Agent Skills: Visual Design System

Visual Design System

Purpose: Publication-grade design tokens and utilities for Nature/JACC/NEJM quality graphics.

Status: Phase 3.1 In Progress - Manim Animations

Which Tool Should I Use?

Use this decision table to pick the right backend before writing any code:

| Task | Best Tool | Output | |------|-----------|--------| | Publication figure (bar, line, forest plot) | Plotly or drawsvg | PNG (300 DPI) | | Infographic card / social slide | Satori or Component Library | PNG, SVG | | Medical diagram (heart, ECG, flowchart) | drawsvg | PNG, SVG | | Clinical trial / drug mechanism template | SVG Templates | PNG, SVG | | Treatment pathway / CONSORT / PRISMA | Architecture Diagrams | PNG | | Animated mechanism / Kaplan-Meier / ECG | Manim | MP4 | | Any of the above with a unified Python API | Component Library | PNG, SVG, HTML |

Quick rule: if it's a static publication figure → Plotly/drawsvg; if it's a polished card or social asset → Satori/Component; if it needs motion → Manim.

Publication Figure Workflow (Create → Validate → Export)

For any publication-grade output, follow this validated sequence:

from skills.cardiology.visual_design_system.tokens import (
    get_color, get_accessible_pair, validate_contrast, get_contrast_ratio
)
from cardiology_visual_system.scripts.plotly_charts import create_comparison_bars, save_chart

# 1. Create the figure using token colors
treatment, control = get_accessible_pair("treatment_control")
fig = create_comparison_bars(
    categories=["Primary", "Secondary"],
    group1_values=[12.3, 8.5],
    group2_values=[18.7, 14.2],
    group1_name="Treatment",
    group2_name="Placebo",
    title="Clinical Trial Results"
)

# 2. Validate contrast before export
if not validate_contrast(treatment, "#ffffff"):
    ratio = get_contrast_ratio(treatment, "#ffffff")
    # Fix: swap to a pre-validated pair
    treatment, control = get_accessible_pair("benefit_risk")

# 3. Validate DPI / scale — always use scale=4 for 300 DPI
# fig.write_image("results.png", scale=4)  # direct Plotly
save_chart(fig, "results.png")            # auto 300 DPI (scale=4)

If validation fails:

| Problem | Cause | Fix | |---------|-------|-----| | validate_contrast returns False | Ratio < 4.5:1 | Swap to a pre-validated pair via get_accessible_pair() | | Render produces no output | Missing dependency | Run pip install drawsvg cairosvg or pip install diagrams | | Manim scene not found | Not in catalog | Run python scripts/render_manim.py --list and check scene_catalog.json | | 300 DPI export looks blurry | Wrong scale | Use scale=4 in fig.write_image() or save_chart(fig, path) | | Font not applied | System font missing | Tokens fall back to Arial; check tokens.get_font_family("primary") |

Quick Start

from skills.cardiology.visual_design_system.tokens import (
    get_tokens,
    get_color,
    get_accessible_pair,
    validate_contrast,
)

# Get a specific color
navy = get_color("primary.navy")  # "#1e3a5f"

# Get a colorblind-safe pair for treatment vs control
treatment, control = get_accessible_pair("treatment_control")

# Validate accessibility
is_safe = validate_contrast("#1e3a5f", "#ffffff")  # True (9.2:1 ratio)

# Get full tokens object
tokens = get_tokens()
palette = tokens.get_color_palette("categorical")  # 7 colorblind-safe colors

Design Philosophy

This system enforces Nature journal standards for all visual output:

| Standard | Requirement | How We Enforce | |----------|-------------|----------------| | Fonts | Helvetica/Arial only | Token system + validation | | Font sizes | 5-8pt for figures | Pre-defined size scale | | Contrast | WCAG AA (4.5:1 min) | Automated validation | | Colorblind | No red-green only | Paul Tol palettes | | Resolution | 300 DPI minimum | Export presets | | Shadows | None in figures | Disabled by default |

Token Categories (Overview)

Full token reference: references/color_palettes.md and references/nature_guidelines.md.

Colors

# Primary
get_color("primary.navy")    # "#1e3a5f"
get_color("primary.blue")    # "#2d6a9f"
get_color("primary.teal")    # "#48a9a6"

# Semantic
get_color("semantic.success")  # "#2e7d32"
get_color("semantic.warning")  # "#e65100"
get_color("semantic.danger")   # "#c62828"
get_color("semantic.neutral")  # "#546e7a"

# Colorblind-safe palettes
tokens.get_color_palette("categorical")      # 7 Paul Tol colors
tokens.get_color_palette("sequential_blue")  # 5-step blue ramp
tokens.get_color_palette("diverging")        # blue ← neutral → red

# Pre-validated accessible pairs
t, c    = get_accessible_pair("treatment_control")    # ('#0077bb', '#ee7733')
b, r    = get_accessible_pair("benefit_risk")         # ('#009988', '#cc3311')
i, p    = get_accessible_pair("intervention_placebo") # ('#0077bb', '#bbbbbb')

# Clinical outcome colors
tokens.get_clinical_color("mortality")           # "#b2182b"
tokens.get_clinical_color("hospitalization")     # "#ef8a62"
tokens.get_clinical_color("symptom_improvement") # "#67a9cf"

# Forest plot colors
colors = tokens.get_forest_plot_colors()
# keys: point_estimate, confidence_interval, null_line, summary_diamond,
#       heterogeneity_low, heterogeneity_moderate, heterogeneity_high

Typography

tokens.get_font_family("primary")   # "Helvetica, Arial, sans-serif"
tokens.get_font_family("monospace") # "Courier New, Courier, monospace"

# Figure elements (Nature 5-8pt standard)
tokens.get_font_size("figure_elements", "panel_label")  # 8pt
tokens.get_font_size("figure_elements", "axis_title")   # 7pt
tokens.get_font_size("figure_elements", "axis_tick")    # 6pt

# Infographic / social media
tokens.get_font_size("infographic_elements", "headline")   # 24pt
tokens.get_font_size("social_media", "carousel_stat")      # 48pt

Spacing & Strokes

# 4px base grid
tokens.get_spacing("xs")  # "4px"  |  tokens.get_spacing("sm")  # "8px"
tokens.get_spacing("md")  # "12px" |  tokens.get_spacing("lg")  # "16px"

tokens.get_stroke_width("hairline")  # "0.5px"
tokens.get_stroke_width("thin")      # "1px"
tokens.get_stroke_width("regular")   # "1.5px"

Validation

CLI

python scripts/token_validator.py                  # Full validation
python scripts/token_validator.py --contrast-report
python scripts/token_validator.py --json

Programmatic

from tokens.index import validate_contrast, get_contrast_ratio

validate_contrast("#1e3a5f", "#ffffff")              # True  (WCAG AA)
validate_contrast("#1e3a5f", "#ffffff", level="AAA") # True  (WCAG AAA)
get_contrast_ratio("#1e3a5f", "#ffffff")             # 9.2

Plotly (Phase 1.4)

Standard publication charts with automatic token integration and 300 DPI export.

from cardiology_visual_system.scripts.plotly_charts import (
    create_comparison_bars, save_chart,
)

fig = create_comparison_bars(
    categories=["Primary", "Secondary"],
    group1_values=[12.3, 8.5],
    group2_values=[18.7, 14.2],
    group1_name="Treatment",
    group2_name="Placebo",
    title="Clinical Trial Results"
)
save_chart(fig, "results.png")  # Auto 300 DPI (scale=4)

# CLI
cd skills/cardiology/cardiology-visual-system/scripts
python plotly_charts.py demo --quality-report
python plotly_charts.py demo --png --output-dir ../outputs
python plotly_charts.py bar -d data.csv -o chart.png

Direct template usage:

from tokens.index import get_plotly_template
import plotly.io as pio

pio.templates["publication"] = get_plotly_template()
fig = px.bar(data, template="publication")
fig.write_image("chart.png", scale=4)

Satori Infographic Pipeline (Phase 1.2)

Generates PNG/SVG infographic cards from structured data via a Node.js renderer.

Templates: stat-card, comparison, process-flow, trial-summary, key-finding

cd satori/
node renderer.js --list
node renderer.js --template stat-card \
  --data '{"value": "26%", "label": "Mortality Reduction", "source": "PARADIGM-HF"}' \
  -o ../outputs/stat-card.png

from scripts.generate_infographic import generate_stat_card, generate_trial_summary

generate_stat_card(
    "26%", "Mortality Reduction",
    sublabel="HR 0.74, 95% CI 0.65-0.85",
    source="PARADIGM-HF",
    output="outputs/stat-card.png"
)

generate_trial_summary(
    "DAPA-HF", "HFrEF patients", "Dapagliflozin 10mg",
    "CV death or HF hospitalization",
    0.74, "0.65-0.85", "<0.001",
    nnt=21,
    output="outputs/trial.png"
)

Output: 1200×630px PNG at 2x scale. Custom dimensions: pass width= / height= to any generator or --width/--height to the CLI.

drawsvg Pipeline (Phase 1.3)

Pure Python SVG for medical diagrams and charts. No Node.js required.

pip install drawsvg cairosvg

Modules: medical_diagrams (heart, ECG, conduction, organ icons), data_charts (bar, grouped bar, line, forest plot), process_flows (algorithm, patient journey, study flow, simple flow).

from drawsvg.medical_diagrams import ecg_wave
from drawsvg.data_charts import forest_plot
from drawsvg.process_flows import study_flow

# ECG waveform
svg = ecg_wave(wave_type="normal", show_labels=True, title="Normal Sinus Rhythm")
svg.save_png("ecg.png")

# Forest plot
studies = [
    {"name": "DAPA-HF",         "estimate": 0.74, "lower": 0.65, "upper": 0.85, "weight": 60},
    {"name": "EMPEROR-Reduced", "estimate": 0.75, "lower": 0.65, "upper": 0.86, "weight": 50},
    {"name": "DELIVER",         "estimate": 0.82, "lower": 0.73, "upper": 0.92, "weight": 70},
]
svg = forest_plot(studies=studies, title="SGLT2 Inhibitors in HF", show_pooled=True)
svg.save_png("forest_plot.png")
# → For component-based forest plot with backend selection, see Component Library below.

# CONSORT study flow
svg = study_flow(enrollment=1500, randomized=1200,
    groups=[
        {"name": "Treatment", "allocated": 600, "discontinued": 45, "analyzed": 555},
        {"name": "Control",   "allocated": 600, "discontinued": 52, "analyzed": 548},
    ],
    title="DAPA-HF Study Flow"
)
svg.save_png("study_flow.png")
# → For CONSORT diagrams with auto-routing, see Architecture Diagrams below.

For full parameter reference (all wave_type values, highlight_chamber options, etc.) see svg_diagrams/.

Component Library (Phase 2.1)

Unified Python API wrapping Satori, Plotly, and drawsvg with consistent interfaces.

Components: StatCard, ComparisonChart, ForestPlot, Timeline, ProcessFlow, DataTable

from components import StatCard, ForestPlot, ComparisonChart, DataTable

card = StatCard(value="26%", label="Mortality Reduction",
                sublabel="HR 0.74, 95% CI 0.65-0.85", source="PARADIGM-HF")
card.render("stat_card.png")                    # auto backend
card.render("stat_card.png", backend="satori")  # infographic style
card.render("stat_card.png", backend="drawsvg") # publication style

# Forest plot with backend selection (see drawsvg section for raw SVG alternative)
plot = ForestPlot(studies=[...], title="SGLT2i in HF", x_label="Hazard Ratio (95% CI)")
plot.render("forest.png", backend="plotly")

table = DataTable(
    title="Baseline Characteristics",
    headers=["Characteristic", "Treatment (n=500)", "Control (n=500)", "P-value"],
    rows=[["Age, years", "65.2 ± 12.1", "64.8 ± 11.9", "0.62"]],
    footer="Values are mean ± SD or n (%)"
)
table.render("baseline.png")

Backend selection:

| Backend | Best For | |---------|----------| | satori | Infographic cards, social media | | plotly | Interactive charts, data viz | | drawsvg | Publication figures, diagrams |

Resolution config:

from components.base import RenderConfig
config = RenderConfig(width=1200, height=630, quality="print")  # 300 DPI
card = StatCard(value="42%", label="Test", config=config)

SVG Infographic Templates (Phase 2.2)

lxml-based SVG placeholder replacement for five standard medical layouts.

Templates: trial_results, drug_mechanism, patient_stats, before_after, risk_factors

Full field reference: see TEMPLATE_REFERENCE.md in svglue_templates/.

cd svglue_templates/
python template_renderer.py --list
python template_renderer.py trial_results --demo -o output.svg
python template_renderer.py trial_results --demo --png --scale 2 -o output.svg

from svglue_templates.template_renderer import render_template, save_svg, save_png
from pathlib import Path

svg = render_template("trial_results", {
    "trial_name": "PARADIGM-HF",
    "primary_hr": "0.80",
    "primary_ci": "95% CI: 0.73-0.87",
    "primary_p": "P < 0.001",
    "source": "McMurray JJV et al. N Engl J Med. 2014",
})
save_svg(svg, Path("trial_results.svg"))
save_png(svg, Path("trial_results.svg"), scale=2)  # 1600×1200 PNG

Architecture Diagrams (Phase 2.3)

Publication-grade clinical pathways and research flow diagrams using mingrammer/diagrams.

pip install diagrams
brew install graphviz  # macOS

Modules: treatment_pathways (HF, ACS, AF algorithms), research_flows (CONSORT, PRISMA, methodology), healthcare_arch (hospital system, cardiology dept, data pipeline).

from arch_diagrams.treatment_pathways import create_heart_failure_pathway
from arch_diagrams.research_flows import create_consort_diagram

create_heart_failure_pathway(output_path="outputs/hf_pathway", format="png")
# → GDMT initiation → ACEi/ARNi → Beta-blocker → MRA → SGLT2i → Device therapy

# CONSORT diagram (see drawsvg section for pure-Python SVG alternative)
create_consort_diagram(
    enrolled=500, randomized=400,
    treatment_n=200, control_n=200,
    treatment_completed=180, control_completed=175,
    treatment_analyzed=200, control_analyzed=200,
    output_path="outputs/consort", format="png"
)

Diagram color coding:

| Element | Color | Meaning | |---------|-------|---------| | Assessment | Blue (#2d6a9f) | Diagnostics | | Decision | Orange (#e65100) | Stratification | | Treatment | Green (#2e7d32) | Active therapy | | Danger/Critical | Red (#c62828) | ICU, exclusions |

python arch_diagrams/treatment_pathways.py  # all pathways
python arch_diagrams/research_flows.py      # all research flows
python arch_diagrams/healthcare_arch.py     # all architecture diagrams

Manim Animations (Phase 3.1)

Educational animations for mechanisms, survival curves, and ECG fundamentals.

python -m venv .venv-manim
.venv-manim/bin/python -m pip install manim

Key scenes:

| Key | Scene Class | Description | |-----|-------------|-------------| | mechanism | MechanismOfActionScene | 4-step mechanism flow with outcome callout | | kaplan_meier | KaplanMeierScene | Stepwise survival curves + HR label | | ecg_wave | ECGWaveScene | Normal sinus rhythm with labels |

Full catalog: manim_animations/scene_catalog.json — categories include cardiometabolic, ACS/CAD, arrhythmia, imaging/DX, statistics, devices, anatomy.

cd skills/cardiology/visual-design-system

python scripts/render_manim.py --list
python scripts/render_manim.py mechanism     --quality m --format mp4
python scripts/render_manim.py kaplan_meier --quality h --preview
python scripts/render_manim.py ecg_wave     --quality l --manim-bin .venv-manim

• Outputs → outputs/manim/
• Colors and fonts sourced from design tokens via manim_animations/theme.py
• Carousel slides with animation_scene route to Manim via carousel-generator-v2

Directory Structure

visual-design-system/
├── SKILL.md
├── tokens/
│   ├── index.py            # Main token loader
│   ├── colors.json
│   ├── typography.json
│   ├── spacing.json
│   └── shadows.json
├── scripts/
│   ├── token_validator.py
│   ├── generate_infographic.py
│   └── render_manim.py
├── satori/                 # Phase 1.2 - React → SVG → PNG
├── svg_diagrams/           # Phase 1.3 - Pure Python SVG
├── components/             # Phase 2.1 - Component Library
├── svglue_templates/       # Phase 2.2 - SVG Templates
│   └── TEMPLATE_REFERENCE.md  # Full field docs for all 5 templates
├── arch_diagrams/          # Phase 2.3 - Architecture Diagrams
├── manim_animations/       # Phase 3.1 - Manim scenes + catalog
├── references/
│   ├── nature_guidelines.md
│   └── color_palettes.md
└── outputs/

References

• Nature Figure Guidelines
• Paul Tol's Colorblind-Safe Palettes
• WCAG 2.1 Contrast Requirements
• Satori Documentation
• references/nature_guidelines.md — full Nature journal figure standards
• references/color_palettes.md — complete palette definitions
• svglue_templates/TEMPLATE_REFERENCE.md — all template field definitions

Last Updated: 2026-01-01 Maintainer: Dr. Shailesh Singh