medsci-skills 4.1.0

package/skills/peer-review/references/reviewer_calibration/compliance_floor.md

@@ -0,0 +1,52 @@
+# Compliance floor — critical-item presence over headline percentage
+
+A reporting-compliance percentage is a coarse summary. Two manuscripts at the same
+percentage can differ enormously in reviewability, because **not all items carry equal
+weight**. Calibrate the reviewer comment to *which* items are missing, not just how many.
+
+## The principle
+
+1. **Critical items are pass/fail.** For each study type there is a small set of items that
+   a methods reviewer or editor treats as non-waivable. If one is MISSING, raise it as a
+   Major comment **regardless of the overall percentage** — an otherwise 90%-compliant paper
+   that cannot be reproduced or whose ground truth is undefined is not "90% acceptable."
+2. **The overall percentage is a secondary signal**, useful for "is the reporting broadly
+   thorough or broadly thin," not for a desk-reject line. Do not assert a numeric floor a
+   journal has not published.
+3. **The journal's own required elements are hard.** A missing structured-abstract heading,
+   a missing required panel (e.g., Key Points / Summary Statement), or the wrong reporting
+   checklist for the study type is a concrete, citable gap — but those facts live in
+   `reviewer_profiles/` and the author guidelines; verify there, never invent them.
+
+## Critical items by guideline (MISSING → Major / reject-risk)
+
+These are the items most consistently treated as non-waivable. Phrase the comment as the
+specific missing item, anchored to where it should appear.
+
+| Guideline (study type) | Critical items (raise as Major if MISSING) |
+|---|---|
+| **CLAIM** (AI in medical imaging) | Model architecture + training procedure; ground-truth/reference-standard definition; data-partition method (train/val/test, leakage control); evaluation metrics with uncertainty |
+| **TRIPOD+AI** (prediction model) | Source and types of input data; outcome definition and timing; how the model was developed and validated; performance with **calibration**, not discrimination only |
+| **STARD-AI / STARD** (diagnostic accuracy) | Reference standard and its rationale; blinding of index-test vs reference readers; participant flow (2×2 recoverable); handling of indeterminate results |
+| **STROBE** (observational) | Eligibility criteria and selection; outcome/exposure definitions; **how missing data were handled**; participant flow (cascade reconciles) |
+| **PRISMA / PRISMA-DTA** (systematic review) | Full search strategy (≥1 database verbatim); flow diagram that reconciles; per-study risk-of-bias; pre-registration/protocol reference |
+
+Extensions name the base instrument too (e.g., STARD **and** STARD-AI) — a manuscript that
+cites only the extension and skips the base items is itself an item to flag.
+
+## Desk-rejection-risk signals (weigh, don't over-assert)
+
+- A **critical item is MISSING** (above) — the strongest single signal.
+- The manuscript uses the **wrong reporting guideline** for its design (e.g., a prediction
+  model reported only against STROBE).
+- A journal **required element is absent** (per `reviewer_profiles/` + author guidelines).
+- The **reporting checklist is not uploaded** or references a stale manuscript version.
+
+When several co-occur, say so in the Confidential Comments to the Editor as a coherent
+"reporting readiness" concern, not as a list of nitpicks.
+
+## Integration
+
+Run `/check-reporting` first for the item-by-item table, then apply this file:
+- `/peer-review` Phase 2 (reporting-guideline check) and Phase 4 self-QC.
+- `/self-review` category G — flag MISSING critical items as Anticipated Major Comments.

package/skills/peer-review/references/reviewer_profiles/AJR.md

@@ -0,0 +1,82 @@
+# American Journal of Roentgenology (AJR) — Reviewer Profile
+
+## Submission System
+- **Platform:** Editorial Manager
+- **Review type:** **Double-blind** (author identity masked from reviewer; do not guess or reference author identity in comments)
+- **Publisher:** ARRS (American Roentgen Ray Society)
+
+## Required Structure (Comments to Author)
+
+Per the AJR Reviewer Guide template, organize the author-facing comments as:
+
+1. **Manuscript summary** — a few unnumbered sentences: purpose, essence of the design, key finding, ultimate conclusion.
+2. **General comments (including impact)** — itemized; strengths and weaknesses plus originality, impact on radiologists' practice, importance, advance-vs-replicate, citation potential.
+3. **Specific comments** — by manuscript section, with page and line numbers where appropriate.
+
+**Numbering:** consecutive across the *entire* review (general + specific together); do **not** restart per section. Favorable comments that need no author response are left unnumbered.
+
+**Section list for original research (IMRAD):** Title, Highlights, Abstract, Introduction, Methods, Results, Discussion, References, Tables, Figures.
+
+**Section list for Review / narrative articles (not IMRAD):** Title, Abstract, Introduction, thematic body sections (the manuscript's own headings), Conclusions, References, Tables, Figures (Methods/Results are N/A). Apply the Phase 2D 8-probe audit (RV1–RV8).
+
+Do **not** focus the review on typographic errors (handled by copyeditors). Do **not** state an accept/reject decision in the author-facing comments.
+
+## Scorecard
+
+No numeric scorecard. The portal exposes two free-text boxes (Comments to Author, Confidential Comments to Editor) plus reviewer CME self-assessment questions (Likert Q1–6, a confidentiality attestation, and an optional named-peer-reviewer credit).
+
+## Recommendation Options (dropdown, percentile-based)
+
+- No Recommendation
+- Accept
+- **Reconsider (for top 25%)**
+- **Reconsider (for top 26–50%)**
+- Reject (for bottom 50%)
+
+There is **no "Major/Minor Revision" option**. A revise-and-resubmit maps to "Reconsider"; a standard major-revision-grade assessment typically maps to **Reconsider (for top 26–50%)** (top 25% is a strong, top-quartile revise).
+
+## Confidential Comments to Editor
+
+Mandatory (must not be left blank). Provide a brief gestalt of merits, impact, and interest to AJR's general radiologist audience, and the fundamental reason for the reconsider/reject/accept recommendation. Do not simply repeat the author-facing comments. State the conflict-of-interest answer (None / explain) at the top.
+
+## Draft Structure Template
+
+```
+## COMMENTS TO THE AUTHORS
+
+Manuscript summary
+<purpose, design, key finding, conclusion>
+
+General comments (including impact)
+<unnumbered strengths>
+1) <impact / originality / weakness>
+
+Specific comments
+
+Abstract (p. X)
+2) ...
+
+Introduction (pp. X–Y)
+3) ...
+
+<body sections / IMRAD sections, continuing the numbering>
+
+References (pp. X–Y)
+N) ...
+
+Tables
+N+1) ...
+
+Figures
+N+2) ...
+
+## CONFIDENTIAL COMMENTS TO THE EDITOR
+Conflict of interest: None.
+<gestalt + recommendation rationale; do not repeat author comments>
+```
+
+## Journal Notes
+
+- **Double-blind** — author identity is masked; never disclose or infer reviewer or author identity in comments.
+- Consecutive numbering across general + specific comments aids editor cross-reference to the author response letter.
+- Recommendation language belongs only in the Confidential box and the dropdown, never in author comments.

package/skills/peer-review/references/reviewer_profiles/EURE.md

@@ -0,0 +1,64 @@
+# European Radiology (EURE) — Reviewer Profile
+
+## Submission System
+- **Platform:** Editorial Manager (Springer Nature)
+- **Review type:** Single-blind
+- **Publisher:** ESR (European Society of Radiology) / Springer Nature
+
+## Scorecard
+
+Follows INSI-style formatting as base (Springer Nature editorial system shared), adjusted per EURE specifics when confirmed from invitation email.
+
+Typical fields (confirm on each invitation):
+
+1. Interest / relevance to readership
+2. Novelty / originality
+3. Scientific quality
+4. Clarity of presentation
+5. Overall recommendation
+
+Rating scale: generally 5-point Likert or High/Medium/Low — **verify from the current invitation form**.
+
+## Recommendation Options
+
+- Accept
+- Minor Revision
+- Major Revision
+- Reject (transfer allowed)
+- Reject
+
+## Draft Structure Template
+
+Use INSI-style base; substitute journal name and scorecard fields when confirmed.
+
+## Journal Notes
+
+- ESR Key Points required in accepted manuscripts — reviewers can suggest strengthening these.
+- Transfer authorization common (to Springer Nature portfolio).
+
+## Confirmed Form Fields (from a recent EURE reviewer invitation)
+
+Editorial Manager scorecard differs from INSI H/M/L base. Actual fields:
+
+1. **Recommendation** (top dropdown): Accept / Minor Revision / Major Revision / Reject (transfer allowed) / Reject
+2. **Transfer Authorization** (2 questions): Yes/No for (a) identifying info transfer, (b) original review transfer. Default Yes/Yes.
+3. **ORCID Reviewer Credit**: Yes/No.
+4. **Level of interest** (4 options):
+   - An exceptional article
+   - An article of importance in its field
+   - An article whose findings are important to those with closely related research interests
+   - An article of limited interest
+5. **Quality of written English** (3 options):
+   - Not suitable for publication unless extensively edited
+   - Needs some language corrections before being published
+   - Acceptable
+6. **Declaration of competing interests**: free-text field, ≤300 chars. Standard "I declare that I have no competing interests." accepted.
+
+Author-facing comments follow the invitation's 4-section scheme:
+1. Comment on study design
+2. Short summary of the manuscript
+3. Major comments (strengths + weaknesses)
+4. Minor weaknesses (only if accept/revision)
+
+Markdown bold (`**...**`) renders as plain text in EM proof — use plain heading numbering instead of bold for sub-section titles if visual emphasis matters.
+

package/skills/peer-review/references/reviewer_profiles/INSI.md

@@ -0,0 +1,57 @@
+# Insights into Imaging (INSI) — Reviewer Profile
+
+## Submission System
+- **Platform:** Editorial Manager (Springer Nature)
+- **Review type:** Single-blind
+- **Publisher:** ESR (European Society of Radiology) / Springer Nature
+
+## Required Scorecard (4 items)
+
+Rating scale: High / Medium / Low
+
+1. Interest
+2. Innovation
+3. Importance
+4. Language editing — categorical. Editorial Manager dropdown labels (verified 2026-05-16 on INSI-D-26-00015R1 confirmation PDF): "No (minor)" / "Yes (minor)" / "Yes (major)". The profile previously listed these as "No minor editing needed" / "Yes minor editing needed" / "Yes major editing needed"; the portal uses the parenthesized short form. Semantics: "No (minor)" = best/most favorable (well written, only minor polish possible); "Yes (minor)" = minor editing needed; "Yes (major)" = major editing needed.
+
+## Additional Required Fields
+
+- Transfer Authorization questions (Yes/Yes) — allow transfer to other Springer Nature journals if rejected.
+- Conflict of interest declaration.
+
+## Recommendation Options
+
+- Accept
+- Minor Revision
+- Major Revision
+- Reject (transfer to another journal allowed)
+- Reject
+
+## Draft Structure Template
+
+```
+## INSI Reviewer Scorecard (Editorial Manager)
+
+| Item | Rating |
+|---|---|
+| Interest | High/Medium/Low |
+| Innovation | High/Medium/Low |
+| Importance | High/Medium/Low |
+| Language editing | No (minor) / Yes (minor) / Yes (major) |
+
+Transfer authorization: Yes / Yes
+Conflict of interest: None.
+
+---
+
+## CONFIDENTIAL COMMENTS TO THE EDITOR
+...
+
+## COMMENTS TO THE AUTHORS
+...
+```
+
+## Confidential Framing — Minor Revision
+
+When recommending Minor Revision, frame concerns as "labeling / clarificatory / reporting issue" vs "methodological flaw" and commit to "resolvable in a single revision cycle, no new analyses required". Helps the editor commit without protracted re-review.
+

package/skills/peer-review/references/reviewer_profiles/KJR.md

@@ -0,0 +1,100 @@
+# Korean Journal of Radiology (KJR) — Reviewer Profile
+
+## Submission System
+- **Platform:** ScholarOne (http://mc.manuscriptcentral.com/kjr)
+- **Review type:** Double-blind
+- **Two reviewers** assigned per manuscript
+
+## Required Scorecard (8 items)
+
+Auto-include in every KJR review draft as its own section, above Confidential Comments to the Editor.
+
+1. Reader interest
+2. Novelty
+3. Clinical impact
+4. Scientific impact
+5. Methodologic quality
+6. Educational or pictorial quality
+7. Clarity of organization, writing, and language
+8. Overall
+
+### Rating Scale (verbatim)
+
+`Not applicable / Excellent / Good / Average (for KJR) / Below Average / Poor`
+
+- **"Average (for KJR)"** is the calibration anchor — the journal's baseline quality bar.
+- Use Good/Excellent only when the manuscript clearly exceeds that bar.
+
+### M-Score Mapping
+
+Journal computes M-Score as mean of per-item ratings on a 0–4 scale:
+
+| Rating | Score |
+|---|---|
+| Excellent | 4 |
+| Good | 3 |
+| Average (for KJR) | 2 |
+| Below Average | 1 |
+| Poor | 0 |
+
+### Recommendation ↔ M-Score Alignment
+
+| Recommendation | M-Score range |
+|---|---|
+| Accept | ~3.5+ |
+| Minor Revision | ~2.5–3.0 |
+| Major Revision | ~1.5–2.5 |
+| Reject: Resubmission allowed | ~1.0–1.5 |
+| Reject | <1.0 |
+
+## Additional Required Fields
+
+- **Conflict of interest** (explicit "None" when applicable) — required form field.
+- **Recommendation options:** Accept / Minor Revision / Major Revision / Reject: Resubmission allowed / Reject.
+- **Web of Science recognition radio (`*` required):** Yes / No. Missing selection blocks submission; recommend Yes for peer-review credit accumulation.
+
+## Pre-submission portal verification (confirmation PDF)
+
+Before clicking submit, export confirmation PDF and verify against draft side by side:
+- [ ] All 8 scorecard items selected (no blank radios)
+- [ ] **Overall scorecard ↔ Recommendation alignment** — scorecard mean (M-Score) should land in the same band as the chosen Recommendation (see Recommendation ↔ M-Score table above). Mismatch is a reviewer credibility distractor for the editor.
+- [ ] Web of Science recognition radio explicitly Yes or No (not blank)
+- [ ] Confidential Comments to Editor body verbatim matches final draft
+- [ ] Comments to Author body verbatim matches final draft (incl. all Major and Minor numbering)
+- [ ] COI field has "none" or actual disclosure (not blank)
+- [ ] Recommendation radio matches final draft (not default Accept)
+
+## Draft Structure Template
+
+```
+## KJR Reviewer Scorecard (ScholarOne)
+
+Rating scale: Not applicable / Excellent / Good / Average (for KJR) / Below Average / Poor
+
+| Item | Rating | Rationale |
+|---|---|---|
+| Reader interest | ... | ... |
+| Novelty | ... | ... |
+| Clinical impact | ... | ... |
+| Scientific impact | ... | ... |
+| Methodologic quality | ... | ... |
+| Educational or pictorial quality | ... | ... |
+| Clarity of organization, writing, and language | ... | ... |
+| Overall | ... | ... |
+
+Conflict of interest: None.
+
+---
+
+## CONFIDENTIAL COMMENTS TO THE EDITOR
+...
+
+## COMMENTS TO THE AUTHORS
+...
+```
+
+## Journal Notes
+
+- Double-blind: ensure comments do not reveal reviewer identity/affiliation.
+- AI policy: reviewers/editors must not upload manuscript content to AI services unless confidentiality is reliably assured; core reviewer functions (generating review comments) must not be delegated to AI. Any non-linguistic AI assistance must be disclosed.
+

package/skills/peer-review/references/reviewer_profiles/README.md

@@ -0,0 +1,32 @@
+# Reviewer Profiles
+
+Canonical per-journal reviewer formatting profiles. Consumed by both the OSS `peer-review` skill (medsci-skills) and user-private peer-review skills (e.g., `~/.claude/skills/peer-review/`).
+
+## Contents
+
+| File | Journal | System | Scorecard |
+|---|---|---|---|
+| KJR.md | Korean Journal of Radiology | ScholarOne | 8 items, Excellent→Poor |
+| RYAI.md | Radiology: Artificial Intelligence | ScholarOne | 5 items, 1–9 |
+| INSI.md | Insights into Imaging | Editorial Manager | 4 items, H/M/L |
+| AJR.md | American Journal of Roentgenology | Editorial Manager | Section-by-section |
+| EURE.md | European Radiology | Editorial Manager | INSI-style base |
+
+## Design Principles
+
+1. **Single source of truth** — each journal's reviewer-form fields live in exactly one file here.
+2. **Form fields, not opinions** — profiles describe what the editorial system expects (fields, scales, options); subjective calibration lives in the reviewer's own guideline.
+3. **OSS-safe** — no PII, no specific reviewer identity, no manuscript content, **no manuscript IDs**, **no specific editor names**, no topic-level hints that could identify a past review. Under COPE reviewer confidentiality obligations, publishing the set of manuscripts a reviewer has handled can itself identify the reviewer. Keep personal precedent logs in a private store (e.g., `~/.claude/skills/peer-review/`) — never commit them here.
+4. **Parallel to find-journal / write-paper profiles** — same directory-of-markdown pattern used elsewhere in medsci-skills.
+
+## Adding a New Journal
+
+1. Copy the closest existing profile as a template.
+2. Record form fields from the first real review invitation (scorecard items, rating scales, required text boxes, recommendation options).
+3. Commit under `{JOURNAL_SHORTNAME}.md` using established abbreviations (KJR, RYAI, INSI, AJR, EURE; full name if no common abbreviation).
+4. Update this README table.
+
+## Consumed By
+
+- `medsci-skills/skills/peer-review/SKILL.md` — OSS public skill.
+- `~/.claude/skills/peer-review/SKILL.md` — user's private overlay (reads same profiles, adds PII-specific guideline path).

package/skills/peer-review/references/reviewer_profiles/RYAI.md

@@ -0,0 +1,86 @@
+# Radiology: Artificial Intelligence (RYAI) — Reviewer Profile
+
+## Submission System
+- **Platform:** ScholarOne
+- **Review type:** Single-blind
+
+## Required Scorecard (5 items)
+
+Rating scale: 1–9 integer (**1 = excellent / Accept, 9 = poor / Reject**; lower is better — confirmed by ScholarOne form labels "Accept (1-3) / Uncertain (4-6) / Reject (7-9)")
+
+1. Scientific Merit
+2. Originality / Importance
+3. Practical Value
+4. Reader Interest
+5. Overall
+
+## Additional Required Fields
+
+The ScholarOne RYAI form has the following fields, **each separate**. Checking their consistency with the review tone is mandatory:
+
+- **Newsworthy** (Yes / No / Unsure radio)
+- **Immediate value for daily practice** (Yes / No / Unsure radio) — a field separate from Newsworthy. Both must be answered.
+- **Statistical review needed** (Yes / No) — location uncertain (sometimes not visible on the confirmation PDF; possibly on a different portal step)
+- **Physician / Non-physician** (radio)
+- **CME credit** (Yes / No)
+- **Web of Science recognition** (Yes / No, required) — submission is blocked if omitted
+- **RSNA Confidentiality and AI Policy checkbox** (required) — submission is blocked if omitted
+- **Specify Journal** (only when "Reject: Another Journal" is selected)
+
+### Consistency rule
+
+**Newsworthy / Daily-practice / Practical Value (1–9) / Major-comment tone — these four must be mutually consistent.**
+
+| Signal combination | Consistency |
+|---|---|
+| Practical Value 1–3 + Newsworthy Yes + Daily-practice Yes | OK (consistently positive) |
+| Practical Value 4–6 + Newsworthy Unsure + Daily-practice Unsure | OK (consistently neutral) |
+| Practical Value 7–9 + Newsworthy No + Daily-practice No | OK (consistently negative) |
+| Practical Value 6 + Newsworthy Unsure + Daily-practice **Yes** | **MISMATCH** — set daily-practice to No or Unsure |
+| Major comments criticizing actionability + Daily-practice Yes | **MISMATCH** |
+
+The Daily-practice field is easily left as Yes by a portal default selection or user error. Check it separately during submission verification.
+
+### Required checkbox checklist (just before submission)
+
+- [ ] RSNA Confidentiality and AI Policy checkbox ✓
+- [ ] Web of Science recognition Yes/No selected
+- [ ] Physician/Non-physician selected
+- [ ] CME credit selected
+- [ ] Recommendation radio selected
+- [ ] Newsworthy + Daily-practice both selected (separate fields)
+- [ ] Statistical review needed — confirm its location, then select (if present on the form)
+
+## Recommendation Options
+
+- Accept
+- Accept with Minor Revision
+- Reject: Encourage major revision and resubmission (**= Major Revision** for practical purposes)
+- Reject
+
+## Draft Structure Template
+
+```
+## RYAI Reviewer Scorecard (ScholarOne)
+
+| Item | Score (1–9) | Rationale |
+|---|---|---|
+| Scientific Merit | ... | ... |
+| Originality / Importance | ... | ... |
+| Practical Value | ... | ... |
+| Reader Interest | ... | ... |
+| Overall | ... | ... |
+
+Newsworthy / daily-practice value: ...
+Statistical review needed: Yes/No
+Conflict of interest: None.
+
+---
+
+## CONFIDENTIAL COMMENTS TO THE EDITOR
+...
+
+## COMMENTS TO THE AUTHORS
+...
+```
+

package/skills/peer-review/skill.yml

@@ -0,0 +1,39 @@
+schema_version: 2
+name: peer-review
+layer: C
+owner_domain: external_peer_review
+when_to_use:
+  - User received a journal review invitation and needs a structured review draft
+  - Drafting reviewer comments for a specific journal's review template (RYAI / INSI / EURE / AJR / KJR)
+  - Constructive developmental feedback on an external manuscript PDF
+when_NOT_to_use:
+  - Reviewing the user's own manuscript (use /self-review)
+  - Drafting the user's manuscript (use /write-paper)
+  - Generating references from memory or modifying the external paper (forbidden)
+inputs:
+  - external_manuscript.pdf
+outputs:
+  - review/{manuscript_id}/{manuscript_id}_review_draft.md
+  - review/{manuscript_id}/{manuscript_id}_submission.md
+deterministic_scripts:
+  - none_required
+side_effects:
+  - writes_review_artifacts
+downstream_consumers:
+  - none
+forbidden_actions:
+  - edit_users_own_manuscript
+  - generate_references_from_memory
+
+# v2.1 quality card
+purpose: "Draft a structured peer-review for a journal-assigned manuscript following the MedSci peer-review guideline (v2.5), with journal-specific formatting."
+safety_boundaries:
+  - "Reviews assigned external manuscripts only; never edits the user's own manuscript."
+  - "References are not generated from memory."
+known_limitations:
+  - "A review is one reviewer's judgement, not an editorial decision."
+  - "No standalone demo; quality depends on the manuscript supplied."
+validation_commands:
+  - "confirm the draft addresses each section against the guideline rubric"
+  - "python3 scripts/check_domain_probe_sync.py --strict"
+evidence_surface: manual_workflow