Lean body that assumes Claude's competence — no definitional padding (no explaining what p values or journals are) and every list item earns its place; the only length comes from justified ready-to-paste output templates.

As an instruction/reporting skill it provides concrete decision rules: an AUTHOR_INPUT_NEEDED protocol, explicit Red lines (do not invent p values, sample sizes, corrections), a copy-pasteable labeled output template, and condition-specific reference routing — absence of code is not penalized since guidance is actionable.

A clear 9-step numbered workflow with explicit final-QA checkpoint (step 9 via reviewer-checklist.md), conditional routing to specific references (steps 5-7), and a bounded-audit instruction when input is partial.

SKILL.md is a concise overview with a Related-files table giving per-file "Open when" triggers; all five referenced files exist and are one level deep with no nested cross-references, matching the clear-overview anchor.

Names multiple concrete actions ("Audit, revise, or draft manuscript statistical reporting") across a defined domain, matching the anchor for listing several specific concrete actions.

Explicitly answers what (audit/revise/draft manuscript statistical reporting) and when ("Use when the user asks to check statistical analysis sections..." / "Also trigger on...") with explicit trigger guidance.

Rich natural-term coverage (p values, confidence intervals, sample size, biological versus technical replicates, randomization, blinding, multiple-comparison correction, reviewer comments about statistics) plus Chinese triggers, all phrased as a user would say them.

Narrow, specialized niche (manuscript statistical reporting for Nature / high-impact journals) with distinct triggers unlikely to fire for unrelated skills; uses third-person voice throughout.