medsci-skills 4.1.0

package/skills/academic-aio/references/checklists/AIO_GENERAL.md

@@ -0,0 +1,354 @@
+# Academic AIO General Checklist
+
+> Machine-readable checklist mirroring `SKILL.md` sections 1–12. Use as the canonical pass/fail audit table for any medical-AI artifact (manuscript, preprint, README, CITATION.cff, HF card). Each item maps to a SKILL.md section so audit findings can be traced back to the rule.
+
+## How to use
+
+1. Copy this checklist into the working artifact directory as `qc/aio_audit.md`.
+2. Determine the artifact's **lifecycle phase** (`pre-draft`, `drafting`, `pre-submission`, `post-acceptance`, `post-publication`) and filter rules whose `applies_to_phase` does not include the current phase — those become NA (do not surface as FAIL).
+3. Mark each remaining item PASS / PARTIAL / FAIL with a one-line reason.
+4. For FAIL items, generate a concrete edit suggestion ranked by `expected_lift` (`high` first, then `medium`, then `low`).
+5. Re-run after edits until ≥ 90 % of applicable items pass.
+6. Where an item carries a `defers_to` link, item-level detail belongs to the linked skill — record only the high-level status here to avoid duplicate audits.
+
+## Schema (v2)
+
+```
+items[].id              : §-numbered rule id
+items[].rule            : one-line description
+items[].applies_to      : artifact types where the rule fires
+items[].applies_to_phase: lifecycle phases where the rule is actionable
+items[].priority        : H | M | L  (severity if missing)
+items[].expected_lift   : high | medium | low  (KJR-case-validated visibility gain)
+items[].defers_to       : optional skill / file owning item-level detail
+```
+
+## Checklist
+
+```yaml
+checklist:
+  schema_version: 2
+  last_updated: "2026-05-11"
+
+  metadata:
+    artifact_path: ""
+    artifact_type: ""        # manuscript | preprint | readme | citation_cff | hf_card | dataset_card
+    artifact_phase: ""       # pre-draft | drafting | pre-submission | post-acceptance | post-publication
+    journal_target: ""
+    audit_date: ""           # YYYY-MM-DD
+    reviewer: ""
+
+  items:
+    # Section 1 — Title and Abstract Optimization
+    - id: 1.1
+      rule: Title three-slot ([Task] + [Modality/anatomy] + [Model class])
+      applies_to: [title]
+      applies_to_phase: [drafting, pre-submission]
+      priority: H
+      expected_lift: high
+    - id: 1.2
+      rule: Structured abstract per journal template; chunk-friendly ≤3-sentence sub-blocks
+      applies_to: [abstract]
+      applies_to_phase: [drafting, pre-submission]
+      priority: H
+      expected_lift: high
+    - id: 1.3
+      rule: First sentence states problem + contribution; last sentence is explicit interpretation
+      applies_to: [abstract]
+      applies_to_phase: [drafting, pre-submission]
+      priority: H
+      expected_lift: high
+    - id: 1.4
+      rule: Taxonomy line names controlled vocabulary (e.g., DTA, foundation-model evaluation)
+      applies_to: [abstract]
+      applies_to_phase: [pre-submission]
+      priority: M
+      expected_lift: low
+    - id: 1.5
+      rule: Quantified primary outcome with 95 % CI in abstract
+      applies_to: [abstract]
+      applies_to_phase: [drafting, pre-submission]
+      priority: H
+      expected_lift: high
+    - id: 1.6
+      rule: Reporting-guideline anchor present (TRIPOD+AI / CLAIM / STARD-AI / TRIPOD-LLM / DECIDE-AI / PRISMA-DTA)
+      applies_to: [abstract, methods]
+      applies_to_phase: [drafting, pre-submission]
+      priority: H
+      expected_lift: high
+      defers_to: /check-reporting   # item-level audit; AIO records anchor-present status only
+    - id: 1.7
+      rule: MeSH and RadLex coverage; consistent UK/US spelling; no abstract-keyword redundancy
+      applies_to: [keywords, abstract]
+      applies_to_phase: [pre-submission]
+      priority: M
+      expected_lift: medium
+
+    # Section 2 — Manuscript-Level
+    - id: 2.1
+      rule: Journal-specific summary box present (Research in context / Key Points / PLS / Main Points)
+      applies_to: [manuscript]
+      applies_to_phase: [drafting, pre-submission]
+      priority: H
+      expected_lift: high   # KJR case: top extracted fragment by AI overviews
+    - id: 2.2
+      rule: Section headings are declarative claims, not generic labels
+      applies_to: [manuscript]
+      applies_to_phase: [drafting, pre-submission]
+      priority: M
+      expected_lift: medium
+    - id: 2.3
+      rule: Numeric claim compression sentence in Methods and Results
+      applies_to: [methods, results]
+      applies_to_phase: [drafting, pre-submission]
+      priority: M
+      expected_lift: medium
+    - id: 2.4
+      rule: Reproducibility block (data, code, weights, prompts, seeds, env)
+      applies_to: [methods, data_availability]
+      applies_to_phase: [pre-submission]
+      priority: H
+      expected_lift: high
+    - id: 2.5
+      rule: Limitations enumerated and named
+      applies_to: [discussion]
+      applies_to_phase: [drafting, pre-submission]
+      priority: M
+      expected_lift: medium
+    - id: 2.6
+      rule: Standalone figure captions restate claim, dataset, and metric
+      applies_to: [figures]
+      applies_to_phase: [pre-submission]
+      priority: M
+      expected_lift: low
+
+    # Section 3 — Preprint, Channel, Indexing
+    - id: 3.1
+      rule: Preprint posted on submission day, OR fast-track justified
+      applies_to: [submission_strategy]
+      applies_to_phase: [pre-submission]
+      priority: H
+      expected_lift: high
+    - id: 3.2
+      rule: Journal preprint policy verified (Sherpa Romeo or IFA)
+      applies_to: [submission_strategy]
+      applies_to_phase: [pre-submission]
+      priority: H
+      expected_lift: medium
+    - id: 3.4
+      rule: Open-access route selected (gold OA preferred; green OA acceptable)
+      applies_to: [submission_strategy]
+      applies_to_phase: [pre-draft, pre-submission]
+      priority: H
+      expected_lift: high
+      defers_to: references/oac_funding_checklist.yaml
+    - id: 3.5
+      rule: Post-acceptance channel checklist scheduled (PMC, ORCID, Scholar, SNS, HF)
+      applies_to: [post_acceptance]
+      applies_to_phase: [post-acceptance]
+      priority: M
+      expected_lift: high
+
+    # Section 5 — GitHub / CITATION.cff / Zenodo / HF
+    - id: 5.1
+      rule: README follows 10-slot canonical order
+      applies_to: [readme]
+      applies_to_phase: [post-acceptance, post-publication]
+      priority: H
+      expected_lift: medium
+    - id: 5.2
+      rule: CITATION.cff present at repo root with ORCID, DOI, version
+      applies_to: [citation_cff]
+      applies_to_phase: [post-acceptance, post-publication]
+      priority: H
+      expected_lift: medium
+    - id: 5.3
+      rule: Zenodo DOI minted via GitHub integration; cited in paper
+      applies_to: [zenodo, manuscript]
+      applies_to_phase: [pre-submission, post-acceptance]
+      priority: H
+      expected_lift: high
+    - id: 5.4
+      rule: Hugging Face model card YAML keys + required prose sections complete
+      applies_to: [hf_card]
+      applies_to_phase: [post-acceptance, post-publication]
+      priority: H
+      expected_lift: medium
+    - id: 5.5
+      rule: Hugging Face dataset card includes PHI/re-identification disclosure
+      applies_to: [dataset_card]
+      applies_to_phase: [post-acceptance, post-publication]
+      priority: H
+      expected_lift: medium
+    - id: 5.6
+      rule: Web-crawler-friendly markdown (declarative headings, alt text, fenced code, JSON-LD)
+      applies_to: [readme, hf_card]
+      applies_to_phase: [post-acceptance, post-publication]
+      priority: M
+      expected_lift: medium
+
+    # Section 6 — Authority / E-E-A-T
+    - id: 6.1
+      rule: Personal author landing page lists all papers with DOIs
+      applies_to: [author_profile]
+      applies_to_phase: [pre-draft, drafting, pre-submission, post-acceptance, post-publication]
+      priority: M
+      expected_lift: medium
+    - id: 6.2
+      rule: Affiliation string consistent across papers
+      applies_to: [author_profile]
+      applies_to_phase: [drafting, pre-submission]
+      priority: H
+      expected_lift: high
+    - id: 6.3
+      rule: ORCID complete and linked to Google Scholar
+      applies_to: [author_profile]
+      applies_to_phase: [pre-draft, drafting, pre-submission, post-acceptance, post-publication]
+      priority: H
+      expected_lift: high
+
+    # Section 7 — LLM-Citation Fabrication Defense
+    - id: 7.1
+      rule: DOI + PMID surfaced in copy-friendly text on landing page and README
+      applies_to: [readme, manuscript]
+      applies_to_phase: [post-acceptance, post-publication]
+      priority: H
+      expected_lift: high
+    - id: 7.2
+      rule: "How to cite" section with BibTeX, APA, Vancouver, plain-text
+      applies_to: [readme]
+      applies_to_phase: [post-acceptance, post-publication]
+      priority: H
+      expected_lift: medium
+    - id: 7.3
+      rule: Scholar alert configured; Perplexity / ChatGPT probes scheduled
+      applies_to: [post_acceptance]
+      applies_to_phase: [post-publication]
+      priority: M
+      expected_lift: low
+
+    # Section 10 — Q&A and Entity-Extraction
+    - id: 10.1
+      rule: Four-question Q&A block (What was known / What this adds / How this changes practice / Why it matters)
+      applies_to: [discussion, supplement]
+      applies_to_phase: [drafting, pre-submission]
+      priority: H
+      expected_lift: high   # KJR case: directly correlated with AI-overview extraction
+    - id: 10.2
+      rule: Glossary block with MeSH / RadLex / UMLS / arXiv IDs
+      applies_to: [methods, supplement]
+      applies_to_phase: [pre-submission]
+      priority: L
+      expected_lift: low   # over-engineering for non-NLP / non-foundation-model papers; apply selectively
+    - id: 10.3
+      rule: Inline citation anchor text uses semantic predicates, not bare numbers
+      applies_to: [manuscript]
+      applies_to_phase: [drafting, pre-submission]
+      priority: M
+      expected_lift: medium
+    - id: 10.4
+      rule: Explicit "Why this is hard" challenge statement at start of Discussion
+      applies_to: [discussion]
+      applies_to_phase: [drafting, pre-submission]
+      priority: H
+      expected_lift: high   # KJR case: top quoted fragment by Perplexity / ChatGPT web
+
+    # Section 11 — First-Mover Timing and Citation-Graph
+    - id: 11.1
+      rule: Topic-peak signals checked before submission timing decision
+      applies_to: [submission_strategy]
+      applies_to_phase: [pre-draft]
+      priority: H
+      expected_lift: high
+    - id: 11.2
+      rule: Editorial-board leverage considered (corresponding author affiliation)
+      applies_to: [submission_strategy]
+      applies_to_phase: [pre-draft]
+      priority: M
+      expected_lift: medium
+    - id: 11.3
+      rule: Target journal verified as PMC-auto-deposit when applicable
+      applies_to: [submission_strategy]
+      applies_to_phase: [pre-draft, pre-submission]
+      priority: H
+      expected_lift: high
+      defers_to: references/oac_funding_checklist.yaml
+    - id: 11.4
+      rule: Discussion anchored in 5–10 high-visibility seminal works
+      applies_to: [discussion]
+      applies_to_phase: [drafting]
+      priority: M
+      expected_lift: medium
+    - id: 11.5
+      rule: Author roster spans ≥ 3 institutions and ≥ 2 disciplines (review papers)
+      applies_to: [authorship]
+      applies_to_phase: [pre-draft]
+      priority: M
+      expected_lift: medium   # not actionable once drafting begins; auto-NA in later phases
+
+    # Section 12 — Cross-Platform Launch Sequencing
+    - id: 12.1
+      rule: Day-0 simultaneous launch (GitHub release, HF card, X/Threads, LinkedIn, landing page)
+      applies_to: [post_acceptance]
+      applies_to_phase: [post-acceptance]
+      priority: H
+      expected_lift: high
+    - id: 12.2
+      rule: Day-1 propagation (ORCID, Scholar, preprint update, institutional page)
+      applies_to: [post_acceptance]
+      applies_to_phase: [post-acceptance]
+      priority: M
+      expected_lift: medium
+    - id: 12.3
+      rule: Week-1 depth posts (LinkedIn long-form, Papers with Code, ResearchGate)
+      applies_to: [post_acceptance]
+      applies_to_phase: [post-acceptance]
+      priority: M
+      expected_lift: medium
+    - id: 12.4
+      rule: Weeks 2–4 refresh signals scheduled
+      applies_to: [post_acceptance]
+      applies_to_phase: [post-publication]
+      priority: L
+      expected_lift: low
+    - id: 12.5
+      rule: Month-1 monitoring (alerts + AI-search probes)
+      applies_to: [post_acceptance]
+      applies_to_phase: [post-publication]
+      priority: M
+      expected_lift: medium
+```
+
+## Output template (paste into qc/aio_audit.md)
+
+```markdown
+# AIO Audit — {artifact_path}
+
+**Phase**: {pre-draft | drafting | pre-submission | post-acceptance | post-publication}
+**Top 5 ranked by expected_lift (high → medium → low):**
+
+| Rank | ID | Rule | Status | Reason | Suggested edit |
+|------|----|------|--------|--------|----------------|
+| 1 | {id} | … | FAIL | … | … |
+| ... |
+
+## Full table (applicable items only — out-of-phase rules auto-NA)
+
+| ID | Rule | Lift | Status | Reason | Suggested edit |
+|----|------|------|--------|--------|----------------|
+| ... |
+
+## Deferred items (audit elsewhere)
+
+- §1.6 → run `/check-reporting`
+- §3.4 / §11.3 → see `references/oac_funding_checklist.yaml`
+
+## Summary
+
+- PASS: X / Y applicable items
+- High-lift fixes pending: …
+```
+
+## Migration from schema v1
+
+If a prior `qc/aio_audit.md` was generated under schema v1 (no `applies_to_phase` / `expected_lift`), regenerate from this v2 file. The id list is unchanged so prior PASS/FAIL state can be re-imported by id.

package/skills/academic-aio/references/journal_summarybox_templates.yaml

@@ -0,0 +1,126 @@
+---
+schema_version: 1
+last_updated: "2026-05-10"
+purpose: |
+  Templates for journal-specific summary boxes used in medical-AI manuscripts.
+  Each entry mirrors the journal's Instructions for Authors (IFA) at the date
+  shown. USER MUST VERIFY the current IFA before applying — journal policies
+  change without notice.
+
+verification_protocol: |
+  Before applying any template:
+  1. Open the journal's current IFA URL (ifa_url field).
+  2. Compare structure / labels / word targets to this YAML.
+  3. If any field has changed since `last_verified`, update this file before
+     deriving the manuscript's box.
+  4. Treat `last_verified` older than 6 months as STALE — re-verify before
+     submission.
+
+journals:
+  - id: lancet-digital-health
+    name: The Lancet Digital Health
+    box_label: Research in context
+    ifa_url: https://www.thelancet.com/journals/landig/about
+    last_verified: "2026-05-10"
+    placement: After Introduction in main text, or as Panel 1.
+    structure:
+      - id: evidence-before
+        label: Evidence before this study
+        word_target: 100-200
+        notes: |
+          Describe the systematic search method (databases, dates, terms) and
+          what was known prior to this study.
+      - id: added-value
+        label: Added value of this study
+        word_target: 100-200
+        notes: |
+          State the contribution and how it advances the field beyond the
+          evidence summarized above.
+      - id: implications
+        label: Implications of all the available evidence
+        word_target: 100-200
+        notes: |
+          How findings (combined with prior evidence) should change clinical
+          practice, policy, or future research.
+    user_must_verify: true
+
+  - id: rsna-radiology
+    name: Radiology (RSNA)
+    box_label: Key Results
+    ifa_url: https://pubs.rsna.org/page/radiology/author-instructions
+    last_verified: "2026-05-10"
+    placement: First page of article, after Abstract.
+    structure:
+      - id: key-results
+        label: Key Results
+        format: 3 bullets, claim-centric, one finding each
+        notes: |
+          Each bullet should state a primary outcome with point estimate and
+          confidence interval where applicable. Avoid hedging language.
+    user_must_verify: true
+
+  - id: rsna-radiology-ai
+    name: Radiology Artificial Intelligence (RSNA)
+    box_label: Key Points
+    ifa_url: https://pubs.rsna.org/page/ai/author-instructions
+    last_verified: "2026-05-10"
+    placement: After Abstract.
+    structure:
+      - id: key-points
+        label: Key Points
+        format: 3 bullets
+        notes: |
+          Same conventions as Radiology Key Results — claim-centric, numeric
+          where possible.
+    user_must_verify: true
+
+  - id: npj-digital-medicine
+    name: npj Digital Medicine
+    box_label: Plain Language Summary
+    ifa_url: https://www.nature.com/npjdigitalmed/submission-guidelines
+    last_verified: "2026-05-10"
+    placement: After Abstract; required for all primary research.
+    structure:
+      - id: pls
+        label: Plain Language Summary
+        word_target: 150-200
+        reading_level: 8th-grade
+        notes: |
+          Avoid jargon, define acronyms inline, prefer everyday vocabulary.
+          Should be readable by a non-specialist physician or informed lay
+          reader. No quantitative thresholds; describe magnitude in plain terms.
+    user_must_verify: true
+
+  - id: nature-medicine
+    name: Nature Medicine
+    box_label: Editor's Summary
+    ifa_url: https://www.nature.com/nm/for-authors
+    last_verified: "2026-05-10"
+    placement: Editorial-supplied at acceptance; not author-drafted in main text.
+    structure:
+      - id: editor-summary
+        label: Editor's Summary
+        word_target: 100-150
+        notes: |
+          Supplied by the editorial team after acceptance. Authors should
+          proactively draft a 100–150 word version for the cover letter and
+          post-acceptance launch materials, so the editor has anchor wording
+          to refine.
+    user_must_verify: true
+
+  - id: jamia
+    name: JAMIA (Journal of the American Medical Informatics Association)
+    box_label: (none mandated; structured abstract used)
+    ifa_url: https://academic.oup.com/jamia/pages/General_Instructions
+    last_verified: "2026-05-10"
+    placement: Structured abstract serves the role of summary box.
+    structure:
+      - id: structured-abstract
+        label: Structured Abstract
+        format: Objective / Materials and Methods / Results / Discussion / Conclusion
+        notes: Treat each section as a chunk-friendly semantic block (≤ 3 sentences).
+    user_must_verify: true
+
+related:
+  - skills/academic-aio/SKILL.md Section 2.1 (manuscript-level summary box)
+  - skills/academic-aio/SKILL.md Section 1.6 (reporting-guideline anchor)

package/skills/academic-aio/references/oac_funding_checklist.yaml

@@ -0,0 +1,129 @@
+---
+schema_version: 1
+last_updated: "2026-05-10"
+purpose: |
+  Open-access compliance matrix for major funders that medical-AI researchers
+  may encounter. Use to decide gold OA vs green OA vs preprint route per paper.
+  Funder policies change; verify the policy_url before applying.
+
+verification_protocol: |
+  All funder policies change without notice.
+  - Verify policy_url before applying decisions.
+  - Treat `last_verified` older than 6 months as STALE.
+  - When in doubt, contact the institution's research-administration office.
+
+funders:
+  - id: plan-s
+    name: cOAlition S (Plan S)
+    region: Europe (multiple national funders)
+    policy_url: https://www.coalition-s.org/plan-s-principles/
+    last_verified: "2026-05-10"
+    requirement: Immediate OA on publication; no embargo allowed.
+    accepted_routes:
+      - Gold OA with CC-BY (preferred)
+      - Green OA via repository deposit at publication (no embargo)
+    apc_cap: ~ EUR 2,500 typical, varies by funder
+    notes: Strict; transformative agreements widely available.
+    user_must_verify: true
+
+  - id: nih-public-access
+    name: US NIH Public Access Policy
+    region: United States
+    policy_url: https://publicaccess.nih.gov/
+    last_verified: "2026-05-10"
+    requirement: |
+      From 2025 onward NIH policy requires manuscript deposit in PMC at
+      publication (no 12-month embargo). Earlier grants may still operate
+      under the legacy 12-month rule.
+    accepted_routes:
+      - Gold OA (publisher deposits to PMC)
+      - Green OA (author deposits AAM to PMC via NIHMS)
+    apc_cap: No cap; APCs are eligible expenses on most grants.
+    user_must_verify: true
+
+  - id: ukri
+    name: UK Research and Innovation (UKRI)
+    region: United Kingdom
+    policy_url: https://www.ukri.org/manage-your-award/publishing-your-research-findings/making-your-research-open/
+    last_verified: "2026-05-10"
+    requirement: Immediate OA; CC-BY preferred.
+    accepted_routes:
+      - Gold OA via transformative agreement (preferred)
+      - Green OA in approved repository at publication
+    user_must_verify: true
+
+  - id: gates-foundation
+    name: Bill & Melinda Gates Foundation
+    region: Global
+    policy_url: https://openaccess.gatesfoundation.org/
+    last_verified: "2026-05-10"
+    requirement: Immediate OA, CC-BY, no embargo.
+    accepted_routes:
+      - Gold OA (mandatory)
+    notes: Strictest among major funders. Closed-access publication is non-compliant.
+    user_must_verify: true
+
+  - id: wellcome
+    name: Wellcome Trust
+    region: UK / global
+    policy_url: https://wellcome.org/grant-funding/guidance/open-access-policy
+    last_verified: "2026-05-10"
+    requirement: Immediate OA, CC-BY, including for preprints.
+    accepted_routes:
+      - Gold OA (preferred)
+      - Green OA at publication
+    user_must_verify: true
+
+  - id: nrf-korea
+    name: National Research Foundation of Korea (NRF)
+    region: South Korea
+    policy_url: https://www.nrf.re.kr/
+    last_verified: "2026-05-10"
+    requirement: |
+      No formal mandatory immediate-OA policy as of 2026-Q2. Gold OA is
+      encouraged and APCs are eligible expenses on most grants.
+    accepted_routes:
+      - Gold OA (encouraged)
+      - Green OA (acceptable)
+    user_must_verify: true
+
+  - id: hira-mohw
+    name: Korean Ministry of Health & Welfare research grants
+    region: South Korea
+    policy_url: https://www.mohw.go.kr/
+    last_verified: "2026-05-10"
+    requirement: Public benefit / repository deposit encouraged; no strict immediate-OA mandate.
+    accepted_routes:
+      - Gold OA (encouraged)
+      - Green OA (acceptable)
+    user_must_verify: true
+
+decision_tree:
+  - step: 1
+    question: Does any funder on this paper require immediate OA?
+    yes_action: |
+      Choose gold OA at a Plan-S compliant journal, OR green OA with
+      repository deposit at publication. Closed-access is not allowed.
+    no_action: Continue to step 2.
+  - step: 2
+    question: Is APC budget available (grant line, institutional waiver, or transformative agreement)?
+    yes_action: |
+      Gold OA preferred — lower friction, faster PMC deposit, cited
+      30-50 % more by AI-search tools (Section 3.4).
+    no_action: |
+      Green OA via preprint at submission + AAM deposit at acceptance.
+      Verify journal's green-OA timeline.
+  - step: 3
+    question: Does the target journal auto-deposit to PMC (Section 11.3)?
+    yes_action: Confirm and proceed. Deposit window typically 4-6 weeks.
+    no_action: |
+      Plan green OA pathway and timeline. Budget the AAM submission step
+      (NIHMS or Europe PMC).
+  - step: 4
+    question: Is a transformative agreement available via the institution?
+    yes_action: Gold OA is effectively free for the author — use it.
+    no_action: Apply APC budget directly or fall back to green OA.
+
+related:
+  - skills/academic-aio/SKILL.md Section 3.4 (open-access choice)
+  - skills/academic-aio/SKILL.md Section 11.3 (PMC-auto-deposit journal preference)

package/skills/academic-aio/references/reporting_guideline_mapping.md

@@ -0,0 +1,39 @@
+# Reporting-Guideline ↔ AIO Rule Mapping
+
+This table maps each AIO rule (sections 1-12 of `SKILL.md`) to the corresponding item(s) in the major medical-AI reporting guidelines. Use it to align the `/check-reporting` audit with the `/academic-aio` audit so the same evidence covers both.
+
+## Core mapping
+
+| AIO rule | TRIPOD+AI 2024 | CLAIM 2024 | STARD-AI 2025 | TRIPOD-LLM 2024 | DECIDE-AI 2022 | Notes |
+|----------|----------------|-------------|----------------|-----------------|----------------|-------|
+| §1.1 Title three-slot | item 1 (title) | item 1 (title and abstract) | item 1 (title) | item 1a (title) | — | All require keyword presence and study-type identification. |
+| §1.2 Structured abstract | item 2 (abstract) | item 1 (title and abstract) | item 2 (abstract) | item 1b (abstract) | — | Each guideline mandates structured form. |
+| §1.5 Quantified primary outcome with CI | item 16 (model performance) | items 28-30 (performance metrics) | items 23-26 (diagnostic estimates) | item 17 (performance with CI) | item 8 (clinical effect) | CI mandatory for all. |
+| §1.6 Reporting-guideline anchor | (compliance declaration) | (compliance declaration) | (compliance declaration) | (compliance declaration) | (compliance declaration) | Cite guideline + checklist in Methods or supplement. |
+| §2.4 Reproducibility block | item 24 (data sharing) + item 25 (code sharing) | items 33-34 (data and code availability) | item 28 (data and code) | items 22-23 (artifacts) | item 12 (artifacts) | All require explicit data/code statement. |
+| §2.5 Limitations enumeration | item 23 (limitations) | item 41 (limitations) | item 27 (limitations) | item 19 (limitations) | item 11 (limitations) | Enumerate, do not narrate generally. |
+| §10.4 Challenge statement | (implicit in Background) | item 4 (rationale) | item 5 (rationale) | item 4 (rationale) | item 4 (rationale) | "Why this is hard" overlaps with rationale items. |
+
+## Workflow
+
+1. Run `/check-reporting` first — produces a PRESENT/PARTIAL/MISSING audit per guideline item.
+2. Run `/academic-aio` — produces the AIO PASS/PARTIAL/FAIL checklist.
+3. For each AIO FAIL or PARTIAL row, check this mapping. If the underlying reporting-guideline item is also MISSING/PARTIAL, fix it once and both audits update.
+4. Items present in `/check-reporting` audit but not in this mapping (e.g., randomization details for RCTs, domain-specific safety items) do not have an AIO consequence and can be addressed independently.
+
+## When the two audits disagree
+
+- AIO PASS + reporting-guideline MISSING — the manuscript looks discoverable but is not formally compliant. Reviewers may still reject. Always fix the reporting-guideline gap.
+- AIO FAIL + reporting-guideline PRESENT — a rule was recorded in compliance form but rendered in a way that LLM extractors cannot parse (e.g., reporting CIs in supplementary instead of inline in the abstract). Move the content into a chunk-friendly location.
+
+## Source
+
+- TRIPOD+AI: Collins et al. BMJ 2024.
+- CLAIM 2024: Tejani et al. Radiology: AI 2024, doi:10.1148/ryai.240300.
+- STARD-AI 2025: Sounderajah et al. Nat Med 2025, doi:10.1038/s41591-025-03953-8.
+- TRIPOD-LLM 2024: Gallifant et al. Nat Med 2024, doi:10.1038/s41591-024-03425-5.
+- DECIDE-AI 2022: Vasey et al. Nat Med 2022, doi:10.1038/s41591-022-01772-9.
+
+## Anti-hallucination
+
+Item numbers above are derived from the most recent published version of each guideline. Verify against the EQUATOR Network entry before citing item numbers in a manuscript — guideline updates renumber items.

package/skills/academic-aio/references/schema_markup_templates/CodeRepository.jsonld

@@ -0,0 +1,32 @@
+{
+  "@context": "https://schema.org",
+  "@type": "SoftwareSourceCode",
+  "name": "<repo name>",
+  "description": "<one-line description of what this repository contains>",
+  "codeRepository": "https://github.com/<org>/<repo>",
+  "programmingLanguage": ["Python", "R"],
+  "license": "https://spdx.org/licenses/Apache-2.0.html",
+  "datePublished": "YYYY-MM-DD",
+  "version": "v1.0.0",
+  "author": [
+    {
+      "@type": "Person",
+      "name": "<First Last>",
+      "identifier": "https://orcid.org/0000-0000-0000-0000"
+    }
+  ],
+  "isPartOf": {
+    "@type": "ScholarlyArticle",
+    "@id": "https://doi.org/10.xxxx/yyyy"
+  },
+  "identifier": {
+    "@type": "PropertyValue",
+    "propertyID": "DOI",
+    "value": "10.5281/zenodo.<id>"
+  },
+  "url": "https://doi.org/10.5281/zenodo.<id>",
+  "softwareRequirements": "Python 3.10+, PyTorch 2.0+",
+  "operatingSystem": ["Linux", "macOS"],
+  "applicationCategory": "Medical AI Research",
+  "keywords": ["<task>", "<modality>", "<model family>"]
+}