Agent-Skills.md

Agent Skills: Scientific Writing

Core skill for the deep research and writing tool. Write scientific manuscripts in full paragraphs (never bullet points). Use two-stage process: (1) create section outlines with key points using research-lookup, (2) convert to flowing prose. IMRAD structure, citations (APA/AMA/Vancouver), figures/tables, reporting guidelines (CONSORT/STROBE/PRISMA), for research papers and journal submissions.

UncategorizedID: drshailesh88/integrated_content_OS/scientific-writing
2

Install this agent skill to your local

pnpm dlx add-skill https://github.com/drshailesh88/integrated_content_OS/tree/HEAD/skills/cardiology/scientific-writing

Skill Files

Browse the full folder contents for scientific-writing.

Download Skill

Loading file tree…

skills/cardiology/scientific-writing/SKILL.md

Scientific Writing

Core skill for producing research manuscripts, evidence-based articles, and publication-quality scientific content with rigorous standards.

Triggers

  • User asks to write a research-style article
  • User needs IMRAD-structured content
  • User wants to write for a scientific audience
  • User is drafting content requiring citations and methodology
  • User needs to follow reporting guidelines (CONSORT, STROBE, PRISMA)

Two-Stage Writing Process

Stage 1: Outline with Research

  1. Create section outlines with bullet points
  2. Use pubmed-database skill for literature gathering
  3. Identify key citations for each point
  4. Structure evidence hierarchy

Stage 2: Convert to Prose

Critical: Always write in full paragraphs with flowing prose. Never submit bullet points in the final manuscript.

IMRAD Structure

| Section | Purpose | Key Elements | |---------|---------|--------------| | Introduction | Why this matters | Background, gap, objective | | Methods | What you did | Design, population, analysis | | Results | What you found | Data, statistics, figures | | Discussion | What it means | Interpretation, limitations, implications |

Reporting Guidelines

For Different Study Types

| Study Type | Guideline | Checklist Items | |------------|-----------|-----------------| | RCTs | CONSORT | Randomization, blinding, flow diagram | | Observational | STROBE | Selection, bias assessment, confounders | | Systematic reviews | PRISMA | Search strategy, screening, synthesis | | Diagnostic studies | STARD | Index test, reference standard, flow | | Case reports | CARE | Timeline, diagnostic reasoning, outcomes |

Citation Standards

Supported Styles

  • AMA - Medical journals (JAMA, NEJM)
  • Vancouver - Numbered citations
  • APA - Psychology, behavioral science
  • Nature - Superscript numbered
  • IEEE - Engineering/technical

Citation Best Practices

  1. Cite primary sources, not reviews (unless reviewing)
  2. Include DOI when available
  3. Verify accuracy with citation-management skill
  4. Prefer recent publications (<5 years) when current evidence exists
  5. Include seminal papers for historical context

Field-Specific Conventions

Cardiology/Medicine

  • Report confidence intervals, not just p-values
  • Include NNT/NNH for clinical relevance
  • Use GRADE for evidence quality
  • Specify intention-to-treat vs per-protocol
  • Report absolute and relative risk

Statistical Reporting

  • Mean (SD) for normal distributions
  • Median (IQR) for skewed data
  • HR with 95% CI for survival analysis
  • OR/RR with 95% CI for binary outcomes
  • Specify statistical tests used

Common Rejection Reasons (Avoid These)

  1. Inadequate statistical descriptions - Always specify tests, assumptions, software
  2. Over-interpretation of results - Stay within what data supports
  3. Poorly described methods - Replicability is essential
  4. Missing reporting checklist items - Use appropriate guidelines
  5. Weak literature contextualization - Show how work fits existing knowledge

Quality Checklist

Before finalizing scientific content:

  • [ ] All claims have citations
  • [ ] Statistics include effect sizes and CIs
  • [ ] Methods are replicable
  • [ ] Limitations acknowledged
  • [ ] Conclusions match evidence strength
  • [ ] No AI tells or promotional language