Community라이팅 & 에디팅github.com

manuscript-figures

Plan, create or revise scientific manuscript figures, panels, tables and captions, including reproducible static plots. Use for manuscript artwork and display legends; exclude interactive Crux bundles and unrelated visualizations.

manuscript-figures란 무엇인가요?

manuscript-figures is a Claude Code agent skill that plan, create or revise scientific manuscript figures, panels, tables and captions, including reproducible static plots. Use for manuscript artwork and display legends; exclude interactive Crux bundles and unrelated visualizations.

지원 대상~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/zachary-zzc/manuscript-writing-skills/tree/HEAD/skills/manuscript-figures

Installed? Explore more 라이팅 & 에디팅 skills: steipete/notion, langchain-ai/langchain, bytedance/podcast-generation · View all 6 →

즐겨 사용하는 AI에게 물어보기

이 에이전트 스킬이 미리 로드된 새 채팅을 엽니다.

문서

Manuscript Figures

Read the shared manuscript contract first. Reuse the task's confirmed Easy, Normal or Expert mode, scientific storyline, evidence and visual profile. This skill does not restart intake or introduce another review mode.

Hard rules for manuscript narrative

Apply the Editorial Guideline (mandatory default) in release narrative and defensive-writing rules before drafting, restructuring, polishing, auditing, or reviewing. They are hard defaults, with special priority for manuscript writing, storyline, idea framing, and innovation claims. When the user supplies manuscript text, perform the required point-by-point defensive-writing audit and give a concrete rewrite for every instance. Preserve material limitations, negative findings, uncertainty, ethics, safety, and reproducibility boundaries as precise evidence or scope statements.

For Nature or Nature Portfolio targets, follow the canonical Nature integration map and load only the relevant nature-shared reference; this manuscript skill remains the owner.

Figure gate for a new or full manuscript

For a new or full manuscript, enter the figure workflow after H4 and before Results prose is treated as complete. Build a claim–panel–data/artwork storyboard that records the panel duty, retained source, comparison, encoding, takeaway, caption and main/SI allocation. Use the shared terminology table for panel labels, axis terms, legend entries, captions, table headers and in-text callouts. For empirical or quantitative evidence, produce representative H6 prototypes from retained evaluated outputs and inspect them at final manuscript size before extending a visual language across the figure set.

Iterate design from the evidence structure. Review at least one credible alternative whenever the visual form, panel composition or encoding is consequential. Quantitative plots must use retained data and reproducible local code or an editable data-linked source. Synthetic, random or illustrative data may appear only in an explicitly labelled layout mockup and never as submission evidence. Every finished display requires source/provenance, script or editable artwork, complete caption, in-text callout and the four-layer audit. If a study genuinely has no displayable evidence, record an evidence-based N/A rationale in the workflow state. Do not silently omit figures or generate a result-looking image to fill a template.

Load only the guidance needed for the requested display work:

  • For figure-set planning, new artwork or a substantive redesign, read scientific figure generation. Preserve the source evidence, evaluated originals, accepted typography and actual target venue. Also read visual form selection and main-figure composition. For every new or redesigned main figure, document the data structure, panel duties, chosen form, rejected alternative and anti-default check.
  • For Python/Matplotlib manuscript plotting, also read figures4papers patterns. Keep its attribution and license notice. This is an implementation reference, not an importable plotting package; existing R and vector workflows remain valid.
  • For comparable multi-panel layouts, revisions that change text or mark geometry, schematics, or microscopy/gel image panels, read the applicable sections of figure verification. Keep measured checks distinct from visual inspection and checks not assessed.
  • For captions, legends or scientific tables, read captions and tables. A bounded caption edit needs the affected display, source facts and narrative checks; it does not require a new storyboard or redesign of the figure set.
  • For generated or redesigned artwork, apply the four visualization audit layers to the affected displays. Load further audit guidance only when the requested scope requires it.
  • For placement or typesetting requests such as loose pages, stranded headings, oversized floats or sparse Supplementary Information, read LaTeX layout and float typesetting. Render and inspect the compiled pages before and after changes.

Make consequential design decisions reviewable with actual evidence and representative artwork under the shared mode. Preserve accepted designs during local revisions. Deliver the requested artwork or caption, relevant source and export material, and the checks actually performed. Keep unsupported source facts explicit rather than inventing data or uncertainty.

Keep quantitative plots in reproducible local Python/R or equivalent source code. image_gen may assist with non-data illustrations or schematic assets, but never with quantitative evidence. If a connected Figma or diagram tool is used, treat it as optional vector/layout assistance only; keep the data, plotting code and final scientific QA in the manuscript project.

Individual skills in this repo

This repo contains 4 individual skills — each has its own dedicated page.

관련 스킬