PMx Skills Index: evidence, 2026-10
What every score in the index rests on
Written by Claude, not by a person. This page is generated by the monthly run from list/2026-10.yml. The index explains how to read it, and Specification 2 defines every score.
Tool Skills
nlmixr2llm · estimation
john-harrold/nlmixr2llm at ac6b240 · tool skill · rank 1 · tags: nlmixr2; estimation
| S | Q | E | M | U |
|---|---|---|---|---|
| 51.0 | 52.1 | 52.0 | 85.6 | 30.0 |
Quality. 4 of 7 applicable rubric items met (57.1), less 0 major and 1 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Vignettes listed at the end; authoring rules cite none. |
| R6 States the limitations of its own output | ✅ yes | “Report estimates with their precision and any caveats (shrinkage, bounds, non-convergence).” |
| R7 Code examples complete enough to run | ✅ yes | Complete theophylline example: “fit <- nlmixr2(one.compartment, theo_sd, est = "saem", saemControl(print = 0))” |
| R8 Examples run in continuous integration | ✅ yes | skill-examples.yaml runs it. |
| R9 Continuous integration passes on the pinned commit | ✅ yes | success on ac6b240. |
| R10 Stores outputs from a real run | no | No stored outputs. |
- Minor error. “BSV% near 0 or > 100% | ETA unsupported by data; remove it” Between-subject variability above 100% CV is common for some parameters (absorption, for example) and is not evidence that the data fail to support the ETA.
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| John Harrold | 1.0 | 13 | record | ORCID 0000-0003-2052-4373, from the DESCRIPTION of john-harrold/ruminate |
Maintenance. Last change to the skill’s files 2026-09-05 (32 days; recency 91.2). Usability 80: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. 3 points from people outside the authors in the last 12 months.
| Person | What they did | Expert | Evidence |
|---|---|---|---|
Bill Denney (billdenney) |
opened issues | yes: ORCID 0000-0002-5759-428X; field h-index 12 | 1, 2, 3 |
Repository stars and forks, used only to break ties: 8 and 3.
nlmixr2llm · simulation
john-harrold/nlmixr2llm at ac6b240 · tool skill · rank 1 (range 1–2) · tags: rxode2; simulation
| S | Q | E | M | U |
|---|---|---|---|---|
| 46.9 | 42.9 | 52.0 | 85.6 | 30.0 |
Quality. 3 of 7 applicable rubric items met (42.9), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Vignettes listed at the end only. |
| R6 States the limitations of its own output | no | Notes package quirks but not the limits of a simulation’s output. |
| R7 Code examples complete enough to run | ✅ yes | “sim <- rxSolve(mod, ev)” with model and event table. |
| R8 Examples run in continuous integration | ✅ yes | skill-examples.yaml runs it. |
| R9 Continuous integration passes on the pinned commit | ✅ yes | success on ac6b240. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| John Harrold | 1.0 | 13 | record | ORCID 0000-0003-2052-4373, from the DESCRIPTION of john-harrold/ruminate |
Maintenance. Last change to the skill’s files 2026-09-05 (32 days; recency 91.2). Usability 80: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. 3 points from people outside the authors in the last 12 months.
| Person | What they did | Expert | Evidence |
|---|---|---|---|
Bill Denney (billdenney) |
opened issues | yes: ORCID 0000-0002-5759-428X; field h-index 12 | 1, 2, 3 |
Repository stars and forks, used only to break ties: 8 and 3.
nlmixr2llm · interop
john-harrold/nlmixr2llm at ac6b240 · tool skill · rank 3 · tags: nlmixr2; NONMEM; Monolix; model conversion
| S | Q | E | M | U |
|---|---|---|---|---|
| 40.4 | 28.6 | 52.0 | 85.6 | 30.0 |
Quality. 2 of 7 applicable rubric items met (28.6), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Vignettes listed at the end only. |
| R6 States the limitations of its own output | ✅ yes | “A translation is only trustworthy once rxode2 reproduces the engine’s own PRED/IPRED.” |
| R7 Code examples complete enough to run | ✅ yes | Complete babelmixr2 call with model and data. |
| R8 Examples run in continuous integration | no | tests/testthat/helper-examples.R skips both SKILL.md blocks: “runs NONMEM and Monolix”, “placeholder paths only”. |
| R9 Continuous integration passes on the pinned commit | no | The examples are not run, so a passing CI says nothing about them. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| John Harrold | 1.0 | 13 | record | ORCID 0000-0003-2052-4373, from the DESCRIPTION of john-harrold/ruminate |
Maintenance. Last change to the skill’s files 2026-09-05 (32 days; recency 91.2). Usability 80: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. 3 points from people outside the authors in the last 12 months.
| Person | What they did | Expert | Evidence |
|---|---|---|---|
Bill Denney (billdenney) |
opened issues | yes: ORCID 0000-0002-5759-428X; field h-index 12 | 1, 2, 3 |
Repository stars and forks, used only to break ties: 8 and 3.
nlmixr2lib · extract-literature-model
nlmixr2/nlmixr2lib at 5ddf9ae · tool skill · unproven: no evidence of use, so not ranked · tags: nlmixr2; model building from literature
| S | Q | E | M | U |
|---|---|---|---|---|
| 47.6 | 57.1 | 48.0 | 98.9 | 0.0 |
Quality. 4 of 7 applicable rubric items met (57.1), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “Never attempt extraction from an abstract alone”; systematic reviews are skipped by default. |
| R2 Cites sources for its recommendations | no | Recommendations point to the skill’s own reference files, not to published sources. |
| R6 States the limitations of its own output | ✅ yes | “Treat its output as a worklist, not a verdict”; the vignette must list “Assumptions and deviations”. |
| R7 Code examples complete enough to run | ✅ yes | Complete shell and R commands for each gate, e.g. the RENDER_GATE block. |
| R8 Examples run in continuous integration | no | No workflow runs the skill’s own code blocks. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | ✅ yes | The repository holds the models and validation vignettes the workflow produced (inst/modeldb, vignettes/articles). |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Bill Denney | 1.0 | 12 | record | ORCID 0000-0002-5759-428X, from nlmixr2lib DESCRIPTION |
Maintenance. Last change to the skill’s files 2026-09-29 (8 days; recency 97.8). Usability 100: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✅.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 14 and 6.
nonmem.utils · r-nonmem.utils
pchelle/nonmem.utils at 1827ac0 · tool skill · unproven: no evidence of use, so not ranked · tags: nonmem.utils; NONMEM
| S | Q | E | M | U |
|---|---|---|---|---|
| 38.5 | 42.9 | 44.0 | 81.5 | 0.0 |
Quality. 3 of 7 applicable rubric items met (42.9), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “Distinct from the task-oriented gof-analysis / covariate-analysis / dataset-analysis skills” |
| R2 Cites sources for its recommendations | ✅ yes | “Read references/API.md before writing code.” |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
| R7 Code examples complete enough to run | no | Function signatures only; no complete example. |
| R8 Examples run in continuous integration | no | CI runs the package tests, not the skill. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | ✅ yes | evals/iteration-1 holds outputs of with-skill and without-skill runs (benchmark.json). |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Pierre Chelle | 1.0 | 11 | record | ORCID 0000-0002-0167-4774, from nonmem.utils DESCRIPTION |
Maintenance. Last change to the skill’s files 2026-08-06 (62 days; recency 83.0). Usability 80: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
pk-modeling-skills · easy-nonmem
kinginsun/pk-modeling-skills at 0074bdd · tool skill · unproven: no evidence of use, so not ranked · tags: NONMEM; PsN
| S | Q | E | M | U |
|---|---|---|---|---|
| 34.3 | 57.1 | 0.0 | 85.6 | 0.0 |
Quality. 4 of 7 applicable rubric items met (57.1), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “use additional PsN tools (vpc, bootstrap, etc.) only when the user asks—keep the default loop to execute + tables + R plots” |
| R2 Cites sources for its recommendations | ✅ yes | “Always read the relevant reference file before inventing options”; references/ holds the NONMEM 7.6 documentation. |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
| R7 Code examples complete enough to run | ✅ yes | “execute run1.mod -directory=run1_execute” with runnable projects under examples/. |
| R8 Examples run in continuous integration | no | No continuous integration. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | ✅ yes | examples/nonmem-pop-pk carries reference .lst and .ext files. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Yufeng Zhang (kinginsun) | 1.0 | 0 | none | not found: commits by a bot; repository owner Yufeng Zhang, CUHK; 310 OpenAlex authors share the name |
Maintenance. Last change to the skill’s files 2026-09-05 (32 days; recency 91.2). Usability 80: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 0 and 0.
nonmem.utils · gof-analysis
pchelle/nonmem.utils at 1827ac0 · tool skill · unproven: no evidence of use, so not ranked · tags: model evaluation; nonmem.utils
| S | Q | E | M | U |
|---|---|---|---|---|
| 31.1 | 28.6 | 44.0 | 72.6 | 0.0 |
Quality. 2 of 7 applicable rubric items met (28.6), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Cites none. |
| R6 States the limitations of its own output | ✅ yes | “Sparse or filtered data causing unstable diagnostics: warn and continue.” |
| R7 Code examples complete enough to run | ✅ yes | Minimal command example reads the table and draws the plots. |
| R8 Examples run in continuous integration | no | CI runs the package tests, not the skill. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Pierre Chelle | 1.0 | 11 | record | ORCID 0000-0002-0167-4774, from nonmem.utils DESCRIPTION |
Maintenance. Last change to the skill’s files 2026-06-02 (127 days; recency 65.2). Usability 80: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. Copied unchanged into pchelle/ferx.utils.
nonmem.utils · poppk-report-workflow
pchelle/nonmem.utils at 1827ac0 · tool skill · unproven: no evidence of use, so not ranked · tags: report writing; nonmem.utils
| S | Q | E | M | U |
|---|---|---|---|---|
| 31.1 | 28.6 | 44.0 | 72.6 | 0.0 |
Quality. 2 of 7 applicable rubric items met (28.6), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Cites none. |
| R6 States the limitations of its own output | ✅ yes | “Use graceful degradation for optional diagnostics … and clearly label skipped sections.” |
| R7 Code examples complete enough to run | ✅ yes | Report function calls with file arguments. |
| R8 Examples run in continuous integration | no | CI runs the package tests, not the skill. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Pierre Chelle | 1.0 | 11 | record | ORCID 0000-0002-0167-4774, from nonmem.utils DESCRIPTION |
Maintenance. Last change to the skill’s files 2026-06-02 (127 days; recency 65.2). Usability 80: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. Copied unchanged into pchelle/ferx.utils.
pk-modeling-skills · monolix-cli
kinginsun/pk-modeling-skills at 0074bdd · tool skill · unproven: no evidence of use, so not ranked · tags: Monolix
| S | Q | E | M | U |
|---|---|---|---|---|
| 27.9 | 42.9 | 0.0 | 85.6 | 0.0 |
Quality. 3 of 7 applicable rubric items met (42.9), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “monolix-cli does not plot” |
| R2 Cites sources for its recommendations | no | Options are described as tested, not tied to a cited document. |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
| R7 Code examples complete enough to run | ✅ yes | Complete monolix.sh and lixoftConnectors calls. |
| R8 Examples run in continuous integration | no | No continuous integration. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | ✅ yes | examples/monolix-theophylline holds a result folder from a real run. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Yufeng Zhang (kinginsun) | 1.0 | 0 | none | not found: commits by a bot; repository owner Yufeng Zhang, CUHK; 310 OpenAlex authors share the name |
Maintenance. Last change to the skill’s files 2026-09-05 (32 days; recency 91.2). Usability 80: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 0 and 0.
ferx.utils · r-ferx
pchelle/ferx.utils at ec220d6 · tool skill · unproven: no evidence of use, so not ranked · tags: ferx
| S | Q | E | M | U |
|---|---|---|---|---|
| 25.1 | 14.3 | 44.0 | 77.1 | 0.0 |
Quality. 1 of 7 applicable rubric items met (14.3), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Points to references/API.md, which the skill folder does not contain. |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
| R7 Code examples complete enough to run | ✅ yes | “fit <- ferx_fit(ex\(model, ex\)data, method = "focei", covariance = TRUE)” on a bundled example. |
| R8 Examples run in continuous integration | no | No workflow runs the skill. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Pierre Chelle | 1.0 | 11 | record | ORCID 0000-0002-0167-4774, from nonmem.utils DESCRIPTION |
Maintenance. Last change to the skill’s files 2026-09-16 (21 days; recency 94.2). Usability 60: licence ✅, frontmatter ✅, install ✅, dependencies ✗, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 0 and 0.
nonmem.utils · covariate-analysis
pchelle/nonmem.utils at 1827ac0 · tool skill · unproven: no evidence of use, so not ranked · tags: covariate modeling; nonmem.utils
| S | Q | E | M | U |
|---|---|---|---|---|
| 24.7 | 14.3 | 44.0 | 72.6 | 0.0 |
Quality. 1 of 7 applicable rubric items met (14.3), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Cites none. |
| R6 States the limitations of its own output | no | Says nothing about shrinkage or the reliability of ETA-covariate plots. |
| R7 Code examples complete enough to run | ✅ yes | Minimal command example. |
| R8 Examples run in continuous integration | no | CI runs the package tests, not the skill. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Pierre Chelle | 1.0 | 11 | record | ORCID 0000-0002-0167-4774, from nonmem.utils DESCRIPTION |
Maintenance. Last change to the skill’s files 2026-06-02 (127 days; recency 65.2). Usability 80: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. Copied unchanged into pchelle/ferx.utils.
nonmem.utils · dataset-analysis
pchelle/nonmem.utils at 1827ac0 · tool skill · unproven: no evidence of use, so not ranked · tags: exploratory data analysis; nonmem.utils
| S | Q | E | M | U |
|---|---|---|---|---|
| 24.7 | 14.3 | 44.0 | 72.6 | 0.0 |
Quality. 1 of 7 applicable rubric items met (14.3), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Cites none. |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
| R7 Code examples complete enough to run | ✅ yes | Minimal command example. |
| R8 Examples run in continuous integration | no | CI runs the package tests, not the skill. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Pierre Chelle | 1.0 | 11 | record | ORCID 0000-0002-0167-4774, from nonmem.utils DESCRIPTION |
Maintenance. Last change to the skill’s files 2026-06-02 (127 days; recency 65.2). Usability 80: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. Copied unchanged into pchelle/ferx.utils.
APMODE · apmode-cli
biostochastics/APMODE at cbdcadf · tool skill · unproven: no evidence of use, so not ranked · tags: APMODE; nlmixr2; Stan
| S | Q | E | M | U |
|---|---|---|---|---|
| 21.6 | 28.6 | 8.0 | 67.3 | 0.0 |
Quality. 2 of 7 applicable rubric items met (28.6), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | Section “When APMODE is the wrong tool”. |
| R2 Cites sources for its recommendations | no | Cites none. |
| R6 States the limitations of its own output | no | Notes a data-exposure caveat for the agentic backend, not the limits of the analysis. |
| R7 Code examples complete enough to run | ✅ yes | Complete “uv run apmode …” commands. |
| R8 Examples run in continuous integration | no | CI runs the package tests, not the skill’s commands. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Sergey A. Kornilov | 1.0 | 2 | record | probable: the only OpenAlex author of that name with 3 or more field works |
Maintenance. Last change to the skill’s files 2026-04-24 (166 days; recency 54.5). Usability 80: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
jinko-skills · jinko
novainsilico/jinko-skills at dd7ff2e · tool skill · unproven: no evidence of use, so not ranked · tags: Jinkō
| S | Q | E | M | U |
|---|---|---|---|---|
| 16.3 | 14.3 | 0.0 | 98.9 | 0.0 |
Quality. 1 of 7 applicable rubric items met (14.3), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “This skill does not make scientific decisions, plan workflows, execute SDK calls, or decide that a task step is complete.” |
| R2 Cites sources for its recommendations | no | Cites none. |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
| R7 Code examples complete enough to run | no | A routing map; no code. |
| R8 Examples run in continuous integration | no | No continuous integration. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | evals/ holds prompts, not outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Nova In Silico | 1.0 | 0 | none | not found: commits by a publishing bot; no authors named in the repository |
Maintenance. Last change to the skill’s files 2026-09-29 (8 days; recency 97.8). Usability 100: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✅.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. A router to the other Jinkō skills.
aNCA · pknca-compatibility
pharmaverse/aNCA at e6b468f · tool skill · unproven: no evidence of use, so not ranked · tags: PKNCA; aNCA; package development
| S | Q | E | M | U |
|---|---|---|---|---|
| 16.3 | 14.3 | 0.0 | 98.4 | 0.0 |
Quality. 1 of 7 applicable rubric items met (14.3), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | ✅ yes | “Prefer source code, NEWS, tests, and linked PKNCA issues/PRs over assumptions” with the upstream repository named. |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
| R7 Code examples complete enough to run | no | No code. |
| R8 Examples run in continuous integration | no | Names the package tests to add; does not run the skill. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Gerardo Jose Rodriguez | 1.0 | 0 | record | ORCID 0000-0003-1413-0060, from aNCA DESCRIPTION |
Maintenance. Last change to the skill’s files 2026-09-25 (12 days; recency 96.7). Usability 100: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✅.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 44 and 17.
Note. A developer skill for maintaining the aNCA package.
jinko-skills · jinko-task-from-nonmem
novainsilico/jinko-skills at dd7ff2e · tool skill · unproven: no evidence of use, so not ranked · tags: Jinkō; NONMEM; model conversion
| S | Q | E | M | U |
|---|---|---|---|---|
| 16.2 | 14.3 | 0.0 | 98.1 | 0.0 |
Quality. 1 of 7 applicable rubric items met (14.3), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Cites none. |
| R6 States the limitations of its own output | ✅ yes | “Never claim a conversion succeeded without [a measured agreement] … Say which reference the numbers came from.” |
| R7 Code examples complete enough to run | no | Only the pip install line; script flags are listed, not shown as a runnable call. |
| R8 Examples run in continuous integration | no | No continuous integration. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | evals/ holds prompts, not outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Nova In Silico | 1.0 | 0 | none | not found: commits by a publishing bot; no authors named in the repository |
Maintenance. Last change to the skill’s files 2026-09-23 (14 days; recency 96.2). Usability 100: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✅.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. Requires a numerical equivalence check against scipy, rxode2 or the NONMEM $TABLE, which the rubric does not reward.
nonmem-skill · nonmem
jaj42/nonmem-skill at 2a32fed · tool skill · unproven: no evidence of use, so not ranked · tags: NONMEM
| S | Q | E | M | U |
|---|---|---|---|---|
| 14.2 | 17.9 | 4.0 | 51.8 | 0.0 |
Quality. 3 of 7 applicable rubric items met (42.9), less 1 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | ✅ yes | Cites Hooker 2007 for CWRES, Bergstrand 2011 for pcVPC, Bauer 2019 and the NONMEM user guides. |
| R6 States the limitations of its own output | no | No statement of the limits of its own output. |
| R7 Code examples complete enough to run | ✅ yes | A complete two-compartment control stream skeleton. |
| R8 Examples run in continuous integration | no | No continuous integration. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | ✅ yes | example/theophylline.res, a saved NONMEM listing from a run of the example. |
- ❌ Major error. references/diagnostics-vpc.md line 30: “> 50%: severe; the data don’t support estimating IIV on that parameter. Consider fixing the omega to 0 or to a literature value.” High ETA shrinkage means the individual estimates carry little information; it does not show that the population variance is absent. Savic and Karlsson 2009 (AAPS J 11:558). Fixing omega to 0 on this rule removes real variability from the model.
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Jona Joachim | 1.0 | 1 | record | probable: the only OpenAlex author of that name with 3 or more field works |
Maintenance. Last change to the skill’s files 2026-05-27 (133 days; recency 63.6). Usability 40: licence ✅, frontmatter ✅, install ✗, dependencies ✗, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 2 and 0.
Note. Known-answer fixture 1.
pk-sim-pbpk · pk-sim-pbpk
gscas/pk-sim-pbpk at 80e514b · tool skill · unproven: no evidence of use, so not ranked · tags: PK-Sim; PBPK
| S | Q | E | M | U |
|---|---|---|---|---|
| 4.9 | 0 | 0.0 | 49.2 | 0.0 |
Quality. 0 of 7 applicable rubric items met (0.0), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Cites none. |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
| R7 Code examples complete enough to run | no | Fragments only, e.g. “sim = ospsuite.Simulation(…)”. |
| R8 Examples run in continuous integration | no | No continuous integration. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| gscas | 1.0 | 0 | none | not found: no name on the GitHub profile |
Maintenance. Last change to the skill’s files 2026-05-08 (152 days; recency 58.4). Usability 40: licence ✗, frontmatter ✅, install ✗, dependencies ✅, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 0 and 0.
Note. No licence.
AI-assisted-MIDD · poppk-estimation
nlmixr2/AI-assisted-MIDD at 3f07128 · tool skill · unproven: no evidence of use, so not ranked · tags: nlmixr2; estimation; model building
| S | Q | E | M | U |
|---|---|---|---|---|
| 30.9 | 37.9 | 24.0 | 78.9 | 0.0 |
Quality. 3 of 7 applicable rubric items met (42.9), less 0 major and 1 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “When NOT to use”: NCA goes to pk-nca, and “NONMEM/Monolix/PKNCA backends via babelmixr2 are not covered by these skills.” |
| R2 Cites sources for its recommendations | no | nlmixr2 vignettes listed at the end; the authoring rules and acceptance thresholds cite none. |
| R6 States the limitations of its own output | ✅ yes | “warn must be recorded as a caveat in the spec (diagnostics.caveats)” |
| R7 Code examples complete enough to run | ✅ yes | Complete Core Process from data to fits, with templates/analysis-poppk.R scaffolded by new_version(“poppk”). |
| R8 Examples run in continuous integration | no | The repository has no continuous integration. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | The README describes worked examples under examples/abc-111 and examples/gs-12345 with outputs passing QC, but neither folder is in the repository. |
- Minor error. “BSV% near 0 or > 100% | ETA unsupported by data; remove it” Carried over from nlmixr2llm estimation, where the same finding was made: between-subject variability above 100% CV is common for some parameters (absorption, for example) and is not evidence that the data fail to support the ETA.
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Mutaz M. Jaber | 1.0 | 6 | record | ORCID 0000-0002-7536-8753, from the DESCRIPTION of Mutaz94/mypkg; sole committer |
Maintenance. Last change to the skill’s files 2026-10-01 (8 days; recency 97.8). Usability 60: licence ✗, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. Adapted from nlmixr2llm estimation; 16% of its lines match that skill, below the 80% copy threshold. Added 2026-10-09, outside the monthly run, at Andy’s request; discovery had missed the repository, created 2026-10-01. Four of its 18 skills were scored. The other 14 are the project, database, logging, run-spec, table, figure, report and QC plumbing around these four, and were not scored: pk-nca-database, pk-nca-figures, pk-nca-logging, pk-nca-qc-tests, pk-nca-report, pk-nca-run-spec, pk-nca-tables, poppk-database, poppk-qc-tests, poppk-report, poppk-run-spec, poppk-tables, pk-project, manage-skills.
AI-assisted-MIDD · pk-nca
nlmixr2/AI-assisted-MIDD at 3f07128 · tool skill · unproven: no evidence of use, so not ranked · tags: PKNCA; NCA
| S | Q | E | M | U |
|---|---|---|---|---|
| 30.9 | 37.9 | 24.0 | 78.9 | 0.0 |
Quality. 3 of 7 applicable rubric items met (42.9), less 0 major and 1 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “Do not use this package for compartmental models or for fitting nonlinear mixed-effects models” |
| R2 Cites sources for its recommendations | no | PKNCA vignettes are listed by topic in references/pknca-resources.md; the NCA rules in SKILL.md cite none. |
| R6 States the limitations of its own output | ✅ yes | “Half-life comes from λz regression on the terminal phase; check adj.r.squared / span.ratio per subject before trusting it.” |
| R7 Code examples complete enough to run | ✅ yes | templates/analysis-nca.R, the scaffolded pipeline from data file to pk.nca(). |
| R8 Examples run in continuous integration | no | The repository has no continuous integration. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | The README describes worked examples under examples/abc-111 and examples/gs-12345 with outputs passing QC, but neither folder is in the repository. |
- Minor error. “Set route = "extravascular" (or "intravascular") and dur = NA_real_ in the dose data frame; PKNCA picks these up by column name.” PKNCAdose() picks up a column named route, but looks for duration, not dur, and otherwise sets duration to 0 (PKNCA source, setDuration.PKNCAdose). An infusion duration entered in dur is silently ignored.
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Mutaz M. Jaber | 1.0 | 6 | record | ORCID 0000-0002-7536-8753, from the DESCRIPTION of Mutaz94/mypkg; sole committer |
Maintenance. Last change to the skill’s files 2026-10-01 (8 days; recency 97.8). Usability 60: licence ✗, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. Added 2026-10-09, outside the monthly run, at Andy’s request; discovery had missed the repository, created 2026-10-01. Four of its 18 skills were scored. The other 14 are the project, database, logging, run-spec, table, figure, report and QC plumbing around these four, and were not scored: pk-nca-database, pk-nca-figures, pk-nca-logging, pk-nca-qc-tests, pk-nca-report, pk-nca-run-spec, pk-nca-tables, poppk-database, poppk-qc-tests, poppk-report, poppk-run-spec, poppk-tables, pk-project, manage-skills.
AI-assisted-MIDD · poppk-simulation
nlmixr2/AI-assisted-MIDD at 3f07128 · tool skill · unproven: no evidence of use, so not ranked · tags: rxode2; simulation
| S | Q | E | M | U |
|---|---|---|---|---|
| 33.2 | 42.9 | 24.0 | 78.9 | 0.0 |
Quality. 3 of 7 applicable rubric items met (42.9), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “Don’t refit, or edit a fitted model’s estimates, inside a simulation.” |
| R2 Cites sources for its recommendations | no | rxode2 vignettes listed at the end; the authoring rules cite none. |
| R6 States the limitations of its own output | ✅ yes | “Source caveats (unapproved source QC, model-file parameter origin) stated with the results.” |
| R7 Code examples complete enough to run | ✅ yes | Complete Core Process and event-table cheatsheet, with templates/analysis-sim.R scaffolded by new_version(“poppk-sim”). |
| R8 Examples run in continuous integration | no | The repository has no continuous integration. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | The README describes worked examples under examples/abc-111 and examples/gs-12345 with outputs passing QC, but neither folder is in the repository. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Mutaz M. Jaber | 1.0 | 6 | record | ORCID 0000-0002-7536-8753, from the DESCRIPTION of Mutaz94/mypkg; sole committer |
Maintenance. Last change to the skill’s files 2026-10-01 (8 days; recency 97.8). Usability 60: licence ✗, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. Adapted from nlmixr2llm simulation. Added 2026-10-09, outside the monthly run, at Andy’s request; discovery had missed the repository, created 2026-10-01. Four of its 18 skills were scored. The other 14 are the project, database, logging, run-spec, table, figure, report and QC plumbing around these four, and were not scored: pk-nca-database, pk-nca-figures, pk-nca-logging, pk-nca-qc-tests, pk-nca-report, pk-nca-run-spec, pk-nca-tables, poppk-database, poppk-qc-tests, poppk-report, poppk-run-spec, poppk-tables, pk-project, manage-skills.
Task Skills
scientific-agent-skills · pkpd-modeling
K-Dense-AI/scientific-agent-skills at 92ace75 · task skill · rank 1 · tags: PK/PD modeling; NCA; population PK
| S | Q | E | M | U |
|---|---|---|---|---|
| 58.4 | 90.0 | 0.0 | 99.2 | 40.0 |
Quality. 9 of 10 applicable rubric items met (90.0), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “it does not establish clinical safety, recommend a patient dose, or certify a regulatory submission” |
| R2 Cites sources for its recommendations | ✅ yes | Cites ICH M12, ICH E14, FDA’s May 2026 BE guidance, ICH M13B and M15, with a source ledger in references/. |
| R3 Says what to do when a check fails | ✅ yes | “Prefer residual inspection, physiological plausibility, sensitivity/profile checks and suitable bootstrap or simulation-based comparisons.” |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | “The 20% extrapolation, 0.8 adjusted R² and two-half-life flags are screening conventions, not universal acceptance rules.” |
| R5 Ties evaluation depth to intended use | ✅ yes | “Match validation to intended use.” |
| R6 States the limitations of its own output | ✅ yes | “This skill computes exploratory quantities and documents assumptions” |
| R7 Code examples complete enough to run | ✅ yes | Command lines for each script, e.g. “python3 nca.py -i profile.csv –dose 100 –route extravascular” |
| R8 Examples run in continuous integration | ✅ yes | skill-tests.yml runs tests/pkpd-modeling/test_scripts.py on every change to skills/. |
| R9 Continuous integration passes on the pinned commit | ✅ yes | Skill Tests run on 526ebce, the last commit touching the skill: success (2026-10-01). |
| R10 Stores outputs from a real run | no | Tests use synthetic fixtures; no stored outputs from a real run. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Timothy Kassis | 1.0 | 0 | record | name in CITATION.cff; OpenAlex record has no field works |
Maintenance. Last change to the skill’s files 2026-10-01 (6 days; recency 98.4). Usability 100: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✅.
Use. 4 points from people outside the authors in the last 12 months.
| Person | What they did | Expert | Evidence |
|---|---|---|---|
Sricharan Samba (ssamba1) |
pull request changing the skill | no | 1, 2 |
Min-Gul Kim (kimmingul) |
installed the skill in a project holding 20 hand-picked skills | yes: name and affiliation (Jeonbuk National University) match OpenAlex A5081300522; field h-index 14 | 1 |
Repository stars and forks, used only to break ties: 47846 and 4321.
nlmixr2llm · reporting
john-harrold/nlmixr2llm at ac6b240 · task skill · rank 2 (range 1–2) · tags: model evaluation; report writing; nlmixr2
| S | Q | E | M | U |
|---|---|---|---|---|
| 50.1 | 50.0 | 52.0 | 85.6 | 30.0 |
Quality. 5 of 10 applicable rubric items met (50.0), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes; ‘What NOT to do’ lists bad practice, not scope. |
| R2 Cites sources for its recommendations | no | Sources are listed once at the end (package vignettes); the checklist items themselves cite none. |
| R3 Says what to do when a check fails | ✅ yes | “treat ETA-based conclusions as weak where shrinkage > ~30%” |
| R4 Qualifies numeric thresholds, or gives none | no | “|CWRES| mostly < 3” is given bare. |
| R5 Ties evaluation depth to intended use | no | No link between diagnostics and the model’s intended use. |
| R6 States the limitations of its own output | ✅ yes | “Summarise in words: … what the diagnostics show, and caveats.” |
| R7 Code examples complete enough to run | ✅ yes | Complete R blocks, e.g. “vpcPlot(fit, n = 500, pred_corr = TRUE, stratify = "DOSE")” |
| R8 Examples run in continuous integration | ✅ yes | Workflow skill-examples.yaml runs the fenced R blocks weekly against CRAN nlmixr2. |
| R9 Continuous integration passes on the pinned commit | ✅ yes | skill-examples.yaml run on ac6b240: success (2026-10-05). |
| R10 Stores outputs from a real run | no | No stored outputs from a real run in the repository. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| John Harrold | 1.0 | 13 | record | ORCID 0000-0003-2052-4373, from the DESCRIPTION of john-harrold/ruminate |
Maintenance. Last change to the skill’s files 2026-09-05 (32 days; recency 91.2). Usability 80: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. 3 points from people outside the authors in the last 12 months.
| Person | What they did | Expert | Evidence |
|---|---|---|---|
Bill Denney (billdenney) |
opened issues | yes: ORCID 0000-0002-5759-428X; field h-index 12 | 1, 2, 3 |
Repository stars and forks, used only to break ties: 8 and 3.
jinko-skills · jinko-task-ichm15-planning
novainsilico/jinko-skills at dd7ff2e · task skill · unproven: no evidence of use, so not ranked · tags: MIDD planning
| S | Q | E | M | U |
|---|---|---|---|---|
| 54.8 | 100.0 | 0.0 | 98.1 | 0.0 |
Quality. 6 of 6 applicable rubric items met (100.0), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “this skill focuses specifically on the planning stage … before model outcomes are available” |
| R2 Cites sources for its recommendations | ✅ yes | Each element cites its ICH M15 section, e.g. “Model Influence … (§2.1.3–2.1.6)”. |
| R3 Says what to do when a check fails | ✅ yes | “If the Question of Interest, or any one Context of Use component … is not stated, ask one focused question” |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | Ratings are low/medium/high with stated justification; no numeric thresholds. |
| R5 Ties evaluation depth to intended use | ✅ yes | “Model Risk … determines the required rigor of model evaluation” |
| R6 States the limitations of its own output | ✅ yes | “list every invented or assumed element … in an explicit ‘Assumptions’ block at the top of the output” |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Nova In Silico | 1.0 | 0 | none | not found: commits by a publishing bot; no authors named in the repository |
Maintenance. Last change to the skill’s files 2026-09-23 (14 days; recency 96.2). Usability 100: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✅.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. Does not need Jinkō access except for the optional upload.
clinpharm-pmx-skills · review-pharmacometrics-deliverable
malekokour/clinpharm-pmx-skills at cefc4ea · task skill · unproven: no evidence of use, so not ranked · tags: document review
| S | Q | E | M | U |
|---|---|---|---|---|
| 53.7 | 83.3 | 28.0 | 92.5 | 0.0 |
Quality. 5 of 6 applicable rubric items met (83.3), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “Do not use for assessing the model itself … or to accept, reject or sign anything.” |
| R2 Cites sources for its recommendations | no | Cites no source. |
| R3 Says what to do when a check fails | ✅ yes | “An absent item is a finding, not a formatting note”; unsupported uses are listed. |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | Gives no numeric thresholds. |
| R5 Ties evaluation depth to intended use | ✅ yes | “Phase 2 — Fitness for intended use … For each, record what the deliverable would need to support it” |
| R6 States the limitations of its own output | ✅ yes | “State plainly which decisions remain with the signer.” |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Malek Okour | 1.0 | 7 | record | probable: the only OpenAlex author of that name with 3 or more field works |
Maintenance. Last change to the skill’s files 2026-08-13 (55 days; recency 84.9). Usability 100: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✅.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 6 and 0.
clinpharm-pmx-skills · review-model-analysis-deliverable
malekokour/clinpharm-pmx-skills at cefc4ea · task skill · unproven: no evidence of use, so not ranked · tags: document review; model evaluation
| S | Q | E | M | U |
|---|---|---|---|---|
| 52.2 | 80.0 | 28.0 | 92.5 | 0.0 |
Quality. 8 of 10 applicable rubric items met (80.0), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “It never re-fits, re-runs or critiques model structure … and it never decides whether a modelling choice was scientifically right.” |
| R2 Cites sources for its recommendations | ✅ yes | Rubrics anchored to ICH M15, the FDA PBPK guidance and the FDA exposure-response guidance through a guidance index. |
| R3 Says what to do when a check fails | ✅ yes | Missing inputs disable named checks and are reported as NEEDS_INPUT. |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | Criteria come from the analysis plan, not from fixed thresholds. |
| R5 Ties evaluation depth to intended use | ✅ yes | “I3 Commissioning question and decision context — without it, ‘conclusions traceable to a decision’ cannot be assessed” |
| R6 States the limitations of its own output | ✅ yes | “Population PK has no rubric in the library … popPK-specific content is marked CANNOT_ASSESS. Do not improvise the missing rubric.” |
| R7 Code examples complete enough to run | no | The skill calls its scripts but shows no runnable command. |
| R8 Examples run in continuous integration | ✅ yes | quality.yml runs scripts/check_all.py and the skill’s script tests. |
| R9 Continuous integration passes on the pinned commit | ✅ yes | quality.yml on f141199: success (2026-10-01). |
| R10 Stores outputs from a real run | no | Eval cases and fixtures, but no stored outputs for this skill. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Malek Okour | 1.0 | 7 | record | probable: the only OpenAlex author of that name with 3 or more field works |
Maintenance. Last change to the skill’s files 2026-08-13 (55 days; recency 84.9). Usability 100: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✅.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 6 and 0.
drug-discovery-agent-skills · pkpd-translation
K-Dense-AI/drug-discovery-agent-skills at 2bd3bde · task skill · unproven: no evidence of use, so not ranked · tags: translational PK/PD; human dose prediction
| S | Q | E | M | U |
|---|---|---|---|---|
| 49.3 | 88.9 | 0.0 | 92.9 | 0.0 |
Quality. 8 of 9 applicable rubric items met (88.9), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “When to stop using these scripts: When the site of action is not plasma …” |
| R2 Cites sources for its recommendations | ✅ yes | Cites the FDA 2005 maximum safe starting dose guidance for its Km factors. |
| R3 Says what to do when a check fails | ✅ yes | “above that, CL and Vz are guesses”; “The default factor of 10 should rise for steep dose–response, irreversible toxicity, or a novel target.” |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | The 20% extrapolation flag and the safety factor of 10 both come with context. |
| R6 States the limitations of its own output | ✅ yes | “it is planning arithmetic, not a regulatory submission” |
| R7 Code examples complete enough to run | ✅ yes | Command lines with full arguments. |
| R8 Examples run in continuous integration | ✅ yes | skill-tests.yml with tests/pkpd-translation. |
| R9 Continuous integration passes on the pinned commit | no | The latest Skill Tests runs failed (2026-09-01 to 09-03); none on the pinned commit. |
| R10 Stores outputs from a real run | ✅ yes | Outputs from real runs are quoted, e.g. “CL = 1.000000, Vz = 10.000000 … exact” against an analytic case. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Timothy Kassis | 1.0 | 0 | record | name in CITATION.cff; OpenAlex record has no field works |
Maintenance. Last change to the skill’s files 2026-08-16 (52 days; recency 85.8). Usability 100: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✅.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 35 and 4.
scientific-agents · clinical-pharmacologist
K-Dense-AI/scientific-agents at 98c7fae · task skill · unproven: no evidence of use, so not ranked · tags: clinical pharmacology
| S | Q | E | M | U |
|---|---|---|---|---|
| 46.4 | 83.3 | 0.0 | 88.9 | 0.0 |
Quality. 5 of 6 applicable rubric items met (83.3), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the persona does not do. |
| R2 Cites sources for its recommendations | ✅ yes | Recommendations name their guidance: ICH E14/S7B, M12, E5, E4. |
| R3 Says what to do when a check fails | ✅ yes | “Flat exposure–response: Wrong metric (total vs unbound), narrow studied range, misaligned sampling times, or PD delay — add effect compartment or time-varying exposure.” |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | “Allometry: pre-specify exponents or estimate with uncertainty”; fixed exponents are qualified. |
| R5 Ties evaluation depth to intended use | ✅ yes | “First classify the deliverable”, with analyses keyed to each. |
| R6 States the limitations of its own output | ✅ yes | “Hedge: distinguish predicted (model) vs observed (study)” |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Timothy Kassis | 1.0 | 0 | record | name in CITATION.cff; OpenAlex record has no field works |
Maintenance. Last change to the skill’s files 2026-09-29 (8 days; recency 97.8). Usability 80: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 200 and 30.
Note. A persona file with no code.
clinpharm-pmx-skills · assess-model-evaluation-diagnostics
malekokour/clinpharm-pmx-skills at cefc4ea · task skill · unproven: no evidence of use, so not ranked · tags: model evaluation; document review
| S | Q | E | M | U |
|---|---|---|---|---|
| 46.3 | 66.7 | 28.0 | 92.5 | 0.0 |
Quality. 4 of 6 applicable rubric items met (66.7), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “Refuses: Decide the outcome of Model evaluation and diagnostics; approve or submit related documents; select or adjust a dose” |
| R2 Cites sources for its recommendations | no | The eight checks cite no source. |
| R3 Says what to do when a check fails | ✅ yes | “For each check, also note … which required input unlocks the check if currently blocked”; gaps coded NEEDS_INPUT, UNKNOWN or CANNOT_ASSESS. |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | Gives no numeric thresholds. |
| R5 Ties evaluation depth to intended use | no | The checks are a generic template with the task name substituted; nothing ties evaluation depth to the model’s use. |
| R6 States the limitations of its own output | ✅ yes | “Never fill gaps with confident invention.” |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Malek Okour | 1.0 | 7 | record | probable: the only OpenAlex author of that name with 3 or more field works |
Maintenance. Last change to the skill’s files 2026-08-13 (55 days; recency 84.9). Usability 100: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✅.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 6 and 0.
Note. Templated: the same eight checks appear in other skills in the library with a different task name. It names no specific diagnostic.
clinpharm-pmx-skills · review-exposure-response-modelling
malekokour/clinpharm-pmx-skills at cefc4ea · task skill · unproven: no evidence of use, so not ranked · tags: exposure-response; document review
| S | Q | E | M | U |
|---|---|---|---|---|
| 46.3 | 66.7 | 28.0 | 92.5 | 0.0 |
Quality. 4 of 6 applicable rubric items met (66.7), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “Refuses: Decide the outcome of Exposure-response modelling …” |
| R2 Cites sources for its recommendations | no | The eight checks cite no source. |
| R3 Says what to do when a check fails | ✅ yes | Coded gap handling, as in the other templated skills. |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | Gives no numeric thresholds. |
| R5 Ties evaluation depth to intended use | no | Generic template; nothing ties review depth to intended use. |
| R6 States the limitations of its own output | ✅ yes | “Never fill gaps with confident invention.” |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Malek Okour | 1.0 | 7 | record | probable: the only OpenAlex author of that name with 3 or more field works |
Maintenance. Last change to the skill’s files 2026-08-13 (55 days; recency 84.9). Usability 100: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✅.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 6 and 0.
Note. Templated, as assess-model-evaluation-diagnostics.
clinpharm-pmx-skills · verify-nca-outputs
malekokour/clinpharm-pmx-skills at cefc4ea · task skill · unproven: no evidence of use, so not ranked · tags: NCA; document review
| S | Q | E | M | U |
|---|---|---|---|---|
| 46.3 | 66.7 | 28.0 | 92.5 | 0.0 |
Quality. 6 of 9 applicable rubric items met (66.7), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “Do not use when the request is … to rerun or re-derive the analysis” |
| R2 Cites sources for its recommendations | no | Rules come from the analysis plan; no published source cited. |
| R3 Says what to do when a check fails | ✅ yes | “Where the plan is silent on a rule the data show was applied, that is an UNKNOWN for a human to resolve” |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | Tolerances come from the plan and are named in every finding. |
| R6 States the limitations of its own output | ✅ yes | “Report reconciliation coverage as a fraction — values checked over values present.” |
| R7 Code examples complete enough to run | no | No runnable command in the skill. |
| R8 Examples run in continuous integration | ✅ yes | quality.yml tests the shared scripts it calls. |
| R9 Continuous integration passes on the pinned commit | ✅ yes | quality.yml on f141199: success. |
| R10 Stores outputs from a real run | no | examples/verify-nca-outputs holds synthetic inputs, not outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Malek Okour | 1.0 | 7 | record | probable: the only OpenAlex author of that name with 3 or more field works |
Maintenance. Last change to the skill’s files 2026-08-13 (55 days; recency 84.9). Usability 100: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✅.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 6 and 0.
aNCA · nca-scientific-change-review
pharmaverse/aNCA at e6b468f · task skill · unproven: no evidence of use, so not ranked · tags: NCA; code review
| S | Q | E | M | U |
|---|---|---|---|---|
| 36.9 | 60.0 | 0.0 | 99.0 | 0.0 |
Quality. 3 of 5 applicable rubric items met (60.0), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | ✅ yes | “identify the applicable CDISC standard/version or metadata authority” |
| R3 Says what to do when a check fails | ✅ yes | “Define regression evidence and manual scientific review needed”; deviations and unresolved compliance risks are recorded. |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | Gives no numeric thresholds. |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Gerardo Jose Rodriguez | 1.0 | 0 | record | ORCID 0000-0003-1413-0060, from aNCA DESCRIPTION |
Maintenance. Last change to the skill’s files 2026-09-30 (7 days; recency 98.1). Usability 100: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✅.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 44 and 17.
Note. A developer skill for maintaining the aNCA package.
scientific-agents · pharmacokineticist
K-Dense-AI/scientific-agents at 98c7fae · task skill · unproven: no evidence of use, so not ranked · tags: PK analysis
| S | Q | E | M | U |
|---|---|---|---|---|
| 36.7 | 61.7 | 0.0 | 88.9 | 0.0 |
Quality. 4 of 6 applicable rubric items met (66.7), less 0 major and 1 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the persona does not do. |
| R2 Cites sources for its recommendations | no | Names guidance documents in a list; few recommendations carry a source. |
| R3 Says what to do when a check fails | ✅ yes | “If VPC fails, stratify by dose, occasion, or covariate; check lag time, absorption model…” |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | Gives no bare thresholds; extrapolated AUC above 20% comes with an action. |
| R5 Ties evaluation depth to intended use | ✅ yes | “Ask what decision the analysis must support.” |
| R6 States the limitations of its own output | ✅ yes | “Hedge mechanistic claims: ‘consistent with renal elimination’ vs. ‘proves renal pathway’” |
- Minor error. “Use Monolix, Simcyp, or GastroPlus for PBPK and DDI simulation” Monolix is a population modelling tool, not a PBPK platform.
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Timothy Kassis | 1.0 | 0 | record | name in CITATION.cff; OpenAlex record has no field works |
Maintenance. Last change to the skill’s files 2026-09-29 (8 days; recency 97.8). Usability 80: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 200 and 30.
Note. A persona file with no code.
jinko-skills · jinko-task-paper2model
novainsilico/jinko-skills at dd7ff2e · task skill · unproven: no evidence of use, so not ranked · tags: model building from literature; QSP
| S | Q | E | M | U |
|---|---|---|---|---|
| 32.4 | 50.0 | 0.0 | 98.9 | 0.0 |
Quality. 5 of 10 applicable rubric items met (50.0), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Works from the source paper; cites no method source. |
| R3 Says what to do when a check fails | ✅ yes | “Record an approved assumption or stop with blocked_evidence before creating a model.” |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | “Freeze targets, comparison metrics, tolerances … before comparing simulations.” |
| R5 Ties evaluation depth to intended use | ✅ yes | “whether the task is structural implementation, numerical reproduction, or both” |
| R6 States the limitations of its own output | ✅ yes | “Fidelity to one paper scenario does not establish general biological validity.” |
| R7 Code examples complete enough to run | ✅ yes | “python scripts/paper2model_report.py validate –spec reproduction-spec.json” |
| R8 Examples run in continuous integration | no | No continuous integration. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | evals/ holds prompts, not outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Nova In Silico | 1.0 | 0 | none | not found: commits by a publishing bot; no authors named in the repository |
Maintenance. Last change to the skill’s files 2026-09-29 (8 days; recency 97.8). Usability 100: licence ✅, frontmatter ✅, install ✅, dependencies ✅, version ✅.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
inflexa · pkpd-clinical-response
inflexa-ai/inflexa at 79c79cb · task skill · unproven: no evidence of use, so not ranked · tags: PK/PD; exposure-response
| S | Q | E | M | U |
|---|---|---|---|---|
| 17.5 | 20.0 | 0.0 | 84.9 | 0.0 |
Quality. 2 of 10 applicable rubric items met (20.0), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | The Do NOT list is bad practice, not scope. |
| R2 Cites sources for its recommendations | no | Cites none; references are the skill’s own files. |
| R3 Says what to do when a check fails | ✅ yes | “A covariate that lowers the objective function without reducing between-subject variability on its parameter has explained nothing. Report both, with the shrinkage.” |
| R4 Qualifies numeric thresholds, or gives none | no | “Use NCA terminal half-life if the terminal phase has fewer than 3 points or R-squared < 0.8” is given bare. |
| R5 Ties evaluation depth to intended use | ✅ yes | Method chosen from the question and the sampling design: two-stage for rich data and covariate questions, NLME otherwise. |
| R6 States the limitations of its own output | no | No statement of the limits of its own output. |
| R7 Code examples complete enough to run | no | Code lives in reference files; the skill shows a decision tree. |
| R8 Examples run in continuous integration | no | No test of the skill in CI. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Robert Krajla | 0.894 | 0 | record | ORCID 0009-0009-7758-0684, from inflexa CITATION.cff |
| Radu Andrei Tanasa | 0.106 | 0 | record | ORCID 0009-0007-2817-4589, from inflexa CITATION.cff |
Maintenance. Last change to the skill’s files 2026-08-31 (37 days; recency 89.9). Usability 80: licence ✅, frontmatter ✅, install ✅, dependencies ✗, version ✅.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 33 and 4.
ai_collection · quantum-compartmental-pkpd
hiyenwong/ai_collection at ce2d037 · task skill · unproven: no evidence of use, so not ranked · tags: simulation
| S | Q | E | M | U |
|---|---|---|---|---|
| 17.3 | 23.3 | 0.0 | 67.8 | 0.0 |
Quality. 3 of 9 applicable rubric items met (33.3), less 0 major and 2 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | ✅ yes | Built on one cited source, arXiv:2605.09691. |
| R3 Says what to do when a check fails | no | No guidance when a check fails. |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | Gives no numeric thresholds. |
| R6 States the limitations of its own output | ✅ yes | “Classical ODE solvers remain more efficient for simple compartmental models” |
| R7 Code examples complete enough to run | no | The circuit reuses rotation angles and encodes no PK quantity; it does not run a PK model. |
| R8 Examples run in continuous integration | no | CI validates skill format only and is failing. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
- Minor error. “Controlled rotations model drug transfer rates” The example circuit has no link between its parameters and any rate constant.
- Minor error. “Variational optimization handles nonlinear mixed-effects models” Unsupported claim; nothing in the skill estimates between-subject variability.
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Robin (hiyenwong) | 1.0 | 0 | none | not found: no surname on the GitHub profile |
Maintenance. Last change to the skill’s files 2026-07-10 (89 days; recency 75.6). Usability 60: licence ✗, frontmatter ✅, install ✅, dependencies ✗, version ✅.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 3 and 0.
Note. Auto-generated from an arXiv paper; three near-identical copies in the same collection.
BiostatAgent · pharmacokinetics
choxos/BiostatAgent at a55c059 · task skill · unproven: no evidence of use, so not ranked · tags: PK analysis; PKNCA; mrgsolve; nlmixr2
| S | Q | E | M | U |
|---|---|---|---|---|
| 15.2 | 20.0 | 0.0 | 61.8 | 0.0 |
Quality. 2 of 10 applicable rubric items met (20.0), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | One closing line: “Follow FDA/EMA guidance for NCA intervals and BE limits”. |
| R3 Says what to do when a check fails | no | Diagnostics are plotted, not interpreted. |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | The only thresholds are the regulatory 80–125% BE limits, stated as such. |
| R5 Ties evaluation depth to intended use | no | Nothing ties the analysis to intended use. |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
| R7 Code examples complete enough to run | ✅ yes | Complete PKNCA, mrgsolve and nlmixr2 examples with inline data. |
| R8 Examples run in continuous integration | no | CI validates the marketplace packaging, not the code. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Ahmad Sofi-Mahmudi | 1.0 | 0 | record | ORCID 0000-0001-6829-0823 on OpenAlex; no field works |
Maintenance. Last change to the skill’s files 2026-05-27 (133 days; recency 63.6). Usability 60: licence ✅, frontmatter ✅, install ✅, dependencies ✗, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 12 and 1.
Note. Same skill as choxos/TidyRModelling (merged as a copy).
PK_Claude · covariate-test
momentummetrix/PK_Claude at c245ed3 · task skill · unproven: no evidence of use, so not ranked · tags: covariate modeling; NONMEM
| S | Q | E | M | U |
|---|---|---|---|---|
| 11.5 | 16.7 | 16.0 | 0 | 0.0 |
Quality. 1 of 6 applicable rubric items met (16.7), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Cites none. |
| R3 Says what to do when a check fails | ✅ yes | Forward-addition and backward-elimination decisions at stated dOFV cut-offs. |
| R4 Qualifies numeric thresholds, or gives none | no | “dOFV >= 3.84 (p < 0.05, 1 df)” and “6.63” are given without the multiplicity or clinical-relevance context. |
| R5 Ties evaluation depth to intended use | no | Nothing ties the covariate search to intended use. |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Justus Bingham | 1.0 | 4 | record | probable: the only OpenAlex author of that name with 3 or more field works |
Maintenance. Last change to the skill’s files 2026-02-16 (233 days; recency 36.2). Usability 40: licence ✗, frontmatter ✅, install ✅, dependencies ✗, version ✗. Repository archived, so M is 0.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. Repository archived.
PK_Claude · nca-analysis
momentummetrix/PK_Claude at c245ed3 · task skill · unproven: no evidence of use, so not ranked · tags: NCA
| S | Q | E | M | U |
|---|---|---|---|---|
| 9.0 | 11.1 | 16.0 | 0 | 0.0 |
Quality. 1 of 9 applicable rubric items met (11.1), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Fetches PKNCA documentation at run time; its own rules cite none. |
| R3 Says what to do when a check fails | no | “Flag any issues” only. |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | Gives no numeric thresholds. |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
| R7 Code examples complete enough to run | no | Points to a template; no runnable code in the skill. |
| R8 Examples run in continuous integration | no | No continuous integration. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Justus Bingham | 1.0 | 4 | record | probable: the only OpenAlex author of that name with 3 or more field works |
Maintenance. Last change to the skill’s files 2026-02-17 (232 days; recency 36.4). Usability 40: licence ✗, frontmatter ✅, install ✅, dependencies ✗, version ✗. Repository archived, so M is 0.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. Repository archived.
PK_Claude · simulate-pk
momentummetrix/PK_Claude at c245ed3 · task skill · unproven: no evidence of use, so not ranked · tags: simulation
| S | Q | E | M | U |
|---|---|---|---|---|
| 9.0 | 11.1 | 16.0 | 0 | 0.0 |
Quality. 1 of 9 applicable rubric items met (11.1), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Cites none. |
| R3 Says what to do when a check fails | no | No guidance when a check fails. |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | Gives no numeric thresholds. |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
| R7 Code examples complete enough to run | no | Points to a template; no runnable code in the skill. |
| R8 Examples run in continuous integration | no | No continuous integration. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Justus Bingham | 1.0 | 4 | record | probable: the only OpenAlex author of that name with 3 or more field works |
Maintenance. Last change to the skill’s files 2026-02-17 (232 days; recency 36.4). Usability 40: licence ✗, frontmatter ✅, install ✅, dependencies ✗, version ✗. Repository archived, so M is 0.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. Repository archived.
PK_Claude · model-eval
momentummetrix/PK_Claude at c245ed3 · task skill · unproven: no evidence of use, so not ranked · tags: model evaluation; NONMEM
| S | Q | E | M | U |
|---|---|---|---|---|
| 8.5 | 10.0 | 16.0 | 0 | 0.0 |
Quality. 1 of 10 applicable rubric items met (10.0), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Cites none. |
| R3 Says what to do when a check fails | no | “If convergence OR covariance failed, report immediately and flag” stops at flagging. |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | “high RSE, high shrinkage” with no numeric cut-offs. |
| R5 Ties evaluation depth to intended use | no | Nothing ties evaluation to intended use. |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
| R7 Code examples complete enough to run | no | Depends on library(aipharma), an internal package that is not public. |
| R8 Examples run in continuous integration | no | No continuous integration. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Justus Bingham | 1.0 | 4 | record | probable: the only OpenAlex author of that name with 3 or more field works |
Maintenance. Last change to the skill’s files 2026-02-16 (233 days; recency 36.2). Usability 40: licence ✗, frontmatter ✅, install ✅, dependencies ✗, version ✗. Repository archived, so M is 0.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. Repository archived.
PK_Claude · er-analysis
momentummetrix/PK_Claude at c245ed3 · task skill · unproven: no evidence of use, so not ranked · tags: exposure-response
| S | Q | E | M | U |
|---|---|---|---|---|
| 8.5 | 10.0 | 16.0 | 0 | 0.0 |
Quality. 1 of 10 applicable rubric items met (10.0), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Cites none. |
| R3 Says what to do when a check fails | no | No guidance when a check fails. |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | Gives no numeric thresholds. |
| R5 Ties evaluation depth to intended use | no | Nothing ties the analysis to intended use. |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
| R7 Code examples complete enough to run | no | Fragments such as glm(…, family = binomial) only. |
| R8 Examples run in continuous integration | no | No continuous integration. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Justus Bingham | 1.0 | 4 | record | probable: the only OpenAlex author of that name with 3 or more field works |
Maintenance. Last change to the skill’s files 2026-02-17 (232 days; recency 36.4). Usability 40: licence ✗, frontmatter ✅, install ✅, dependencies ✗, version ✗. Repository archived, so M is 0.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. Repository archived. Silent on confounding between exposure and prognosis.
satori · scientific-clinical-pharmacology
nahisaho/satori at 20787d9 · task skill · unproven: no evidence of use, so not ranked · tags: clinical pharmacology; population PK
| S | Q | E | M | U |
|---|---|---|---|---|
| 3.8 | 0 | 0.0 | 37.9 | 0.0 |
Quality. 2 of 10 applicable rubric items met (20.0), less 1 major and 1 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | no | No statement of what the skill excludes. |
| R2 Cites sources for its recommendations | no | Cites none. |
| R3 Says what to do when a check fails | no | No guidance when a check fails. |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | Gives no numeric thresholds. |
| R5 Ties evaluation depth to intended use | no | Nothing ties the analysis to intended use. |
| R6 States the limitations of its own output | no | No statement of the limits of its output. |
| R7 Code examples complete enough to run | ✅ yes | Complete Python functions. |
| R8 Examples run in continuous integration | no | CI runs the package’s JavaScript tests, not the skill. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | No stored outputs. |
- ❌ Major error. popk_estimation(): “eta_cl = 0.0 / eta_v = 0.0” while returning “omega_cl” and “omega_v” The between-subject variances never enter the objective function, so the function is a pooled fit labelled as NLME and its omega estimates are arbitrary.
- Minor error. simulate_pk(): “t_half = 0.693 * params.get("v", params.get("v1", 50)) / params["cl"]” for the 2-compartment model For a two-compartment model 0.693·V1/CL is not the terminal half-life.
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Hisaho Nakata | 1.0 | 0 | none | not found: the only OpenAlex author of that name is a physicist, affiliation does not match |
Maintenance. Last change to the skill’s files 2026-02-15 (234 days; recency 35.9). Usability 40: licence ✗, frontmatter ✅, install ✅, dependencies ✗, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 3 and 2.
Note. Known-answer fixture 2. Skill text is in Japanese.
AI-assisted-MIDD · pk-data-validation
nlmixr2/AI-assisted-MIDD at 3f07128 · task skill · unproven: no evidence of use, so not ranked · tags: dataset assembly; NCA; nlmixr2
| S | Q | E | M | U |
|---|---|---|---|---|
| 33.9 | 44.4 | 24.0 | 78.9 | 0.0 |
Quality. 4 of 9 applicable rubric items met (44.4), less 0 major and 0 minor errors.
| Item | Evidence | |
|---|---|---|
| R1 States what the skill does not do | ✅ yes | “Mapping the validated columns to PKNCA objects is covered by pk-nca (references/pipeline.md); to an nlmixr2 model by poppk-estimation.” |
| R2 Cites sources for its recommendations | no | No recommendation names a source. |
| R3 Says what to do when a check fails | ✅ yes | “If a check fails, fix the mapping, or tell the user what is wrong in the data. Never drop, loosen or skip a check to get a pass without the user’s agreement” |
| R4 Qualifies numeric thresholds, or gives none | ✅ yes | Gives no thresholds of its own: “Never add a limit the user or protocol did not give.” |
| R6 States the limitations of its own output | no | Lists what each check catches, but not what a passing validation leaves unchecked. |
| R7 Code examples complete enough to run | ✅ yes | describe_pk_data() and validate_pk_data() calls, with scripts/pk_data_checks.R. |
| R8 Examples run in continuous integration | no | The repository has no continuous integration. |
| R9 Continuous integration passes on the pinned commit | no | Follows R8. |
| R10 Stores outputs from a real run | no | The README describes worked examples under examples/abc-111 and examples/gs-12345 with outputs passing QC, but neither folder is in the repository. |
Expertise.
| Author | Share | Field h-index | OpenAlex | Match |
|---|---|---|---|---|
| Mutaz M. Jaber | 1.0 | 6 | record | ORCID 0000-0002-7536-8753, from the DESCRIPTION of Mutaz94/mypkg; sole committer |
Maintenance. Last change to the skill’s files 2026-10-01 (8 days; recency 97.8). Usability 60: licence ✗, frontmatter ✅, install ✅, dependencies ✅, version ✗.
Use. No issue, pull request or comment from anyone outside the authors in the last 12 months, and no deliberate installation found in another project.
Repository stars and forks, used only to break ties: 1 and 0.
Note. Added 2026-10-09, outside the monthly run, at Andy’s request; discovery had missed the repository, created 2026-10-01. Four of its 18 skills were scored. The other 14 are the project, database, logging, run-spec, table, figure, report and QC plumbing around these four, and were not scored: pk-nca-database, pk-nca-figures, pk-nca-logging, pk-nca-qc-tests, pk-nca-report, pk-nca-run-spec, pk-nca-tables, poppk-database, poppk-qc-tests, poppk-report, poppk-run-spec, poppk-tables, pk-project, manage-skills.