Agent Skills
› brycewang-stanford/Awesome-Journal-Skills
› lancet-writing
lancet-writing
GitHub用于撰写或精简《柳叶刀》原创研究文章，确保符合IMRaD结构及约3000-3500字篇幅限制。规范讨论部分的谨慎表述与四步逻辑，控制参考文献数量，并遵循英式英语及全球视角的行文风格。
Trigger Scenarios
起草或修改《柳叶刀》原创研究文章正文
正文超字数或讨论部分夸大结论
文章结构偏离IMRaD标准
作者不断添加内容导致篇幅超标
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill lancet-writing -g -y
SKILL.md
Frontmatter
{
"name": "lancet-writing",
"description": "Use to structure and hold the length of a Lancet Article main text — Introduction, Methods, Results, Discussion (~3000–3500 words), the Research in context panel, a limited reference list (~30), and a cautious, globally minded Discussion that does not overstate."
}
Main-Text Writing & Structure (lancet-writing)
When to trigger
- Drafting or trimming a Lancet original-research Article.
- The main text is over length, or the Discussion overstates the findings.
- The structure wanders (lab/clinic chronology instead of IMRaD).
- Co-authors keep adding paragraphs and references and the paper is creeping past budget.
Article structure (IMRaD + the panel)
The Lancet original-research Article uses standard IMRaD with the Research in context panel:
| Section | Holds |
|---|---|
| Introduction | The problem, the gap, and the specific objective — brief (the systematic search lives in the panel, not a long literature review). |
| Methods | Design, setting/countries, participants, randomisation/allocation/blinding (or observational design), outcomes, sample size, statistical analysis (pre-specified), registration, ethics. Methods are detailed and stay in the main text (clinical journals expect full Methods, unlike Science's supplement model). |
| Results | Recruitment/flow (with the flow diagram), Table 1, primary outcome, secondary outcomes, subgroups, harms — in a logical order. |
| Discussion | Principal findings → comparison with other studies → strengths and limitations → implications. |
| Research in context panel | The mandatory three-part box (see lancet-research-in-context). |
Length and reference budget
- Main text ~3000–3500 words (Articles; confirm the current cap — formats and limits change).
- References limited (~30) for an Article — cite the key and the systematic-search-anchored literature, not everything.
- Tables/figures within the journal's display-item allowance (see
lancet-figures-tables); extended methods/results → appendix. - An appendix / supplementary material carries the protocol summary, additional analyses, full subgroup tables, and the reporting checklist.
The Discussion — four moves, cautiously
Write the Discussion in this order, and keep it cautious:
- Principal findings — restate the main result plainly, without re-listing every number.
- Comparison with other studies — situate against prior evidence (consistent with the panel's systematic search); explain agreement/disagreement.
- Strengths and limitations — be candid; name confounding, generalisability, missing data, power.
- Implications — for practice, policy, and research, with calibrated causal language.
The Lancet prizes caution, not overstatement. Match claims to design: an observational study shows association, not causation; a single trial rarely "proves" — it "supports" or "provides evidence for." A non-inferiority trial cannot claim superiority.
House style
- Accessible, globally minded prose: avoid US-only framing; define context for an international readership; spell out abbreviations on first use.
- British spelling is the house style; use the journal's conventions for numbers, units (SI), and dates.
- Lead Results paragraphs with the clinical finding, then the supporting statistics.
What Lancet editors expect from the main text
The Lancet's editors read for a globally minded, cautious voice: a brief Introduction (the systematic search lives in the panel), full Methods in the main text, and a Discussion judged on restraint — observational studies show association, a single trial supports rather than proves, a non-inferiority trial never claims superiority. Patterns reviewers flag: a literature-review Introduction; Methods pushed offstage; "proves"/superiority overstatement; an over-budget reference list. Confirm current limits in the author guidelines.
Worked micro-example (illustrative numbers — not real data)
A hypothetical Lancet Article on a multi-country observational cohort.
Length audit (illustrative):
Main text 3 980 words -> over ~3 500 by ~480; trim Introduction and Discussion
References 47 -> over ~30; keep the load-bearing 30
Claim calibration:
DRAFT: "This proves the exposure causes the outcome."
FIX: "The exposure was associated with the outcome (adjusted HR 1.34,
95% CI 1.18-1.52); residual confounding cannot be excluded."
The fix trims to budget and recasts the overstatement as a calibrated association (95% CI).
Reviewer-pushback patterns and the venue-specific fix
- "The Introduction reads as a literature review." → Cut it to problem-gap-objective; move synthesis to the panel.
- "The conclusions overstate the design." → Replace "proves"/causal language with "associated with"/"supports"; remove superiority framing from a non-inferiority trial.
- "The paper is over length / reference budget." → Trim to load-bearing claims and ~30 citations; push extended analyses to the appendix.
Output format
【Sections present】 Introduction / Methods / Results / Discussion / Research-in-context panel — all? yes/no
【Methods in main text?】 yes (good) / pushed to appendix (FIX)
【Main-text word count】 N → over/under ~3000–3500 by M
【Reference count】 N → vs ~30 budget
【Discussion four moves】 principal findings / comparison / strengths-limitations / implications — all? yes/no
【Overstatement check】 causal language calibrated to design? yes/no
【Next】 lancet-ethics
Anti-patterns
- Do not write a long literature-review Introduction — the systematic search belongs in the panel.
- Do not overstate: no causal claims from observational designs, no "proves," no superiority claim from a non-inferiority trial.
- Do not let the Discussion re-list Results numbers instead of interpreting them.
- Do not blow the reference budget by citing exhaustively; cite the load-bearing literature.
Version History
- 1839142 Current 2026-07-05 14:02
