Claude Scientific Writing Skill Guide

The prompt

Use the scientific-writing skill to draft an abstract for ./manuscript.md. Target 250 words, one flowing paragraph, NO labeled sections (no 'Background:' / 'Methods:'). Pull the primary endpoint and effect size from @results.md. Open with the gap, state the design, give the headline number with its 95% CI, end with the one-sentence implication. Avoid undefined abbreviations on first use.

What slides.md looks like

## Abstract

Wearable-derived sleep metrics are rarely validated against
polysomnography in free-living adults, leaving clinicians
uncertain how to interpret them. We conducted a prospective
cross-sectional study of 312 adults who wore a consumer
device for one night of in-lab polysomnography. Total sleep
time agreed closely (mean bias 6.4 min; 95% CI -3.1 to 15.9),
but deep-sleep duration was overestimated by 22 min on
average. These findings suggest aggregate sleep duration is
trustworthy for population screening, whereas stage-level
output should not yet guide individual diagnosis.

Pairs with

The prompt

Use the scientific-writing skill to write the Methods section from ./protocol-notes.md. Full paragraphs, past tense, passive where it reads naturally. Cover: design, setting, participants with inclusion/exclusion, measures and instruments (with versions), and the statistical plan (the model, the software, the alpha). Keep the inclusion/exclusion criteria as a short list — Methods is the one place lists are allowed. Define every abbreviation at first use.

What slides.md looks like

## Methods

### Study design and setting
We conducted a prospective cross-sectional validation study
at a single academic sleep laboratory between January and
August 2025. The protocol was approved by the institutional
review board (#2024-1187), and all participants provided
written informed consent.

### Participants
Adults were eligible if they were 18-65 years old and
reported habitual sleep of 6-9 hours. We excluded:

- diagnosed sleep disorders
- shift work in the prior 3 months
- current sedative-hypnotic use

Pairs with

The prompt

Use the scientific-writing skill to draft the Results section from @analysis-output.csv and @tables.md. Report findings in flowing paragraphs in hypothesis order. Reference 'Table 2' and 'Figure 1' rather than repeating their numbers. Quote only the primary effect, its CI, and the test statistic inline. NO bullet points in Results. Do not interpret — interpretation belongs in the Discussion.

What slides.md looks like

## Results

Of 340 enrolled adults, 312 (91.8%) completed the in-lab
night and contributed analyzable recordings (Table 1).
Device-derived total sleep time agreed closely with
polysomnography, with a mean bias of 6.4 minutes
(95% CI, -3.1 to 15.9; Figure 1). Agreement was weaker for
sleep staging: the device overestimated deep-sleep duration
by a mean of 22.0 minutes (95% CI, 17.4 to 26.6), and the
intraclass correlation for REM duration was modest
(ICC 0.51; Table 2). Subgroup estimates by age band are
reported in Supplementary Table S3.

Pairs with

The prompt

Use the scientific-writing skill to write a Discussion from ./manuscript.md. Structure: (1) restate the principal finding in one sentence, (2) compare with prior work using @refs.bib, (3) mechanism or interpretation, (4) limitations stated plainly, (5) implications and one next study. Full paragraphs only. Do not overstate causation — this is a cross-sectional design. End on the clinical implication, not the limitations.

What slides.md looks like

## Discussion

In this validation of a consumer wearable against
polysomnography, aggregate sleep duration was accurate but
stage-level estimates were not. This pattern is consistent
with prior actigraphy work [12,14] yet contrasts with the
manufacturer's stage-accuracy claims. A plausible mechanism
is that single-channel accelerometry captures gross
immobility well but lacks the EEG signal that defines deep
sleep. This study has limitations. Its cross-sectional,
single-night design cannot establish night-to-night
reliability, and the laboratory setting may not generalize
to home use. For screening, the device is fit for purpose;
for staging, it is not yet ready.

Pairs with

The prompt

Use the scientific-writing skill to convert the citations in ./manuscript.md from APA author-date to AMA numbered superscript. Renumber in order of first appearance, convert every in-text (Author, Year) to a superscript number, and rebuild the reference list in AMA order. Flag any reference missing a DOI or page range so I can fix it. Do not invent missing fields — list them instead.

What slides.md looks like

<!-- In-text: AMA numbered superscript -->
Wearable validation against polysomnography remains
limited.^1,2 Aggregate duration is generally accurate.^3

## References

1. de Zambotti M, Cellini N, Goldstone A, et al. Wearable
   sleep technology in clinical and research settings.
   Med Sci Sports Exerc. 2019;51(7):1538-1557.
2. ...

<!-- Needs review (missing fields): -->
- Ref 7: no DOI found
- Ref 11: page range incomplete

Pairs with

The prompt

Use the scientific-writing skill to reformat ./manuscript.md for submission to a target journal. The guidelines: 3,000-word main text, structured abstract (250 words), Vancouver references, separate title page with author affiliations, and one combined Methods+Materials section. Produce a compliance checklist of what changed and flag any section now over the word limit. Keep the science identical — only restructure.

What slides.md looks like

## Reformat summary

| Requirement            | Status   | Note                    |
|------------------------|----------|-------------------------|
| Main text <= 3000 words| OVER     | 3,180 — trim Discussion |
| Structured abstract    | DONE     | 4 labeled sections      |
| Vancouver references   | DONE     | renumbered [1]-[31]     |
| Title page split out   | DONE     | title-page.md created   |
| Methods + Materials     | DONE     | merged into one section |

Action: cut ~180 words from the Discussion's prior-work
paragraph to meet the limit.

Pairs with

The prompt

Use the scientific-writing skill to write a one-page cover letter to the Editor-in-Chief for ./manuscript.md submitted to a clinical journal. Cover: the gap, our one-sentence finding, why it fits this journal's readership, and the standard attestations (original work, not under review elsewhere, all authors approved, no undisclosed conflicts). Professional tone, no hyperbole. Leave the editor name and date as placeholders.

What slides.md looks like

Dear [Editor-in-Chief],

We are pleased to submit "Validation of a Consumer Wearable
Against Polysomnography in Free-Living Adults" for
consideration as an Original Article.

Clinicians increasingly encounter patient-reported wearable
sleep data but lack validation evidence to interpret it. In
312 adults, we found aggregate sleep duration was accurate
while stage-level estimates were not — a distinction directly
relevant to your readership of sleep clinicians.

This manuscript is original, is not under consideration
elsewhere, and has been approved by all authors. The authors
report no conflicts of interest.

Pairs with

The prompt

Use the scientific-writing skill to draft a point-by-point response to reviewers from ./reviewer-comments.md and the revised ./manuscript.md. For each comment: quote it verbatim, write a courteous reply, state exactly what changed, and cite the new line or section. Where we disagree, explain the rationale respectfully rather than only complying. Group by reviewer. Keep the tone collegial even on the harsh comments.

What slides.md looks like

# Response to Reviewers

## Reviewer 1

**Comment 1.1.** "The single-night design cannot speak to
night-to-night reliability."

*Response.* We agree and thank the reviewer. We have added
this explicitly as a limitation (Discussion, para. 4,
"its cross-sectional, single-night design cannot establish
night-to-night reliability") and softened the abstract's
claim accordingly.

**Comment 1.2.** "Consider reporting Bland-Altman limits of
agreement."

*Response.* Added. Limits of agreement now appear in
Figure 1 and are quoted in Results, para. 2.

Pairs with

The prompt

Use the scientific-writing skill to write a 150-word plain-language summary of ./manuscript.md for a general audience (around an 8th-grade reading level). No undefined jargon, no statistics notation. Explain what we asked, what we found, and why it matters for someone wearing a sleep tracker. Do not overstate — be clear that staging is the part that isn't accurate yet. One short paragraph.

What slides.md looks like

## Plain-language summary

Millions of people wear watches and rings that claim to
track their sleep, but few studies have checked those numbers
against the gold-standard test done in a sleep lab. We
compared one popular device with the lab test in 312 adults.
The device was good at estimating how long people slept
overall — close enough to be useful for general health
checks. But it was not accurate at telling apart the
different stages of sleep, such as deep sleep. So if your
tracker says you got a certain amount of deep sleep, treat
that number with caution. The total hours, though, are
probably about right.

Pairs with

The prompt

Use the scientific-writing skill to write captions for every figure and table in ./manuscript.md. Each caption must be self-contained: state what is shown, the sample size, what the error bars or shading represent, and define every abbreviation used in that exhibit. Match the journal's style (sentence case, bold figure number). Do not duplicate Results prose — the caption describes the exhibit, the text interprets it.

What slides.md looks like

**Figure 1.** Agreement between device-derived and
polysomnography total sleep time (n = 312). Bland-Altman
plot; the solid line shows mean bias (6.4 min) and dashed
lines the 95% limits of agreement (-58 to 71 min). Each
point is one participant. PSG, polysomnography.

**Table 2.** Stage-level agreement by sleep stage (n = 312).
Values are intraclass correlation coefficients (ICC) with
95% confidence intervals. ICC < 0.50 indicates poor
agreement. REM, rapid eye movement.

Pairs with