medsci-skills 5.17.0 → 5.18.0

package/README.md

@@ -11,6 +11,7 @@
 [![CI](https://img.shields.io/github/actions/workflow/status/Aperivue/medsci-skills/validate.yml?branch=main&style=flat-square&label=CI)](https://github.com/Aperivue/medsci-skills/actions/workflows/validate.yml)
 ![Skills](https://img.shields.io/badge/Skills-55-brightgreen?style=flat-square)
 [![npm](https://img.shields.io/npm/v/medsci-skills?style=flat-square&label=npm&color=cb3837)](https://www.npmjs.com/package/medsci-skills)
+[![npm downloads](https://img.shields.io/npm/dw/medsci-skills?style=flat-square&label=npm%20downloads&color=cb3837)](https://www.npmjs.com/package/medsci-skills)
 [![Watch the 2-min intro](https://img.shields.io/badge/▶_Watch-2--min_intro-FF0000?style=flat-square&logo=youtube&logoColor=white)](https://youtu.be/MclQ_RIofpE)
 [![good first issues](https://img.shields.io/github/issues/Aperivue/medsci-skills/good%20first%20issue?style=flat-square&label=good%20first%20issues&color=7057ff)](https://github.com/Aperivue/medsci-skills/contribute)
 
@@ -41,8 +42,9 @@
 
 ## What is MedSci Skills?
 
-MedSci Skills is an open-source Claude Code skill collection for **clinical
-research — the manuscript and the medical-AI model alike**. It helps
+MedSci Skills is an open-source **Agent Skills** collection for **clinical
+research — the manuscript and the medical-AI model alike** — designed to be driven
+directly by AI coding agents (Claude Code, Codex, Cursor, and GitHub Copilot). It helps
 physician-researchers and biomedical/medical-engineering investigators move from
 literature search, study design, statistics, and figures to reporting-guideline
 compliance, citation/reference auditing, numerical-consistency checks, and
@@ -155,15 +157,16 @@ protocol scaffolding, screening/extraction structure, PRISMA-consistent counts a
 diagram, pooled-estimate figures, and a manuscript draft. *Safety:* screening and
 extraction decisions stay with the human review team.
 
-## Live Demos: Three Study Types, Three Full Pipelines
+## Live Demos: Four Study Types, Four Full Pipelines
 
-Three public datasets. Three study types. Each produces a complete manuscript, publication-ready figures, and a reporting compliance audit.
+Four public datasets. Four study types. Demos 1–3 each produce a complete manuscript, publication-ready figures, and a reporting-compliance audit; Demo 4 runs the medical-AI **model-engineering lane** end to end (scaffold → gates → training → evaluation → interpretability).
 
 | Demo | Dataset | Study Type | Compliance |
 |------|---------|------------|------------|
 | [Demo 1: Wisconsin BC](demo/01_wisconsin_bc/) | `sklearn` built-in | Diagnostic accuracy | STARD 2015 |
 | [Demo 2: BCG Vaccine](demo/02_metafor_bcg/) | `metafor::dat.bcg` (13 RCTs) | Meta-analysis | PRISMA 2020 |
 | [Demo 3: NHANES Obesity](demo/03_nhanes_obesity/) | CDC NHANES 2017-18 | Epidemiology (survey) | STROBE |
+| [Demo 4: PneumoniaMNIST CNN](demo/04_pneumoniamnist_cnn/) | `medmnist` (CC BY 4.0) | Medical-AI model engineering (CNN) | CLAIM / TRIPOD+AI |
 
 ### Demo 1: Diagnostic Accuracy — Wisconsin Breast Cancer
 
@@ -285,6 +288,8 @@ The E2E pipeline (`orchestrate --e2e`) produces everything up to `qc/`. The `sub
 
 ## What's New
 
+**v5.18** — reliability & workflow-integrity batch. A new deterministic **response-claim gate** for `/revise` and `/peer-review` (verifies a response letter's "we added / we now cite X" against the *revised manuscript body* — a claimed-but-absent edit is caught before it reaches a reviewer; **detectors 50 → 51**), a **reframe / headline-change survivor scan** (`--retired-term` / `--old-value`, finds stale framing or superseded values left in the body, supplement, and figure legends after a reframe), a pre-drafting **backbone full-text readiness gate** for `/write-paper` (the backbone article is *used*, not merely detected), a **skill-registry consistency validator** (`capabilities.yml` ⇄ `skill.yml`, CI-enforced), AI-tool **citation-framing** guidance for `/academic-aio`, and **Demo 4** (PneumoniaMNIST CNN, the model-engineering lane end to end). Additive and backward-compatible; **55 skills / 44 guidelines / 51 integrity detectors / 22 domain-probe modules**. (See the [CHANGELOG](CHANGELOG.md) for v5.0–v5.17.)
+
 **v5.17** — model-engineering produce-side depth, **completion**. Deployment safety + the final roadmap/candidate items: a new **`uncertainty-imaging`** skill + `check_uncertainty_reporting` gate (calibrated per-case uncertainty / OOD guard on a held-out set / abstention at a pre-specified target / calibration-under-shift — for a deployment-framed model), an **MLOps wiring reference** for `model-scaffold` (experiment tracking, config/data/environment versioning, CI-for-ML — pointing to W&B / MLflow / nnU-Net, reimplementing nothing), and an **`architecture-zoo` graph-neural-net family card** (GCN / GraphSAGE / GAT / GIN / BrainGNN for brain connectomes) that closes the last candidate gap. The six-item produce-side depth roadmap and its candidate list are now complete. Additive and backward-compatible; **55 skills / 44 guidelines / 50 integrity detectors / 22 domain-probe modules**. (See the [CHANGELOG](CHANGELOG.md) for v5.0–v5.16.)
 
 **v5.16** — model-engineering produce-side depth, clinical fine-tuning focus (Items 3–4 of the [produce-side depth roadmap](docs/roadmap_model_engineering_depth.md)). A new **`radiomics-ml`** skill + `check_radiomics_ml` gate for the most common solo-doable clinical-ML workflow — radiomics / tabular features → any classical learner (LASSO / SVM / random forest / XGBoost / …) → a clinical outcome, no GPU — with a learner-agnostic nested-CV / calibration / stability gate; plus a **`model-scaffold` fine-tuning mode** (`--task finetune` + `--from-pretrained`) that adapts a pretrained backbone on collected clinical data with a frozen→unfrozen schedule, discriminative learning rates, and a pretrained-weight provenance record (`PRETRAINED.md` + a `PRETRAINED_PROVENANCE_MISSING` verdict on the existing `check_training_hygiene`, plus a MedSAM-adaptation + train-only diffusion-augmentation guide). Additive and backward-compatible; **54 skills / 44 guidelines / 49 integrity detectors / 22 domain-probe modules**. (See the [CHANGELOG](CHANGELOG.md) for v5.0–v5.15.)
@@ -647,7 +652,7 @@ matrices without CSV-coordinate comments are flagged as structural risks even wh
 are currently correct, since the next revision will re-introduce the same failure mode.
 
 ### Reference Safety (Phase 1)
-Projects declare their source-of-truth layout in `SSOT.yaml`, and a `qc/migration_complete` marker gates strict enforcement. `/verify-refs` is the sole writer of `qc/reference_audit.json`. The `MEDSCI_VERIFY_REFS_MODE` env var (`auto` default, `warn`, `enforce`, `off`) controls behavior — `auto` blocks only when both SSOT.yaml and the migration marker are present, otherwise warns. Legacy projects freeze as warn-only; new projects opt in via `scripts/migrate_project_to_ssot.py`. An optional PostToolUse hook (not shipped in this repo — document only) can invoke `/verify-refs` automatically on manuscript saves for users who install it locally at `~/.claude/hooks/verify-refs-guard.sh`; the regression suite (`tests/test_phase1c_hooks.sh`) runs end-to-end only when that local hook is present and is skipped otherwise.
+Projects declare their source-of-truth layout in `SSOT.yaml`, and a `qc/migration_complete` marker gates strict enforcement. `/verify-refs` is the sole writer of `qc/reference_audit.json`. The `MEDSCI_VERIFY_REFS_MODE` env var (`auto` default, `warn`, `enforce`, `off`) controls behavior — `auto` blocks only when both SSOT.yaml and the migration marker are present, otherwise warns. Legacy projects freeze as warn-only; new projects opt in via `scripts/migrate_project_to_ssot.py`. An optional PostToolUse hook (not shipped in this repo — document only) can invoke `/verify-refs` automatically on manuscript saves for users who install it locally at `~/.claude/hooks/verify-refs-guard.sh`; the regression suite (`tests/test_phase1c_hooks.sh`) runs end-to-end only when that local hook is present and is skipped otherwise. The hook gates saves under `*/submission/*/manuscript/*.{docx,md}` and `*/revision/R*/…circulation….docx` (enforce-eligible), and — added for issue #14 — also **warn-only** on pre-submission and mentor-circulation drafts that previously skipped the audit entirely: `*/outgoing/*.{docx,md}`, `*/8_Review_Comments/*/outgoing/*.{docx,md}`, and any `*/circulation/*.{docx,md}`. Warn-only patterns surface a missing audit without blocking rapid iteration and never enforce, regardless of SSOT/migration state.
 
 ### Meta-Analysis Failure Modes
 `/meta-analysis` ships empirical failure-mode references (data integrity, review orchestration, submission package drift, post-submission release ops) with four automation hooks: `scripts/prisma_5way_consistency.py` (DI-6 PRISMA number consistency), `scripts/extraction_consensus_log_init.py` (DI-1 dual-extraction scaffold), `scripts/tag_cleanup_gate.sh` (DI-8 placeholder tag gate), and `scripts/verify_package_integrity.py` (SPD SHA-256 manifest for submission bundles).
@@ -782,12 +787,18 @@ from one of these:
 
 | Want to add… | How | Issue |
 |---|---|---|
-| **A journal profile** (submission rules for a journal we don't cover) | `/add-journal`, or copy an existing `journal_profiles/*.md` | [#115](https://github.com/Aperivue/medsci-skills/issues/115) |
+| **A CSL citation style** for a journal that lacks one | drop one `.csl` into `manage-refs/citation_styles/` | [#117](https://github.com/Aperivue/medsci-skills/issues/117) |
+| **A de-identification locale pack** for one more country | add one patterns file under `deidentify/` | [#116](https://github.com/Aperivue/medsci-skills/issues/116) |
+| **A README translation** (e.g., zh-CN) | one translated `README` file | [#119](https://github.com/Aperivue/medsci-skills/issues/119) |
 | **A figure exemplar** (ROC, KM, forest, Bland–Altman, confusion matrix…) | one `make-figures/references/exemplar_plots/*.md` anatomy model | [#118](https://github.com/Aperivue/medsci-skills/issues/118) |
-| **A CSL citation style** for a journal that lacks one | drop a `.csl` into `manage-refs/citation_styles/` | [#117](https://github.com/Aperivue/medsci-skills/issues/117) |
-| **A de-identification locale pack** for one more country | add patterns to `deidentify/` | [#116](https://github.com/Aperivue/medsci-skills/issues/116) |
+| **A journal profile** (submission rules for a journal we don't cover) | `/add-journal`, or copy an existing `journal_profiles/*.md` | [#115](https://github.com/Aperivue/medsci-skills/issues/115) |
 | **A reporting checklist or peer-review exemplar** | one reference file in the matching skill | [#120](https://github.com/Aperivue/medsci-skills/issues/120) |
-| **A README translation** (e.g., zh-CN) | a translated `README` | [#119](https://github.com/Aperivue/medsci-skills/issues/119) |
+
+Each of these adds **exactly one new file** — none requires editing a count, a generated
+page, or the build config, and the catalog-consistency check auto-derives profile counts from
+disk so it won't flag you. **No assignment needed: just open a PR — the first one that passes
+CI wins, and a maintainer handles any bookkeeping in review.** See the
+[5-minute first-PR quickstart](CONTRIBUTING.md#quickstart-your-first-pr-5-minutes) to get going.
 
 Every contribution is gated the same way the maintainers are: it must be a self-contained
 file, pass the CI (`validate.yml` — PII scan, structure, catalog consistency), and carry no
@@ -801,6 +812,8 @@ or a [detector request](https://github.com/Aperivue/medsci-skills/issues/new?tem
 [`docs/maintainer_workflow.md`](docs/maintainer_workflow.md) (review + release process),
 and [`SECURITY.md`](SECURITY.md) (vulnerability reporting + the medical-scope boundary).
 A change that touches a medical/research claim needs Clinical-Lead review.
+Community members who have reported, verified, or shaped the project are credited in
+[`CONTRIBUTORS.md`](CONTRIBUTORS.md).
 
 ## In the Wild
 
@@ -836,9 +849,9 @@ software citation is how a tool like this earns academic recognition, and it tak
 }
 
 @article{nam2026agentic,
-  author  = {Nam, Yoojin and Kim, Namkug},
-  title   = {{Agentic Skills for Auditable and Reproducible Medical Research Writing:
-             An Integrity-gated Architecture for LLM-Assisted Clinical Manuscripts}},
+  author  = {Nam, Yoojin and Jeong, Jinhoon and Kim, Namkug},
+  title   = {{Deterministic Integrity Gates for LLM-Assisted Clinical Manuscript
+             Preparation: An Auditable Biomedical Informatics Architecture}},
   year    = {2026},
   journal = {arXiv preprint arXiv:2606.09500},
   url     = {https://arxiv.org/abs/2606.09500}

package/metadata/distribution_files.json

@@ -58,8 +58,13 @@
     },
     {
       "path": "skills/academic-aio/SKILL.md",
-      "size": 31396,
-      "sha256": "339f937d2218e11c8f6576df9bb6d5b2b827634092d8f4e6c1b6110e092076b5"
+      "size": 32276,
+      "sha256": "d893714e70af45b6f9d162d5220875685995e4a2c428331348689566e672989c"
+    },
+    {
+      "path": "skills/academic-aio/references/ai_tool_citation_framing.md",
+      "size": 4064,
+      "sha256": "e3c8403cc698a45bb5626b233c5cbf18eb283e6a606d6232471f0f0c657d0fec"
     },
     {
       "path": "skills/academic-aio/references/case_studies/kjr_mllm_2025.md",
@@ -1033,8 +1038,8 @@
     },
     {
       "path": "skills/design-study/SKILL.md",
-      "size": 19220,
-      "sha256": "d4a05f6dd2530620bd1ec3e1c64ba0625aae25def12892247e61a77724132cd6"
+      "size": 20296,
+      "sha256": "98644b7750b65ff9bb38af7cb6e0b3300169f498332a3f143394cc35f5ba7ff7"
     },
     {
       "path": "skills/design-study/references/dag_adjustment.md",
@@ -2268,8 +2273,8 @@
     },
     {
       "path": "skills/make-figures/skill.yml",
-      "size": 2205,
-      "sha256": "b973c376998b15cefe9bffdf580889f905b60ddc2dead00cc5cbc900a6fd50f5"
+      "size": 2249,
+      "sha256": "6a8ebbe077c2073bb546a044bbecb5d483dbe2d2a50b31adabba6cf5e3a87990"
     },
     {
       "path": "skills/make-figures/templates/official/NOTES.md",
@@ -2373,8 +2378,8 @@
     },
     {
       "path": "skills/manage-refs/SKILL.md",
-      "size": 18165,
-      "sha256": "49adc82dea2b5d7eb93946b2cd8143d66d50b53169d3ed18f4bd738bfe3af39f"
+      "size": 18627,
+      "sha256": "abb0f38acbff93ffdff68be6a29b68757027fd7ffec545abd90df2df4113c08c"
     },
     {
       "path": "skills/manage-refs/citation_styles/README.md",
@@ -2718,13 +2723,13 @@
     },
     {
       "path": "skills/mllm-eval/references/evaluation_axes.md",
-      "size": 10857,
-      "sha256": "49d77ab63feae5dba5cdae7e47d9de6ea97ee3b9eea4b9cba39ec8b0e185be72"
+      "size": 11609,
+      "sha256": "0bad98d58685c638132c891c7c4d8b3c4620fe0988994d3307458382167d3d92"
     },
     {
       "path": "skills/mllm-eval/scripts/check_mllm_eval_completeness.py",
-      "size": 9380,
-      "sha256": "fe24a9b0dfcce7d826c29be59f2534c768daf73e1e98f545e8a7791bc0709079"
+      "size": 9386,
+      "sha256": "bb714a45449a26ccafca9e988a95aaa6af92fb8f3b79cfc03f8b47768f37aaec"
     },
     {
       "path": "skills/mllm-eval/scripts/mllm_eval_completeness_challenge/fixture/plan_bad.md",
@@ -2808,23 +2813,23 @@
     },
     {
       "path": "skills/model-evaluation/SKILL.md",
-      "size": 5780,
-      "sha256": "334b4ca87a2f672446563f3fb6d78c0585386fac12aee29e10dc09465e892193"
+      "size": 7380,
+      "sha256": "b39293e54402fa79ac73de827201f4cd98bb2cc70fd1cd4fc8da4e940d429b3b"
     },
     {
       "path": "skills/model-evaluation/references/metric_guide.md",
-      "size": 2454,
-      "sha256": "8d09ca7ce9fb9f66ee4942689294d9b12ae1d892ac67769cd68fdc38a4e220ee"
+      "size": 6194,
+      "sha256": "b8cb061cf40b0736961a4f06eb7f6b60f238a0e19f7b47e07f3cb38a82097f52"
     },
     {
       "path": "skills/model-evaluation/references/metric_selection_grounding.md",
-      "size": 9588,
-      "sha256": "56723e73b2b74299140d353921d1dba471bedf9aea0b8e769fc69995c4733995"
+      "size": 10256,
+      "sha256": "535282ae5cfbf2593c3e40926143e2ebd081ebbd78ac972cb0e6aeae839d1e77"
     },
     {
       "path": "skills/model-evaluation/scripts/check_metric_reporting.py",
-      "size": 9562,
-      "sha256": "fd6c0205f652651a5a43910cb415ed8442100fc553ce04aa5d5905d97743a0bf"
+      "size": 17964,
+      "sha256": "f1315745ceba694cb2437a412618d86cd7f0a7f4e652f5ced2cc207b67eaa731"
     },
     {
       "path": "skills/model-evaluation/scripts/metric_reporting_challenge/fixture/clf_bad.md",
@@ -2846,6 +2851,31 @@
       "size": 224,
       "sha256": "08227fbe46829b8eecdfe15680e1bb32087f1e6c5a353b3df6055c67d8300fc0"
     },
+    {
+      "path": "skills/model-evaluation/scripts/metric_reporting_challenge/fixture/generative_bad.md",
+      "size": 134,
+      "sha256": "f4ac7e4d2d6dd12125e832a55363579181546d68411f49984680f334db5c8cb4"
+    },
+    {
+      "path": "skills/model-evaluation/scripts/metric_reporting_challenge/fixture/generative_good.md",
+      "size": 437,
+      "sha256": "d218437b09d8d80fa13aa52a03c65fd6156d95e99204762cafaa7021c4b4b21f"
+    },
+    {
+      "path": "skills/model-evaluation/scripts/metric_reporting_challenge/fixture/interactive_bad.md",
+      "size": 128,
+      "sha256": "be714934c18566a89c6c7b06951a420d8e2cbfca814e1daf26df87866252a185"
+    },
+    {
+      "path": "skills/model-evaluation/scripts/metric_reporting_challenge/fixture/interactive_good.md",
+      "size": 500,
+      "sha256": "a45e14485fb02f0ee790350235c652ea2b871aa16cfd89e12a32e49cd6e1b9ca"
+    },
+    {
+      "path": "skills/model-evaluation/scripts/metric_reporting_challenge/fixture/multiclass_bad.md",
+      "size": 128,
+      "sha256": "0ab44d8dcdf4ec8281c7afb7db021e61bb368f2f30bd5ec49ed5bcf962319421"
+    },
     {
       "path": "skills/model-evaluation/scripts/metric_reporting_challenge/fixture/seg_bad.md",
       "size": 119,
@@ -2863,13 +2893,13 @@
     },
     {
       "path": "skills/model-evaluation/scripts/metric_reporting_challenge/verify.sh",
-      "size": 1663,
-      "sha256": "4aa6db49d3552f01a54ddb70484df5d214806d805b836b24beda81d9aca32bd6"
+      "size": 2642,
+      "sha256": "3efa9aff1f4ed223b7114740cac9cb67801456ffbfa3e5f7a7f1df7f46ada565"
     },
     {
       "path": "skills/model-evaluation/skill.yml",
-      "size": 2921,
-      "sha256": "9713dbab40c54ca88324e7dcd74d3142890f9625ad68d708ba40b0cf57a7b9ee"
+      "size": 2944,
+      "sha256": "d055072bcfb86d84b90f0b29790a262a496a95e4fe8ffbf38d02a28bf9db258e"
     },
     {
       "path": "skills/model-scaffold/SKILL.md",
@@ -2928,8 +2958,8 @@
     },
     {
       "path": "skills/model-validation/SKILL.md",
-      "size": 10836,
-      "sha256": "5cfd4ac6d75b3a474408b10d8eead3fdf7be8a44e1578dcac332cc2203ae75fe"
+      "size": 11524,
+      "sha256": "ac326c03451282a5adeeb3592f511dac47720bba9338dba761bf6ea4679d7c0b"
     },
     {
       "path": "skills/model-validation/references/validation_design.md",
@@ -3008,8 +3038,8 @@
     },
     {
       "path": "skills/peer-review/SKILL.md",
-      "size": 69384,
-      "sha256": "94033b84b41304d7dbd7919c84c7e9502bdaff1b7e2ef8f5dd7ec9cd7147c532"
+      "size": 70338,
+      "sha256": "246beb8a3c3c6eef78180a6ab1d87fa49db4f98d55964af5b3ab0ce5d0e7e3b0"
     },
     {
       "path": "skills/peer-review/references/aczel_2021_reviewer2_patterns.md",
@@ -3478,8 +3508,8 @@
     },
     {
       "path": "skills/render-pdf-doc/SKILL.md",
-      "size": 7490,
-      "sha256": "2de13f525ab617dc3a5d67403f8668c26829d927732ae2ac718158b109861243"
+      "size": 8300,
+      "sha256": "fad21da8e60907c212b6d216ffe591502941c8defa4ce212b521306e00332836"
     },
     {
       "path": "skills/render-pdf-doc/references/known_pitfalls.md",
@@ -3493,8 +3523,8 @@
     },
     {
       "path": "skills/render-pdf-doc/scripts/check_deps.sh",
-      "size": 984,
-      "sha256": "84d5b4d3455a66f21baaa2a3be1f01b9cba58c9e50331138c7132d23531f211d"
+      "size": 2183,
+      "sha256": "782efdda541c0a84a8d978bb076d43a1d3a026a00b878205dd6d922a197186f6"
     },
     {
       "path": "skills/render-pdf-doc/scripts/infer_colwidths.py",
@@ -3503,8 +3533,8 @@
     },
     {
       "path": "skills/render-pdf-doc/scripts/render_pdf.sh",
-      "size": 2874,
-      "sha256": "aad89324341121270be0093a05ff944927696180a27a97068f2faf906321e987"
+      "size": 3958,
+      "sha256": "b5a9ebcbb25b1ec85577c9864f60f4f363e90f88896903cff48f7626f5b6b95a"
     },
     {
       "path": "skills/render-pdf-doc/scripts/scan_glyph_coverage.py",
@@ -3513,8 +3543,8 @@
     },
     {
       "path": "skills/render-pdf-doc/skill.yml",
-      "size": 3131,
-      "sha256": "af400eb4f819fd982dc0df71ed6ad8b003987a164fbff1cf467c9d1d3c0c0a6c"
+      "size": 3133,
+      "sha256": "77fe112e5a013d3169290132f9882fbb05b3ebe9317729960582895509c21568"
     },
     {
       "path": "skills/render-pdf-doc/templates/anchor-doc.md",
@@ -3598,18 +3628,23 @@
     },
     {
       "path": "skills/revise/SKILL.md",
-      "size": 27775,
-      "sha256": "2da4f80e879c2d3ff31d2af435cb27ecae0ba09f1014b8ebbd799ac2472ff1ea"
+      "size": 28904,
+      "sha256": "c1851293ea0f59c658a1a5b4b303667de3e55f693ee5fddc2148d7245bab60ea"
     },
     {
       "path": "skills/revise/references/r2r_voice.md",
       "size": 17042,
       "sha256": "d6e657ab4d864e0c12c6d3c8e7db5060dfcea0c56d0f9afccd876702f8b8adcf"
     },
+    {
+      "path": "skills/revise/scripts/check_response_claims.py",
+      "size": 9798,
+      "sha256": "ef5e67cf757e146b43c578753fcafd10632676b08eec00328eb5100e20dc79b6"
+    },
     {
       "path": "skills/revise/skill.yml",
-      "size": 1873,
-      "sha256": "972bcb3f09675ab830f27dd447948059d7de583a7e7317501be7212f72d82683"
+      "size": 2261,
+      "sha256": "421670bd78193e01966d08d08b22f044a6c396cba0bed99ae50fbb8f6c6564a5"
     },
     {
       "path": "skills/search-lit/SKILL.md",
@@ -3953,8 +3988,8 @@
     },
     {
       "path": "skills/sync-submission/SKILL.md",
-      "size": 27787,
-      "sha256": "4da14c76c6c9326d31ee93e9515854291cba2c48692eb85cf5d9f6f4301ce465"
+      "size": 28938,
+      "sha256": "deb2d8b9d9939e84f2a0eaf57584b900b62eadf1d398ea60a49e3720d7ab174b"
     },
     {
       "path": "skills/sync-submission/references/journal_availability_policy.json",
@@ -3993,8 +4028,8 @@
     },
     {
       "path": "skills/sync-submission/scripts/check_cross_artifact_stale.py",
-      "size": 8286,
-      "sha256": "8316f2779f2cd523b9316736396074e5c06ea8270f7cb29286b4511e74a6fc1e"
+      "size": 11679,
+      "sha256": "613cdbab508fa2767e7b445624829d645c3a917e8f3d58a72dd66bcac7fd0586"
     },
     {
       "path": "skills/sync-submission/scripts/check_disclosure_availability.py",
@@ -4098,8 +4133,8 @@
     },
     {
       "path": "skills/verify-refs/references/manual_checkpoint_guide.md",
-      "size": 3727,
-      "sha256": "86832ad71dbeaf8287258d0ccab6e32ffd75e9fa5787b9240703c8a951b18d4e"
+      "size": 4031,
+      "sha256": "4da6bd649d672bcae7505098d4712d6fb253fea8fa3be41c95870587cc6553a4"
     },
     {
       "path": "skills/verify-refs/scripts/verify_cli.sh",
@@ -4108,13 +4143,13 @@
     },
     {
       "path": "skills/verify-refs/scripts/verify_refs.py",
-      "size": 41415,
-      "sha256": "4377e25e6b125ace31fd35df9ffdd21a413f0f70ba1e429e5bfcf50c0a29510e"
+      "size": 43072,
+      "sha256": "ce0db6183942afdb2dfd55733c8b1330dee1b10ce091ce99c31497f83d9c4e0b"
     },
     {
       "path": "skills/verify-refs/skill.yml",
-      "size": 2248,
-      "sha256": "3933c313777557ad8ac82b9ed6389fba12675880588af85193791f3facc636e9"
+      "size": 2291,
+      "sha256": "608e80234418a3560c9662e83a786b20ad16a340b934bfea59e2421712595fc5"
     },
     {
       "path": "skills/version-dataset/SKILL.md",
@@ -4138,8 +4173,8 @@
     },
     {
       "path": "skills/write-paper/SKILL.md",
-      "size": 67054,
-      "sha256": "62b3cbff56a139ed20a783c462fb6dde6b7a932bb057762a2d1c6382c8d044c1"
+      "size": 68059,
+      "sha256": "d05e1bbde228a87abf459ec19484e560dc08b7e88213e36655b680ebecf7dd8f"
     },
     {
       "path": "skills/write-paper/references/exemplar_abstract.md",
@@ -4631,10 +4666,15 @@
       "size": 7437,
       "sha256": "6bc4032302265ba2c834212e77fbe78b8ddeddba716c833ecf8c65e85e586320"
     },
+    {
+      "path": "skills/write-paper/scripts/gate_backbone_fulltext.py",
+      "size": 8358,
+      "sha256": "fa06f0d38c59b319a6cb892f4fb0beeccbf147ec4377a87d0f4b35a49d607050"
+    },
     {
       "path": "skills/write-paper/skill.yml",
-      "size": 2191,
-      "sha256": "06bd12d85f94702b6e23f74ea8e7009a746d31d247e437f4ba08ea6bd26f7d18"
+      "size": 2466,
+      "sha256": "185eb0948913e0fd0c0f6ddcf4bea30ff7d6b50185b5c6c401b51600039b745c"
     },
     {
       "path": "skills/write-protocol/SKILL.md",

package/metadata/distribution_manifest.json

@@ -1,6 +1,6 @@
 {
   "schema_version": 1,
-  "version": "5.17.0",
+  "version": "5.18.0",
   "owned_skills": [
     "academic-aio",
     "add-journal",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "medsci-skills",
-  "version": "5.17.0",
+  "version": "5.18.0",
   "description": "MedSci Skills — a medical/scientific research skill suite for AI coding agents (Claude Code, Codex, Cursor, Copilot). The npm package is a terminal-friendly installer shortcut; the canonical distribution remains the GitHub repository and the Claude Code plugin marketplace.",
   "license": "SEE LICENSE IN LICENSE",
   "homepage": "https://github.com/Aperivue/medsci-skills#readme",

package/skills/academic-aio/SKILL.md

@@ -107,6 +107,9 @@ This pattern is the canonical shape LLM extractors parse first.
 ### 2.4 Reproducibility block
 Include a labeled block (typically end of Methods or a standalone Data/Code Availability section) listing: data availability and license, code availability with DOI, model weights and checkpoints, prompts and configuration files, random seeds, compute environment. This block is disproportionately scraped by AI agents when they cite a paper as reproducible.
 
+### 2.4a Citing an AI-assisted tool by use-class
+If the work used an AI-assisted tool (a verification/QA suite, analysis code, or a generative drafting assistant), frame the mention by *what it did*, not by hiding it — under current journal wariness, a proud in-text citation of a **generative** use invites suspicion, while the same placement for a **verification** use reads as rigor (like citing a reference manager or a linter). Split the mention: **verification/QA and analysis** → a **Software / Code-availability statement** (citable); **generative** drafting/humanizing → the journal's **AI-use disclosure field**, not a citation. A self-citation by the tool's author additionally requires a COI disclosure, and you should cite only the functions the work actually used. Full use-class table and rules: `${CLAUDE_SKILL_DIR}/references/ai_tool_citation_framing.md`.
+
 ### 2.5 Limitations enumeration
 List limitations explicitly and name each one (generalizability, spectrum bias, dataset shift, single-center training, label noise). Papers with enumerated limitations score higher for trustworthiness in LLM summarization benchmarks.
 

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

@@ -0,0 +1,53 @@
+# Citing an AI-assisted research tool safely (framing by use-class)
+
+When a manuscript used an AI-assisted tool — a reference-verification / QA suite, a
+statistical-analysis helper, or a generative drafting assistant — *where and how you
+name it* changes how an editor or reviewer reads it. Under current journal wariness
+about AI, a proud in-text citation of a **generative** use can invite suspicion or a
+desk-reject, while the identical placement for a **verification** use reads as rigor
+(the same way citing R, SPSS, or a reference manager does). The safe move is to frame
+the mention by what the tool actually *did*, not to hide it.
+
+This applies to any AI-assisted tool; it also applies to **self-citation** by a tool's
+author (e.g. citing MedSci Skills in your own paper), which additionally requires a
+conflict-of-interest disclosure.
+
+## The three use-classes
+
+| Use-class | What it covers | Where it belongs | Citable like software? |
+|---|---|---|---|
+| **Verification / QA** (rigor-signalling) | Reference/citation verification (DOI/PMID against PubMed/CrossRef), reporting-guideline compliance checks, deterministic integrity gates, numerical-consistency checks | **Software / Code-availability statement** (or Methods, as a named tool with version) | **Yes** — cite it plainly, like a reference manager or a linter. It signals rigor. |
+| **Analysis** (neutral) | Statistical-analysis code, figure generation, data-transformation scripts | Methods (named, with version) and/or Code-availability | **Yes** — neutral and citable, like citing R/Python packages. |
+| **Generative** (disclosure, not citation) | Drafting or rewriting prose, "humanizing", summarizing, idea generation | The journal's **AI-use disclosure field / statement** (per its policy), *not* a proud in-text citation | **No** — declare it in the disclosure field; do not farm it into the running text. |
+
+## Rules of thumb
+
+- **Prefer the deterministic, verifiable functions for the cleanest cite.** A tool's
+  reproducible/verifiable capabilities (reference verification, compliance gates,
+  analysis code) are the ones that read as rigor and are safest to name in a Software
+  statement. Its generative capabilities belong in the disclosure field.
+- **Match the placement to the use-class**, not to the tool. The *same* tool can be
+  citable (its QA gate ran your references) *and* disclosure-only (its drafting helper
+  touched your prose) in the *same* paper — split the mention accordingly.
+- **Do not self-cite generatively.** If the paper did not materially use the generative
+  parts, do not add a self-citation for them to inflate a citation count; cite only the
+  functions the work actually used.
+- **Pair a self-citation with a COI disclosure.** If you authored the tool, state the
+  relationship (e.g. "Author X develops the cited toolkit") in the COI/competing-interests
+  section — the same standard as any other intellectual/financial interest.
+- **Honor the target journal's AI policy first.** Where a journal specifies *where* AI
+  use must be declared (Methods vs Acknowledgements vs a dedicated field), that placement
+  wins for the generative use-class; the Software-statement guidance above is for the
+  verification/analysis classes, which are tool-use, not AI-authorship.
+
+## Why this is guidance, not a deterministic gate (yet)
+
+A deterministic check ("flag an AI-tool citation placed in running-text Methods for a
+*generative* use") would need both a maintained tool-name allowlist and a reliable
+classifier of *which use-class* a given sentence describes — the latter is high false-
+positive without context the grep cannot see. The reliable, low-FP part (a tool named in
+a Software/Code-availability statement) is already the recommended state, so there is
+nothing to flag. If a bounded, allowlist-driven placement check proves worthwhile on real
+manuscripts, it can be added later; until then this stays advisory. (See
+`~/.claude/rules/manuscript-style-classical.md` §7/§15 for AI-disclosure placement and the
+self-applicability rule.)

package/skills/design-study/SKILL.md

@@ -191,6 +191,19 @@ scale validity; a low overall ICC is interpretable only if raters at least conve
 - Require each item to be judged standalone; discourage cross-item references in free-text, which
   signal non-independent rating.
 
+**Human-as-operator arm (interactive / promptable AI).** The reader-study patterns above assume the
+human is a *rater / reference* judging outputs. Interactive / promptable segmentation (SAM2, MedSAM2,
+nnInteractive) inverts this: the human is the *operator* who places the prompts, so the measured object
+is the human-operated system's **accuracy + interaction count + time + learning curve**, not a rating.
+Design for it explicitly:
+- Define the operator population and their onboarding; a **learning curve** (performance vs case index)
+  is a first-class outcome, not noise to average away.
+- Fix the prompting protocol (allowed prompt types, stopping rule, target Dice) identically to any
+  simulated-prompting arm so the two are comparable — **protocol fidelity**, checked in `/model-validation`.
+- Pre-specify the interaction and timing metrics; their deterministic reporting gate is
+  `/model-evaluation --task interactive`. (A design document is free-form prose, so the deterministic
+  anchor for these items sits at the reporting stage, not on the protocol text.)
+
 For an AI-system-versus-human-expert benchmark specifically, route to `/design-ai-benchmarking`, which
 extends this subsection with arm definition, LLM-as-judge versus human-as-judge adjudication, and a
 structured export schema.

package/skills/make-figures/skill.yml

@@ -50,5 +50,6 @@ known_limitations:
 validation_commands:
   - "Rscript scripts/generate_flow_diagram.R"
   - "python3 scripts/validate_pptx_mac_compat.py <file>"
+  - "python3 tests/test_pptx_mac_compat.py"
   - "bash scripts/render_core_figures_challenge/verify.sh"
 evidence_surface: demo