health-estimate / burden-of-disease modeling studies (GBD and GBD-satellite, comparative-risk / population-attributable-fraction, cause-of-death and prevalence/incidence estimation, with or without forecasts), GATHER 2016 (in-house faithful summary; CC BY, Stevens et al. Lancet 2016;388:e19-23 / PLoS Med 2016;13(6):e1002056). Pairs with

    /analyze-stats

    references/analysis_guides/burden_decomposition_forecasting.md

    for the analytic methods.
• Fail-fast contract: if a routed guideline has no vendored checklist file, the skill does not silently construct items from memory. It halts with a

  MISSING_CHECKLIST_CONTRACT_VIOLATION

  and surfaces the gap. A from-memory assessment is allowed only with the explicit

  --allow-from-memory

  opt-in, and that report must be clearly labelled NON-AUTHORITATIVE. See Step 2 and

  scripts/check_checklist_exists.py

  .
• Critical-item floor:

  ${CLAUDE_SKILL_DIR}/references/critical_item_floor.md

  -- the small set of non-waivable items per study type (presence outranks the headline %), plus the AI/radiomics methodological-quality / risk-of-bias instruments (PROBAST+AI, METRICS/RQS, APPRAISE-AI) kept distinct from their reporting counterparts. Loaded in Step 4f.

/analyze-stats

QUADAS3.md

QUADAS-3 has two protocol-stage phases, and this skill usually runs too late for them. Phase 1 (state the synthesis question) and phase 2 (define the ideal test accuracy trial each judgement is made against) are review-level and belong in the protocol, alongside the review-specific guidance for answering each signalling question. Reaching them for the first time during manuscript QC means writing the comparator after seeing the results. If they are missing, say so as a limitation rather than reconstructing them — and route the protocol work to

/meta-analysis

Phase 1. Phases 3–6 are what a QC pass can genuinely run.

• If the study involves AI/ML, always apply the AI extension in addition to the base guideline.
  • Exception — TRIPOD: TRIPOD+AI 2024 (Collins et al., BMJ 2024) is a complete rewrite, not an addendum to TRIPOD 2015 (Moons et al., Ann Intern Med 2015). For non-AI prediction models, use TRIPOD 2015 only. For AI/ML prediction models, use TRIPOD+AI 2024 only. Do NOT apply both simultaneously.
• STARD-AI (Sounderajah et al., Nat Med 2025) extends STARD 2015 with 14 new and 4 modified items (40 total). For AI diagnostic accuracy studies, use STARD-AI (which incorporates all STARD 2015 items). Do NOT apply both STARD 2015 and STARD-AI simultaneously — STARD-AI supersedes STARD 2015 for AI studies.
• TRIPOD-LLM (Gallifant et al., Nat Med 2025) is the reporting guideline for studies that develop, fine-tune, prompt, or evaluate a large language model for a clinical/biomedical task. It extends the TRIPOD family (TRIPOD 2015 → TRIPOD+AI 2024 → TRIPOD-LLM 2025); name the base instrument and the extension and cite each. It is modular — task-specific items (Annotation, Prompting, Summarization, Instruction-tuning) are N/A when that component is absent. Use TRIPOD-LLM for LLM studies in place of TRIPOD+AI; pair with MI-CLEAR-LLM when LLM accuracy is an evaluated outcome. The vendored checklist is an educational summary (own-words paraphrase of item intent); complete the official instrument for a submission checklist.
• MI-CLEAR-LLM is a supplementary checklist (8 item categories in the 2025 update; the 2024 original had 6), not a standalone reporting guideline. Always pair it with the study's primary guideline (e.g., STARD-AI for AI diagnostic accuracy, CLAIM for imaging AI). Apply MI-CLEAR-LLM whenever the study evaluates LLM accuracy as an outcome — do NOT apply it merely because the manuscript was written with LLM assistance. Its scope is LLM accuracy studies (including VLMs interpreting images); it does not apply at study level to studies where a generative model produces the images under study (see next bullet).
• Generative-AI images as the study object (a generative model synthesizes images and the study evaluates their realism, controllability, real-vs-synthetic distinguishability, or model-vs-model quality) has no single dominant checklist. Assemble: CLAIM 2024 (imaging-AI umbrella; model-development items N/A when commercial models are used as-is) + FUTURE-AI traceability + MI-CLEAR-LLM transparency items only (prompt/model/version/params/runs — for generation provenance, not study-level compliance) on the generator side; STARD-AI (for real-vs-synthetic detection) + GRRAS (reader reliability) + MRMC reporting on the evaluation side. Map applicable items and cite base + extension; never claim wholesale compliance. Full decision aid:

  ${CLAUDE_SKILL_DIR}/references/genai_image_study_object_decision_aid.md

  .
• If multiple guidelines apply (e.g., a diagnostic accuracy study that is also an AI study), check against all relevant guidelines and merge into one report.
• If the user requests a specific guideline, use that one regardless of auto-detection.

1. Run the fail-fast guard first for every guideline you intend to apply:
   bash

   python "${CLAUDE_SKILL_DIR}/scripts/check_checklist_exists.py" --guideline "STARD-AI"

   • Exit 0 → the vendored checklist exists; read it from

     ${CLAUDE_SKILL_DIR}/references/checklists/

     and proceed.
   • Exit 1 (

     MISSING_CHECKLIST_CONTRACT_VIOLATION

     ) → the guideline is routed but no checklist file is vendored. Do not construct items from memory. Halt, report the violation to the user, and stop unless they explicitly opt in (next bullet).
   • Exit 2 (

     UNKNOWN_GUIDELINE

     ) → the name is not recognised; confirm the correct guideline with the user.
2. No silent fallback. A from-memory checklist is permitted only when the user explicitly accepts it — re-run the guard with

   --allow-from-memory

   (exit 0 + a NON-AUTHORITATIVE warning). In that case the output report MUST carry a prominent banner that the assessment was constructed from model knowledge and is not backed by a vendored checklist, and

   submission_safe

   must not be asserted on its basis.

1. Title and abstract
2. Introduction
3. Methods (all subsections)
4. Results (all subsections)
5. Discussion
6. Tables, figures, and their captions
7. Supplemental materials (if available)
8. References (for registration numbers, protocol references)

• Status: PRESENT / PARTIAL / MISSING / N/A
• Location: Section name and paragraph or approximate position (e.g., "Methods, paragraph 3")
• Notes: What was found (if PRESENT/PARTIAL) or what should be added (if MISSING)

• Results section contains only factual findings: no interpretation, no "why" explanations, no prior literature comparisons, no evaluative adjectives without numbers.
• Discussion section does not introduce new data not presented in Results.
• Flag any boundary violation as a separate finding in Part C Action Items with the label

  [BOUNDARY]

  .

1. records screened = records identified − duplicates removed
2. records sought-for-retrieval = records screened − records excluded (screening)
3. reports retrieved = sought − reports not retrieved
4. studies included = reports assessed for eligibility − reports excluded (with reasons)

• Body text PRISMA numbers (e.g., "315 records identified, 122 duplicates removed, 186 records screened") match Figure 1 box labels 1:1.
• Reasons for exclusion (Methods + Figure legend) agree on counts and category names.

qc/prisma_figure_audit.json

python3 ${CLAUDE_SKILL_DIR}/scripts/check_prisma_figure.py \
  --md <manuscript.md> --figure <Figure 1 source: .md manifest / caption / text export> \
  --out qc/prisma_figure_audit.json

1

[PRISMA-FIGURE]

fixable_by_ai: false

2

1. Extract numbers from manuscript Results / PRISMA flow paragraph (regex: integers near keywords

   identified

   ,

   duplicates

   ,

   screened

   ,

   excluded

   ,

   sought

   ,

   retrieved

   ,

   assessed

   ,

   included

   ).
2. Extract numbers from Figure 1 source — preferred order: (a)

   analysis/figures/Figure1_PRISMA.md

   markdown manifest, (b) caption text in

   manuscript.md

   , (c) PPTX text run if

   .pptx

   exists, (d) manual entry from PNG/SVG.
3. Cross-check

   analysis/figures/_figure_manifest.md

   (produced by

   /make-figures

   ): verify that the row whose

   Type = prisma

   (or

   Type = prisma-dta

   ) points at the same file path used as the audit source, and that the row's

   Critic

   field is

   yes

   or

   partial

   (not

   no

   ). A missing manifest row, mismatched path, or

   Critic = no

   flag logs

   [MANIFEST-XREF]

   (advisory) — the arithmetic check still runs against the source identified in step 2. Skip this sub-step if

   _figure_manifest.md

   does not exist (older projects).
4. Run 4 arithmetic checks; emit PRESENT / MISSING / MISMATCH per equation.
5. Run 2 cross-reference checks; emit PRESENT / MISSING / MISMATCH per number.
6. Output

   qc/prisma_figure_audit.json

   and a short table.

[PRISMA-FIGURE]

fixable_by_ai: false

${CLAUDE_SKILL_DIR}/references/step4d_prisma_figure_audit.md

~/.claude/rules/numerical-safety.md

compliance_pct

fixable_by_ai

Item | Recommendation | Reported in page/section

<!-- INTERNAL AUDIT — NOT FOR SUBMISSION. This is the /check-reporting working
report, not the official journal checklist. Do not upload to a submission portal. -->

/sync-submission

check_checklist_dump_leak

${CLAUDE_SKILL_DIR}/references/report_templates.md

• Part A — Summary. Header (manuscript file, version token, guideline, date), the PRESENT/PARTIAL/MISSING/N-A count table, and overall compliance. The headline is the critical items (Step 4f), not the percentage: report

  {present}/{total}

  and name every missing critical item with the section it belongs in.
• Part B — Item-by-item checklist. One row per item:

  # | Section | Item | Status | Location | Notes

  .
• Part C — Action items (MISSING and PARTIAL only), ordered by: items most journals enforce strictly (ethics approval, registration, sample size) → items in Methods (easiest to fix) → everything else.
• Part D — Machine-readable JSON, appended as a fenced block. MUST be present under

  --json

  or when called from

  /write-paper

  Phase 7, which parses it.

• compliance_pct

  —

  present / (total_items - na) * 100

  , one decimal.

• action_items

  — MISSING and PARTIAL only; PRESENT and N/A are excluded.

• fixable_by_ai

  —

  true

  when the fix inserts or expands text using information already in the manuscript or inferable from it;

  false

  when it needs external facts the author alone holds (registration number, IRB approval number, protocol details).

• suggested_fix

  — concrete draft text, insertable as written.

• source_sha256

  — first 12 hex chars of the SHA-256 of the manuscript bytes, so a stale report cannot be silently attributed to a newer manuscript.

references/report_templates.md

• PARTIAL means the item is mentioned but lacks specificity. For example:
  • "We used appropriate statistical tests" = PARTIAL (which tests?)
  • "We used the Mann-Whitney U test for continuous variables and Fisher's exact test for categorical variables" = PRESENT
• A vague reference does not count as PRESENT. The detail level must match what the guideline expects.

• For MISSING items, provide a draft sentence the user can insert.
• For PARTIAL items, point to the exact gap and suggest specific additions.
• Reference the specific manuscript section where the addition should go.

1. Study registration number (CONSORT, PRISMA, STARD)
2. Registration / amendment date consistency (PRISMA 2020, PRISMA-DTA, CONSORT, SPIRIT) — run Step 4c whenever a registration identifier is present
3. Sample size justification (CONSORT, STROBE, STARD)
4. Missing data handling (all guidelines)
5. Blinding details (CONSORT, STARD)
6. Funding and conflicts of interest (all guidelines)
7. Ethics approval with committee name and approval number (all guidelines)
8. Data availability statement (increasingly required)
9. AI-specific: training/validation/test split details (TRIPOD+AI, CLAIM, STARD-AI)
10. AI-specific: model architecture and hyperparameters (TRIPOD+AI, CLAIM, STARD-AI)
11. AI-specific: failure mode analysis (CLAIM, STARD-AI)
12. AI-specific: fairness/bias assessment (STARD-AI)
13. AI-specific: commercial interests and data/code availability (STARD-AI)
14. Power-aware framing of a null result (STROBE 16a / 18 / 20) — for an observational study whose headline is a non-significant association, a flat "X was not associated with Y" overreads the data when the analysis is not powered to exclude a clinically meaningful effect. Mark item 18/20 PARTIAL unless the manuscript states the precision as an exclusion (e.g., "the 95% CI excluded an eGFR difference larger than ~1.7") or reports a minimum detectable effect — "no effect" vs "could not exclude an effect of size X" are different claims, and a negative conclusion needs the latter.
15. Confounder-selection rationale, not "adjust for everything that differs" (STROBE 16a explicitly asks which confounders were adjusted for and why) — flag a kitchen-sink adjustment set chosen because variables differ in Table 1. The Methods must give a causal rationale (DAG / prior literature) and must not adjust for a mediator or consequence of the outcome (over-adjustment, e.g. serum uric acid in an eGFR model); both an unjustified inclusion and an unjustified omission are item-16a gaps.

151 + 108 + 39 + 1 + 1 + 4 = 304

python "${CLAUDE_SKILL_DIR}/scripts/prisma_cascade_check.py" \
    --round1 2_Screening/round1.tsv \
    --round2 2_Screening/round2.tsv \
    --round3 2_Screening/round3_adjudication.tsv \
    --manuscript manuscript.md \
    --out qc/prisma_cascade.json

1. Reads the round TSVs and counts

   INCLUDE

   /

   EXCLUDE

   /

   MAYBE

   decisions per round.
2. Computes the cascade arithmetic from raw decisions (no prose).
3. Optionally grep the manuscript for matching stage-count claims and emits per-stage drift when the prose disagrees.

manuscript_drift

• If the manuscript file cannot be read, ask the user for the correct path.
• If the study type is ambiguous, ask the user to confirm before selecting a guideline.
• If a checklist item is genuinely unclear in its applicability, mark as N/A with justification.
• This is a pre-screening tool. Always remind the user that final compliance should be verified by all co-authors and ideally by a methodologist.

• Checklist content and compliance report: English
• Communication with user: Match user's preferred language
• Medical terms: English only

• Never fabricate references. All citations must be verified via

  /search-lit

  with confirmed DOI or PMID. Mark unverified references as

  [UNVERIFIED - NEEDS MANUAL CHECK]

  .
• Never invent clinical definitions, diagnostic criteria, or guideline recommendations. If uncertain, flag with

  [VERIFY]

  and ask the user.
• Never fabricate numerical results — compliance percentages, scores, effect sizes, or sample sizes must come from actual data or analysis output.
• If a reporting guideline item, journal policy, or clinical standard is uncertain, state the uncertainty rather than guessing.