Section-by-section academic manuscript writing assistant for quantitative research. Version 1.0.0.
Scientific Paper Writer is a Claude plugin that guides researchers through writing a complete manuscript in expert writing order -- starting with Methods, then Results, Discussion, and frontmatter last. It supports 11 study types, verifies every citation against multiple databases, and exports to LaTeX or DOCX with venue-specific formatting.
Key principles:
- No fabrication -- all statistics are extracted directly from your analysis outputs, figures, and tables. Values that cannot be verified are flagged with
[PLEASE VERIFY]. - Guideline-aligned -- every section follows the appropriate reporting guideline (PRISMA 2020, STROBE, CONSORT, TRIPOD+AI, STARD 2015, CHEERS 2022, COREQ/SRQR, PRISMA-NMA, PRISMA-IPD).
- Expert writing order -- Methods first, Abstract and Title last, following the order that experienced researchers use.
- Citation verification -- every external citation is verified against PubMed, CrossRef, Semantic Scholar, and bioRxiv/medRxiv before inclusion. Unverifiable citations are always flagged, never silently included.
claude plugin install github:olaTechie/scientific-paper-writerDownload the zip from the Releases page, then:
claude plugin install ./scientific-paper-writer-v1.0.0.zipgit clone https://github.com/olaTechie/scientific-paper-writer.git
claude plugin install ./scientific-paper-writerType / in a Claude session. You should see all 8 commands listed below.
- Create a
PAPER_CONTEXT.mdfile in your working directory (see template below) - Follow the expert writing order:
/write-methods Write Methods aligned with the reporting guideline
/write-results Write Results from your analysis outputs
/write-discussion Write Discussion with verified literature citations
/write-frontmatter Write Abstract, Introduction, and Title
/write-captions Write figure and table captions
/write-references Build and verify the reference list
/generate-figures Generate publication-quality figures (R/Python)
/export-manuscript Export to LaTeX or DOCX
| Command | Description | Example |
|---|---|---|
/write-methods [study-type] |
Write the Methods section aligned with the reporting guideline for your study type | /write-methods systematic-review |
/write-results [study-type] |
Write the Results section, extracting all numbers from source materials | /write-results rct |
/write-discussion [section] |
Write the Discussion with literature comparison and citation verification | /write-discussion main-findings |
/write-frontmatter [section] |
Write Abstract, Introduction, and Title after the body is complete | /write-frontmatter abstract |
/write-captions [folder-path] |
Generate publication-ready captions for figures and tables | /write-captions ./figures/ |
/write-references [style] |
Verify citations and format the reference list | /write-references vancouver |
/generate-figures [type] |
Generate figures using R or Python scripts | /generate-figures forest-plot |
/export-manuscript [format] [venue] |
Export the manuscript to LaTeX and/or DOCX | /export-manuscript latex jama |
| Identifier | Reporting Guideline | Key Features |
|---|---|---|
systematic-review |
PRISMA 2020 | PRISMA flow, risk of bias, pooled estimates, heterogeneity, subgroup/sensitivity analyses |
network-meta-analysis |
PRISMA-NMA | Network geometry, inconsistency testing, SUCRA/P-score rankings, league tables |
ipd-meta-analysis |
PRISMA-IPD | One-stage/two-stage models, participant-level subgroups, missing data handling |
dose-response |
PRISMA 2020 + dose-response extensions | Restricted cubic splines, non-linearity testing, threshold detection |
bayesian-meta-analysis |
PRISMA 2020 + Bayesian extensions | Prior specification, MCMC diagnostics, posterior summaries, model comparison |
rct |
CONSORT 2010 | CONSORT flow, ITT/PP analyses, primary/secondary outcomes, adverse events |
observational |
STROBE | Cohort/case-control/cross-sectional, crude + adjusted estimates, bias assessment |
economic-evaluation |
CHEERS 2022 | Cost-effectiveness analysis, ICER, sensitivity analyses, CEAC, budget impact |
prediction-model |
TRIPOD+AI | Model development, calibration, discrimination, internal/external validation |
diagnostic-accuracy |
STARD 2015 | STARD flow, sensitivity/specificity, ROC analysis, likelihood ratios |
qualitative |
COREQ / SRQR | Thematic/framework analysis, reflexivity, trustworthiness, transferability |
| Format | Description |
|---|---|
| Markdown | Default output for all writing commands. Clean .md ready for manuscript editing. |
| LaTeX (generic) | Standard article template with line numbers and double spacing. |
| LaTeX (JAMA) | JAMA Network template with Key Points box and structured abstract sections. |
| LaTeX (Lancet) | Lancet template with Research in Context panel. |
| DOCX | Pandoc-based export with professional reference template (Times New Roman, double-spaced). |
Every citation passes through a 4-source verification pipeline before inclusion:
- PubMed (first) -- search by title or author+year. Retrieve PMID, DOI, full metadata.
- CrossRef (second) -- query the CrossRef API. Match by title similarity (>85% threshold).
- Semantic Scholar (third) -- query the Semantic Scholar Graph API. Extract paper ID, DOI, venue.
- bioRxiv/medRxiv (fourth) -- search preprints. Check whether a published version exists.
If a citation cannot be verified through any source, it is flagged:
[UNVERIFIED -- not found in PubMed, CrossRef, Semantic Scholar, or bioRxiv. Please check manually]
Citations are never fabricated. Missing metadata fields are marked (e.g., [year unknown], [DOI unknown]).
Create this file in your working directory to pre-load study context across all commands.
# Paper Context
## Study Details
- Study type: systematic review and meta-analysis
- Primary outcome: vaccine effectiveness against laboratory-confirmed pertussis
- Population: adults aged >=18 years
- Target journal: JAMA Network Open
- Citation style: Vancouver
- Abstract word limit: 350
## Folder Structure
- Figures folder: ./figures/
- Draft report: ./drafts/report.pdf
- Analysis outputs: ./results/
## Conventions
- UK English spelling throughout
- VE% = (1 - OR) x 100
- Report heterogeneity: I-squared, tau-squared, Q, 95% prediction interval
- Distinguish absolute VE from relative VEScientific Paper Writer works standalone but can optionally enhance its capabilities when the claude-scientific-writer plugin is also installed.
When detected, the integrations engine:
- Uses enhanced literature search capabilities if available
- Coordinates shared PAPER_CONTEXT.md between plugins
- Avoids duplicating functionality that the companion plugin provides
Detection is automatic and graceful -- the plugin works identically whether or not the companion is installed.
Experienced researchers write papers in this order, not the reading order:
1. Methods /write-methods What you did (most concrete)
2. Results /write-results What you found (from outputs)
3. Discussion /write-discussion What it means (needs Results)
4. Abstract /write-frontmatter Summary (needs all sections)
5. Introduction /write-frontmatter Context (needs Abstract)
6. Title /write-frontmatter Distillation (needs everything)
7. Captions /write-captions Standalone figure descriptions
8. References /write-references Final verification pass
This order ensures each section is written with the information it depends on already complete.
scientific-paper-writer/
.claude-plugin/
plugin.json
commands/
write-methods.md
write-results.md
write-discussion.md
write-frontmatter.md
write-captions.md
write-references.md
generate-figures.md
export-manuscript.md
skills/
core/ Core prose conventions and verification flags
citations/ Multi-source citation verification engine
visual/ Figure interpretation and caption generation
output/ LaTeX, DOCX, and abstract format engine
integrations/ Optional plugin detection
study-types/
systematic-review/ PRISMA 2020
network-meta-analysis/ PRISMA-NMA
ipd-meta-analysis/ PRISMA-IPD
dose-response/ Dose-response meta-analysis
bayesian-meta-analysis/ Bayesian meta-analysis
rct/ CONSORT 2010
observational/ STROBE
economic-evaluation/ CHEERS 2022
prediction-model/ TRIPOD+AI
diagnostic-accuracy/ STARD 2015
qualitative/ COREQ / SRQR
templates/
latex/ LaTeX templates (generic, JAMA, Lancet)
figures/ R script templates (forest, PRISMA, funnel)
docx/ DOCX reference template
hooks/
hooks.json
build-release.sh
LICENSE
README.md
No external API keys required. The plugin uses Claude's built-in web search, PubMed tools, and bioRxiv tools for citation verification and literature search.
For figure generation, R (with metafor, DiagrammeR packages) or Python (with matplotlib) should be available on your system. If neither is available, the plugin generates script files for manual execution.
For DOCX export, pandoc is required: brew install pandoc (macOS) or apt install pandoc (Linux).
MIT License. See LICENSE for details.
