scientific-manuscript-writing

Develop, revise, audit, and format high-level scientific manuscripts, Supplementary Information, captions, figures, and publication Word files when evidence-led storytelling, contribution framing, cross-document consistency, or journal-ready visual design is required. Do not use for ordinary reports, essays, or administrative documents.

¿Qué es scientific-manuscript-writing?

scientific-manuscript-writing is a Claude Code agent skill that develop, revise, audit, and format high-level scientific manuscripts, Supplementary Information, captions, figures, and publication Word files when evidence-led storytelling, contribution framing, cross-document consistency, or journal-ready visual design is required. Do not use for ordinary reports, essays, or administrative documents.

Compatible con~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/yizheng-hub/scientific-manuscript-writing-skill/tree/HEAD/skills/scientific-manuscript-writing

Installed? Explore more Investigación y análisis de datos skills: affaan-m/uspto-database, affaan-m/scholar-evaluation, affaan-m/literature-review · View all 6 →

Preguntar en tu IA favorita

Abre un nuevo chat con esta habilidad de agente ya precargada.

Documentación

Scientific Manuscript Writing

Turn the supplied evidence into a clear scientific claim, a figure-led evidence chain, a self-contained Supplementary Information document, and publication-ready files. Follow the target journal and supplied templates first. Never transfer project-specific terms, results, or assumptions into a new study.

Route the task

Core workflow

  1. Read the journal rules, template, research materials, data, code outputs, current figures, and references. Record missing or unverified inputs without inventing them.
  2. State one central scientific claim and map each supporting finding to its analysis, figure panels, innovation, and contribution. Resolve evidence problems before polishing prose.
  3. Freeze the verified evidence base. Do not use a preferred direction, ranking, or significance result as the condition for accepting an analysis.
  4. Organize Results around scientific findings and figures. Cite and interpret every panel in order. Use conceptual or framework panels only when a Methods overview genuinely needs them; Results panels must carry data or analysis.
  5. Build Title, Abstract, Introduction, Discussion, SI, captions, and any cover letter around the same contribution. Keep claims, numbers, terminology, units, colors, and numbering synchronized.
  6. For a full manuscript or journal-ready package, run the required cold-read roles as separate agents or isolated contexts when orchestration is available; do not blend them into one simulated multi-role review. Integrate only after independent reports are frozen, fix supported issues, and repeat affected technical, field-expert, reader, visual, and publication-file checks until they pass.

Non-negotiable standards

  • Do not fabricate data, methods, results, references, or certainty. Mark unresolved facts for verification.
  • Write a discovery-led paper, not an academic project report. Remove code variable and column names, script/notebook/cell identifiers, debugging chronology, discarded-trial narration, internal audit labels, AI-review language, and hypothetical reviewer rebuttals from narrative prose and figure labels. Preserve final stopping/selection rules, negative controls, convergence or replication failures, robustness evidence, exclusions, and protocol deviations when they affect validity. Where exact identifiers are needed in an SI data dictionary or reproducibility mapping, pair each one with a human-readable scientific term.
  • Do not scatter defensive sentences after findings or repeatedly state what the study does not prove. Keep real study-level limitations brief and integrated at the end of Discussion. Preserve conditions or uncertainty that materially change a claim.
  • Treat non-positive or null results as evidence when they change the scientific interpretation. State the resulting insight; do not hide it or narrate it as a failed experiment.
  • Use simple, idiomatic English. Lead with the innovation and finding, retain only decision-relevant numbers, and explain the contribution before technical detail overwhelms it.
  • Define abbreviations independently in the main text, SI, and every figure caption that must stand alone.
  • Make captions objective and self-contained; keep interpretation, mechanism, contribution, and extended analysis in the main or SI narrative.
  • Do not create Extended Data as part of the default general deliverable. Use the main manuscript and SI unless the user separately requests a journal-specific structure.
  • Preserve user edits and template conventions. Make surgical changes, then render and visually inspect every page and every figure at final size.

Skills relacionados