Conciseness

The body is dense, expert-level prose with no padding and no explanation of concepts Claude already knows — every paragraph carries non-obvious domain judgment ("AIC picks the three-compartment model. BIC and the F test both reject it"). It falls short of the score 5 anchor only because some inline detail could move to the bundle: the Software ecosystem section duplicates per-version, dated changelog material ("2.0.0 (2026-02-12) changed dataset row indices to start at 1", the NONMEM 7.5-to-7.6 feature list) that already lives in references/software-ecosystem.md, and the closing citation protocol (arXiv fetch/version-resolution rules) runs long. It is well above score 3, which requires unnecessary explanation or looseness throughout.

Actionability

Every workflow ships a copy-paste-ready command with real flags and arguments ("python3 nca.py -i profile.csv --dose 100 --route extravascular --partial-auc 0-24", the simulate/two-scenario invocation, the tdm_bayes invocation), a script-to-question table routes users to the right tool, I/O conventions are uniform ("All take --format table|tsv|json... Exit code is 0 for no findings, 1 when findings were raised, 2 for bad input"), and real recorded outputs illustrate each common case. This matches the score 5 anchor: fully executable commands covering the common cases, with no gaps.

Workflow Clarity

The Workflow section sequences nine numbered analyses, each with explicit validation checkpoints: "Check the dataset before running anything" for popPK, RSE/correlation flags after fitting ("Q3 has 98% RSE - not estimable"), attainment fractions after simulation, and NCA extrapolation/window findings. Findings plus exit codes create the validate-then-decide feedback loop ("any of them can gate a workflow"), and the closing ten-item checklist enumerates the failure modes the skill exists to prevent. This matches the score 5 anchor (clear sequence, explicit validation, feedback, checklist); none of these analyses are destructive or batch operations, so no cap applies.

Progressive Disclosure

The body is a true overview that pushes depth into 14 one-level-deep reference files (all verified to exist and contain substantive content), 2 assets, and 11 scripts, with no nested references. Navigation is well-signaled both in a dedicated References section with a scope line per file ("references/population-pk.md — NLME estimation, covariate building, BLQ M1-M7, diagnostics, VPC") and at point of need in the text ("see references/population-pk.md for estimation methods"). This matches the score 5 anchor: clear overview, well-signaled one-level-deep references, easy navigation.

Specificity

The description enumerates concrete capability areas — "non-compartmental analysis, compartmental and population PK, PK/PD and exposure-response, TMDD, PBPK orientation, bioequivalence, allometric scaling and first-in-human dose, drug interaction prediction, and Bayesian therapeutic drug monitoring" — which map one-to-one onto the skill's nine bundled workflows. This matches the anchor "lists multiple specific concrete actions; comprehensive coverage"; it is not score 4 because there is no meaningful coverage gap, and not score 3 because far more than 1-2 actions are named.

Completeness

It explicitly answers both questions: the "what" in the opening enumeration of capabilities, and the "when" via "Use when analysing concentration-time data, deriving exposure metrics, fitting PK or PD models, or evaluating dosing regimens" followed by a concrete "Triggers include..." list. This is the score 5 pattern (clear what AND when with concrete trigger phrases); the score 4 anchor ("when could be more explicit") does not apply because the when-clause is fully explicit.

Trigger Term Quality

Roughly 35 natural trigger terms including synonyms users actually say: "population PK"/"popPK", "drug-drug interaction"/"DDI", "pharmacokinetics", "AUC", "Cmax", "lambda z", "half-life", "clearance", "volume of distribution", plus tool names ("NONMEM", "nlmixr2", "Pharmpy", "Monolix") a user would name directly. This is comprehensive natural-term coverage with synonyms, matching the score 5 anchor; no common phrasing in the domain is missing.

Distinctiveness Conflict Risk

A deep pharmacometrics niche with unambiguous, domain-exclusive triggers ("RSABE", "ABEL", "MABEL", "lambda z", "TMDD", "ICH M12", "MIPD") that no adjacent skill (general statistics, chemistry, data analysis) would fire on. Matches the score 5 anchor "clear niche with distinct triggers; minimal conflict risk"; overlap risk with generic skills is negligible.