Scientific Research

Overview

Master skill for all scientific research workflows. Covers the full research lifecycle: literature reviews, hypothesis generation, statistical analysis, manuscript writing, figure creation, scientific diagrams, presentation slides, citation management, peer review, and grant proposals.

Each domain has a dedicated reference file with detailed guidance. This skill routes you to the right reference based on your intent.

Task Router

Intent	Reference File	When to Use
Writing papers, manuscripts, IMRAD structure	references/writing.md	Drafting or revising any section; also cover letters and rebuttals
Creating publication figures, plots	references/visualization.md	Journal-ready plots; also multi-panel layouts and supplementary figures
Scientific diagrams, schematics, flowcharts	references/schematics.md	AI-generated diagrams via Gemini; also graphical abstracts
Presentation slides, conference talks	references/slides.md	PowerPoint/Beamer decks; also poster talks and thesis defenses
Literature reviews, systematic reviews	references/literature-review.md	Multi-database search + PRISMA; also scoping and narrative reviews
Hypothesis generation, experimental design	references/hypothesis.md	Competing hypotheses with predictions; also power/sample-size planning
Citation management, BibTeX, DOI lookup	references/citations.md	Search, validate, deduplicate; also format conversion between styles
Peer review, manuscript evaluation	references/peer-review.md	Structured review with major/minor comments; also rebuttal responses
Grant proposals (NSF, NIH, DOE, DARPA, NSTC)	references/grants.md	Full proposals; also LOIs, budget justifications, and resubmissions
Statistical tests, power analysis, reporting	references/statistics.md	Test selection + APA reporting; also Bayesian analysis and meta-analysis
Qualitative methods, thematic analysis, interviews	references/qualitative-research.md	Reflexive TA 6-phase process; also grounded theory, phenomenology, mixed methods
Open science, preregistration, reproducibility	references/open-science.md	OSF workflow, registered reports, FAIR data, open code, reproducibility checklists
Research ethics, IRB, decolonial methodology	references/research-ethics.md	Belmont principles, CBPR, OCAP, CRediT authorship, data privacy (GDPR)
Data management, FAIR principles, DMPs	references/data-management.md	NSF/NIH DMP requirements, metadata standards, versioning (DVC), repository selection
Causal inference, DAGs, multilevel models, SEM	references/advanced-statistics.md	DAGitty, propensity scores, IV, DiD, lme4, lavaan; also causal ML
Science communication, policy briefs, public engagement	references/research-communication.md	Policy briefs, press releases, social media threads, video abstracts, graphical abstracts

Before Starting

Gather these answers before any research task — they determine which references to load and how to calibrate output:

1. Discipline? (e.g., biomedical, social science, engineering, ecology) — drives terminology, citation style, reporting guidelines
2. Research stage? (literature review, hypothesis, data collection, analysis, writing, revision) — determines workflow entry point
3. Output format? (journal manuscript, grant proposal, conference slides, poster, thesis chapter) — sets quality thresholds and structure
4. Target venue? (Nature, PNAS, field-specific journal, NSF, NIH) — dictates formatting, word limits, figure specs
5. Data type? (continuous, categorical, mixed, qualitative) — guides statistical test selection

Quick Reference

Statistical Test Selector (most common cases):

• 2 groups, normal → independent t-test | non-normal → Mann-Whitney U
• 3+ groups, normal → one-way ANOVA | non-normal → Kruskal-Wallis
• Outcome prediction → linear regression (continuous) | logistic regression (binary)
• Association → Pearson r (normal) | Spearman rho (non-normal)

Citation Format by Field:

• Social sciences → APA (Author, Year) | Medicine → AMA superscript | Biomedical → Vancouver [1] | Engineering/CS → IEEE [1]

Reporting Guideline Selector:

• RCT → CONSORT | Observational → STROBE | Systematic review → PRISMA | Animal → ARRIVE | Prediction model → TRIPOD

Cross-Cutting Principles

Visual Enhancement (All Documents)

Every scientific document should include AI-generated figures. Use the schematics workflow (see references/schematics.md) to generate diagrams via Gemini MCP tools:

• Papers: Graphical abstract + 5-8 figures minimum
• Literature reviews: PRISMA flow diagram + thematic synthesis diagrams
• Grant proposals: Methodology flowchart + timeline/Gantt chart
• Presentations: 1-2 visuals per slide
• Hypothesis reports: Framework diagram + experimental design flowchart

Writing Style

• Always write in full paragraphs for manuscripts -- never submit bullet points in final text
• Use two-stage process: (1) outline with key points, (2) convert to flowing prose
• Match field-specific terminology to discipline and target journal
• Follow IMRAD structure unless venue requires otherwise

Citation Rigor

• Verify all DOIs before submission using validation scripts
• Prioritize high-impact papers (Tier 1 venues, high citation counts)
• Use consistent citation style throughout (APA, Vancouver, Nature, etc.)
• Run metadata enrichment on incomplete BibTeX entries

Statistical Standards

• Always check assumptions before running tests
• Report effect sizes with confidence intervals alongside p-values
• Distinguish statistical from practical significance
• Pre-register analyses when possible

Reproducibility

• Document all methods with enough detail for replication
• Share code and data via repositories
• Follow reporting guidelines (CONSORT, STROBE, PRISMA, ARRIVE)
• Version-control all source files and generated figures

Domain Summaries

Scientific Writing

Write manuscripts using IMRAD structure with proper citations, figures, and reporting guidelines. Two-stage process: outline first, then convert to prose. Example: for a Nature submission, use Vancouver citations, 3000-word limit, structured abstract, and 6-8 display items. See references/writing.md.

Scientific Visualization

Create publication-ready figures with multi-panel layouts, error bars, significance markers, and colorblind-safe palettes. Export as PDF/EPS/TIFF. Example: Nature requires 89mm (single column) or 183mm (double) width at 300+ DPI. See references/visualization.md.

Scientific Schematics

Generate publication-quality diagrams using Gemini MCP tools with iterative quality refinement. Supports neural network architectures, biological pathways, CONSORT/PRISMA flowcharts, system diagrams, and circuit schematics. Quality thresholds vary by document type (6.5-8.5/10). See references/schematics.md.

Scientific Slides

Build slide decks for conference talks, seminars, and defenses. Supports PowerPoint and LaTeX Beamer. Emphasizes visual engagement, minimal text, proper citations, and story-driven narrative. Slide count targets ~1 per minute. See references/slides.md.

Literature Review

Conduct systematic, multi-database literature reviews (PubMed, arXiv, bioRxiv, Semantic Scholar). Use PICO framework for clinical reviews (e.g., P: Type 2 diabetes patients, I: GLP-1 agonists, C: metformin, O: HbA1c reduction). Phases: planning, search, screening, extraction, synthesis. See references/literature-review.md.

Hypothesis Generation

Systematic hypothesis formulation from observations. Generate 3-5 competing hypotheses, evaluate against quality criteria (testability, falsifiability, parsimony), design experimental tests, formulate quantitative predictions. LaTeX report output with colored boxes. See references/hypothesis.md.

Citation Management

Search Google Scholar and PubMed, extract metadata from DOI/PMID/arXiv, validate BibTeX entries, format and deduplicate bibliographies. Scripts for automated search, extraction, validation, and formatting. See references/citations.md.

Peer Review

Structured manuscript evaluation: initial assessment, section-by-section review, methodological rigor, reproducibility, figure integrity, ethical considerations, writing quality. Hierarchical feedback with major/minor comments. Presentation review via image-based inspection. See references/peer-review.md.

Research Grants

Write competitive proposals for NSF, NIH, DOE, DARPA, and Taiwan NSTC. Agency-specific formatting, review criteria, budget preparation, broader impacts, specific aims, timeline planning, and resubmission strategies. See references/grants.md.

Statistical Analysis

Guided test selection, assumption checking, power analysis, effect sizes, and APA reporting. Example output: "t(98)=3.82, p<.001, d=0.77, 95% CI [0.36, 1.18]". Always check assumptions first, report effect sizes with CIs, and use non-parametric alternatives when normality is violated. See references/statistics.md.

Qualitative Research

Conduct rigorous qualitative studies using Reflexive Thematic Analysis (Braun & Clarke 6-phase), Grounded Theory (theoretical sampling, constant comparison, saturation), and interview design. Quality criteria: credibility, transferability, dependability, confirmability. Tools: NVivo, ATLAS.ti, Taguette. Mixed methods integration. See references/qualitative-research.md.

Open Science

Implement preregistration via OSF or AsPredicted, write Registered Reports, share data under FAIR principles, and deposit code with Zenodo DOIs. Discipline-specific reproducibility checklists (psychology, biomedical, computational). Open practice badges. See references/open-science.md.

Research Ethics

Navigate IRB review levels, design valid informed consent, apply CBPR and OCAP principles for community and Indigenous research, and embed decolonial reflexivity. CRediT authorship taxonomy, conflict of interest disclosure, GDPR data privacy for research. See references/research-ethics.md.

Data Management

Write NSF and NIH Data Management Plans, implement FAIR principles with appropriate metadata standards (Dublin Core, DataCite, EML, BIDS), version data with DVC or Git LFS, and select the right repository (Zenodo, Dryad, ICPSR, discipline-specific). License selection: CC0 vs CC BY. See references/data-management.md.

Advanced Statistics

Causal inference with DAGs (DAGitty, backdoor criterion, collider bias), propensity scores, IV, RDD, and DiD. Multilevel models (lme4/statsmodels, ICC, random slopes). Longitudinal analysis and multiple imputation. SEM/CFA with lavaan/semopy (CFI, RMSEA, SRMR). Prediction vs causal model distinction; Double ML and causal forests. See references/advanced-statistics.md.

Research Communication

Translate findings for non-academic audiences: 1-2 page policy briefs with numbered recommendations, press releases with inverted-pyramid structure, Twitter/X 10-tweet threads, video abstracts (90-second structure), and graphical abstracts. Altmetric tracking. Audience analysis, curse of knowledge, analogy and metaphor. See references/research-communication.md.

Integration Notes

These domains work together in research workflows:

• Literature review feeds hypothesis generation and writing (introduction/discussion)
• Statistical analysis produces results reported in writing and visualized via visualization
• Schematics generates diagrams for writing, slides, grants, and literature reviews
• Citation management supports writing, literature review, and grants
• Peer review evaluates manuscripts produced by writing and visualization
• Slides presents findings from all other domains

Layer 3 Skills

These atomic skills provide deeper specialization within the scientific research domain:

Skill	Use when
writing	IMRAD manuscripts, cover letters, rebuttals
visualization	Publication-quality figures, multi-panel layouts
schematics	AI-generated diagrams via Gemini, graphical abstracts
slides	Conference talks, PowerPoint/Beamer decks
literature-review	Systematic reviews, PRISMA, multi-database search
hypothesis	Competing hypotheses, experimental design
citations	BibTeX, DOI lookup, bibliography management
peer-review	Manuscript evaluation, rebuttal responses
grants	NSF, NIH, DOE, DARPA proposals
statistics	Test selection, power analysis, APA reporting
qualitative-research	Thematic analysis, grounded theory, interviews
open-science	Preregistration, OSF, FAIR data, reproducibility
research-ethics	IRB, informed consent, CBPR, CRediT
data-management	FAIR DMPs, metadata standards, versioning
advanced-statistics	Causal inference, DAGs, multilevel models, SEM
research-communication	Policy briefs, press releases, video abstracts