ahsanmandhar-ui/research-poster-maker

Research Poster Maker Skill

Produce conference-ready scientific research posters that read as if art-directed by an experienced academic graphic designer (suitable for Nature conferences, BMJ meetings, ACC, ASN, European congresses, and university symposia)—never generic AI infographics.

All file paths below are relative to this skill's root directory (research-poster-maker/).

---

Core Non-Negotiable Rules

1. Zero Scientific Fabrication: Never invent sample sizes, effect estimates, confidence intervals, p-values, exclusion counts, citations, grant numbers, or affiliations. Preserve exact decimal precision, units, and statistical significance framing.
2. Distill Prose—Do Not Paste Walls of Text: Transform manuscript paragraphs into concise, high-clarity analytical statements and parallel bullets (~350–600 body words total) while preserving exact meaning.
3. Never Shrink All Fonts to Cure Overcrowding: If content overflows a layout at legible viewing-distance sizes, rewrite verbose prose, prune secondary background points, rebalance columns, or switch layouts.
4. Restrained Editorial Design: Avoid identical rounded cards around every section, decorative clip-art/emojis, drop shadows, rainbow gradients, and knee-jerk medical blue. Use clean typographic rules, intentional whitespace, high WCAG contrast, and real data visualizations.
5. Honest Print & Technical Reporting: Never upscale a low-resolution raster image and claim it is 300 DPI. Never claim CMYK offset-press PDF/X compliance on sRGB outputs. Report exact dimensions, .pptx scale factor (if >56 in), and PDF export backend transparently.

---

Step-by-Step Workflow

Step 1 — Inspect Available Research Material First

Before asking the user questions, locate and inspect all supplied manuscripts (.docx, .pdf, .md, .txt), datasets (.csv, .xlsx), and existing figures/logos:

python scripts/analyze_research.py <input_files_or_folder> --output-dir <work_dir>

Review <work_dir>/research_extraction.json and <work_dir>/extraction_summary.md to check the 18 canonical fields (Title, Authors/Affiliations, Background, Research Gap, Objectives, Study Design, Methods, Population, Sample Size, Statistical Methods, Primary Outcome, Secondary Outcomes, Important Findings, Tables/Figures, Limitations, Conclusion, References, Funding/Acknowledgements).

Step 2 — Request Only Essential Missing Information

Do not present a long questionnaire when the manuscript already supplies the study details. Ask concisely only when:

• Critical scientific facts are missing from the source (e.g., missing authors/affiliations or missing primary outcome numbers) and cannot be inferred.
• Conference board dimensions/orientation were not specified (offer a sensible regional default—e.g., 48 × 36 in landscape or ISO A0 portrait—clearly labeled as an assumption).
• Institutional branding/logo or study QR URL is desired.

Step 3 — Select Architecture & Consult Reference Docs as Needed

Load detailed reference guides from docs/ on demand:

• Dimensions, 56-in PowerPoint ceiling & print specs: docs/print-specifications.md
• Color palettes & editorial anti-patterns: docs/design-system.md
• Typography pairings & viewing-distance point scales: docs/typography.md
• 8 Layout composition strategies: docs/layout-recipes.md
• 15 Scientific chart/diagram types: docs/visualization-guide.md
• Scientific integrity & extraction rules: docs/scientific-integrity.md
• WCAG 2.1 contrast & Okabe-Ito colorblind safety: docs/accessibility.md

Available templates in templates/<id>/template.json:

• clinical-research | systematic-review | meta-analysis | experimental-research | epidemiology | minimalist-editorial

Available layout recipes:

• three_column_classic | two_column_editorial | asymmetric_grid | results_focused | timeline_flow | central_visual | clinical_workflow | systematic_review_prisma

Available palettes:

• academic_navy | charcoal_ivory | deep_teal | burgundy_offwhite | forest_cream | monochrome_editorial (plus custom hex overrides)

Step 4 — Present Brief Pre-Generation Design Specification

Before final rendering, present a concise specification block:

Poster: Clinical Research
Orientation: Landscape
Dimensions: 48 × 36 inches (explicit / assumed default)
Layout: clinical_workflow (30% : 41% : 29%)
Typography: IBM Plex Sans
Palette: Oxford Navy & Crisp White (#0F2742 / #FFFFFF / #D55E00)
Visual focus: Kaplan-Meier primary endpoint curve + subgroup outcomes table
Output: PPTX + PDF + PNG + Content & Design Reports

Step 5 — Author poster_spec.json & Execute Production Pipeline

Construct <work_dir>/poster_spec.json strictly from verified source data:

{
  "template_id": "clinical-research",
  "dimensions_preset": "48x36",
  "layout_recipe": "clinical_workflow",
  "palette": "academic_navy",
  "font_pairing": "clinical_precision",
  "study_badge": "PROSPECTIVE MULTICENTER COHORT · STROBE",
  "title": "Exact Verified Study Title in Sentence Case",
  "authors_line": "First Author, MD¹; Second Author, PhD²",
  "affiliations_line": "¹Dept. of Medicine, Institution A; ²Dept. of Biostatistics, Institution B",
  "study_url": "https://doi.org/...",
  "logo_path": null,
  "items": [
    {
      "id": "background",
      "column": 0,
      "item_type": "text",
      "style": "editorial_rule",
      "title": "01  Clinical Background & Objective",
      "paragraphs": ["Concise context distilled from source."],
      "bullets": ["Objective: Precise primary hypothesis."]
    },
    {
      "id": "primary_outcome",
      "column": 1,
      "item_type": "callout",
      "style": "focal_callout",
      "title": "Primary Clinical Endpoint",
      "hero_metric": "aHR 0.68 (95% CI 0.56–0.82; p < 0.001)",
      "paragraphs": ["Exact analytical summary of primary finding."]
    },
    {
      "id": "fig1",
      "column": 1,
      "item_type": "figure",
      "title": "03  Cumulative Event-Free Survival",
      "chart_spec": {
        "id": "fig1_km",
        "chart_type": "kaplan_meier",
        "width_in": 16.5,
        "height_in": 9.2,
        "x_label": "Follow-up (Months)",
        "y_label": "Event-Free Survival",
        "data": { "groups": [], "risk_times": [], "stat_summary": "..." }
      },
      "caption": "Figure 1. Kaplan-Meier survival curves with number at risk (N = ...)."
    },
    {
      "id": "tbl1",
      "column": 2,
      "item_type": "table",
      "title": "04  Multivariable Outcomes",
      "table_data": {
        "headers": ["Endpoint", "Adjusted HR (95% CI)", "p-value"],
        "rows": [["Primary composite", "0.68 (0.56–0.82)", "<0.001"]],
        "col_weights": [1.9, 1.3, 0.7]
      },
      "footnote": "Adjusted for pre-specified baseline covariates."
    }
  ],
  "colophon": {
    "references": ["[1] Author et al. Journal Year;Vol:Pages."],
    "funding": "Funding: Grant ID.",
    "correspondence": "Correspondence: [email protected]"
  },
  "missing_information_notes": []
}

Run the unified generator (which runs chart generation, layout solving, PPTX build, PDF export, PNG preview rendering, and multi-layer QA validation):

python scripts/generate_poster.py \
  --spec <work_dir>/poster_spec.json \
  --extraction <work_dir>/research_extraction.json \
  --output-dir <work_dir>/output

Individual pipeline stages can also be invoked independently via scripts/generate_charts.py, scripts/export_poster.py, scripts/generate_preview.py, and scripts/validate_poster.py.

Step 6 — Visual Inspection & Bounded Iterative Refinement

1. Read <work_dir>/output/validation_report.md to confirm Technical, Scientific, Accessibility, and Visual checks passed.
2. Visually inspect <work_dir>/output/poster_preview.png using image-viewing tools whenever available. Check overall visual hierarchy, column balance, chart legibility, alignment, and whitespace.
3. If any issue is found (Generate → Export → Render → Inspect → Validate → Revise → Finalize), adjust poster_spec.json (condense verbose bullets, adjust column assignments or aspect_ratio, or switch layout_recipe) and re-run. Stop once objective checks pass or honestly report any unresolved constraint (such as low-resolution user-supplied photography).

---

Mandatory Deliverables Checklist

Verify all five files exist in <output_dir> before concluding:

1. poster.pptx — Editable native PowerPoint poster (shapes, text runs, booktabs tables, high-DPI/native charts).
2. poster.pdf — Print-ready PDF preserving exact physical dimensions and embedded fonts.
3. poster_preview.png — Full-poster rendered preview image.
4. poster_content.md — Extracted and distilled poster content with completeness notes.
5. design_specification.md — Complete physical, typographic, color-contrast, and QA audit report (accompanied by validation_report.md).