AI Agent Hub
Back to skills
Research Writing Assistant icon

Research Writing Assistant

Professional Updated 2026.08.30

Paste the following prompt into your AI chat to install this skill:

Please install @user_41e893c2/kay-20260812-1 according to https://skillhub.cn/install/skillhub.md.

About this skill

What Problem It Solves

The hard part of research writing is not producing longer text, but turning analysis outputs into traceable paper prose: claims must stay within the data, citations must not be invented, and figures, tables, and body text must use the same metric definitions. This skill is aimed at medical and bioinformatics paper writing, especially when you already have analysis artifacts such as results/*/figures/Figure*.pdf, downstream/*/*.csv, FIGURE_LEGENDS.txt, and plan/feasibility.md, but not yet a coherent Introduction, Methods, Results, and Discussion structure.

How It Works

The workflow is structured rather than single-pass polishing:

  • Brainstorming: first define the topic, section outline, and evidence chain before drafting prose.
  • Evidence and introduction: use evidence-driven writing to organize background and motivation, with no fabricated citations.
  • Results planning: connect experiment-results-planning, FIGURE_LEGENDS.txt, and plan/figures-manifest.md so each figure maps to a specific results paragraph.
  • Chapter writing: use chapter-level and medical-writing rules to shape the manuscript voice.
  • Figure legends and results paragraphs: write by figure number, keeping numeric values, thresholds, sample sizes, and visual contents consistent.
  • Self-review: run a pre-submission check focused on boundary conditions, citation traceability, and factual consistency.

The default outputs are typically .md or .tex, making it useful as a manuscript draft or LaTeX writing intermediate layer, not as a replacement for the full analysis pipeline.

Scope and Caveats

  • Good for: structuring paper sections, organizing figure legends, aligning table citations, pre-submission review, and drafting in Markdown or LaTeX.
  • Not good for: running analyses, recalculating statistics, or generating new biological conclusions; those should remain in upstream analysis skills.
  • The writing must preserve method conditions, sample sizes, thresholds, and conclusion boundaries; removing them in the name of de-AI-ifying the text would weaken the paper.
  • Every citation and data reference should be traceable to an actual source, rather than added merely to make the manuscript appear complete.

Use Cases

  • {'title': 'From analysis outputs to Results prose', 'desc': 'Turn `results/*/figures/Figure*.pdf` and `FIGURE_LEGENDS.txt` into figure-by-figure Results paragraphs.'}
  • {'title': 'Structure introduction and evidence chain', 'desc': 'Organize the Introduction before submission, linking motivation, background citations, and project findings.'}
  • {'title': 'Align Methods with cohort notes', 'desc': 'Use `plan/feasibility.md` to draft cohort definitions, method conditions, and sample boundaries consistent with the data.'}
  • {'title': 'Pre-submission review and consistency check', 'desc': 'Check figure values, thresholds, sample sizes, conclusion limits, and citation traceability before submitting.'}

Best For

  • {'title': 'Bioinformatics researchers', 'desc': 'Need to turn statistical, downstream, and figure results into medical paper prose without rerunning analysis.'}
  • {'title': 'Medical or life-science authors', 'desc': 'Preparing Introduction, Methods, Results, and Discussion while preserving sample sizes, thresholds, and conclusion limits.'}
  • {'title': 'LaTeX or Markdown writers', 'desc': 'Want `.md` or `.tex` drafts that can be reviewed, edited, and submitted directly.'}
  • {'title': 'Pre-submission reviewers', 'desc': 'Responsible for checking figure-text consistency, table citations, legends, and citation traceability.'}