Skill: Mermaid to GIF

Convert Mermaid diagrams into animated GIFs with rich animation effects. Supports .mmd files and extracting ```mermaid code blocks from .md files.

Prerequisites: FFmpeg, Python 3.8+, Playwright (pip install playwright && playwright install chromium)

---

When to Use

• The user wants to convert Mermaid diagrams to animated GIFs
• The user has .mmd files or .md files containing mermaid code blocks
• The user needs animated visuals for presentations, documentation, or social media
• The user wants to batch-convert all mermaid blocks in a document

---

Context-Aware Style Selection

IMPORTANT: When converting mermaid blocks from .md files, read the surrounding markdown context to choose the most appropriate animation style for each diagram. Do NOT blindly apply the same style to all blocks.

Decision Guide

1. Read the markdown text around each mermaid block — understand what the diagram is illustrating
2. Match the style to the semantic meaning:

Context Clue	Recommended Style	Reasoning
Data pipeline, ETL flow, request/response path	pulse-flow	Flowing dashed lines convey data movement
Architecture layers, org chart, hierarchy	progressive	Elements activate layer-by-layer
Step-by-step process, tutorial walkthrough	highlight-walk	Spotlight guides the reader through each step
System overview, title diagram, simple reference	wave	Brightness ripple adds life without distraction
Sequence diagram with message flow	progressive	Messages activate one by one in conversation order
Class/ER diagram (reference/static)	progressive or wave	Structure lights up or gets a subtle ripple

3. Consider special handling:

   • If the surrounding text says "data flows from A to B", use pulse-flow even for a simple flowchart
   • If the text describes "three layers" or "two tiers", use progressive to activate layer-by-layer
   • If the diagram is decorative or supplementary, use wave to keep it simple
   • For very large or complex diagrams, prefer wave or shorter --duration to keep GIF size reasonable

4. Per-block style override: When batch-processing a .md file, you may need to run the script multiple times with different styles, extracting specific blocks. Or process the whole file with a sensible default and re-run individual blocks that need different treatment.

Example: Context-Aware Processing

## Data Ingestion Pipeline        ← context: "pipeline" → pulse-flow
[mermaid block: graph LR with ETL stages]

## System Architecture            ← context: "architecture" → progressive
[mermaid block: graph TD with layers]

## Quick Reference                ← context: "reference" → wave
[mermaid block: simple diagram]

---

Default Workflow

Single .mmd file

python ${CLAUDE_SKILL_ROOT}/scripts/mermaid_to_gif.py diagram.mmd

Markdown file with mermaid blocks

python ${CLAUDE_SKILL_ROOT}/scripts/mermaid_to_gif.py document.md -o ./images/

This extracts all ```mermaid code blocks and generates one GIF per block.

Multiple files

python ${CLAUDE_SKILL_ROOT}/scripts/mermaid_to_gif.py *.mmd -o ./gifs/
python ${CLAUDE_SKILL_ROOT}/scripts/mermaid_to_gif.py doc1.md doc2.md -o ./gifs/

Replacing mermaid blocks in markdown

After generating GIFs, replace the original ```mermaid code blocks with image references:

![Flow Diagram](images/document-1.gif)

Use descriptive alt text based on the diagram content. The image path should be relative to the markdown file.

---

Animation Styles

All styles keep the diagram fully visible from frame 1 — no elements start hidden or fade in from zero. Every style adds motion while the user can see the complete diagram structure at all times.

Style	Effect	Best For
progressive (default)	All elements start dimmed (25% opacity), activate sequentially to full brightness; edges draw in with stroke animation	Flowcharts, architecture, hierarchy
highlight-walk	All elements start dimmed (15%); a spotlight with blue glow moves through each element, leaving visited ones bright	Step-by-step process, tutorials
pulse-flow	All elements fully visible; edges become flowing dashed lines (uniform dash size and speed)	Data flow, pipelines, request paths
wave	All elements fully visible; a brightness pulse + blue glow ripple sweeps through elements sequentially	Simple diagrams, overviews, reference

Animation Details

• progressive: Elements start at 25% opacity (diagram structure always visible). Nodes, edges, and labels activate in interleaved order (node → edge → node → edge) following flow direction. Edges use stroke-dashoffset to draw in visually. Activation is fast (8% of total duration per element).
• highlight-walk: All elements start at 15% opacity. A spotlight (with blue glow) moves through elements in order, leaving visited elements at 90% opacity. The whole diagram is visible as a "ghost" before the spotlight reaches each element.
• pulse-flow: All elements at full opacity. Edge paths get a uniform dashed pattern (10px dash + 6px gap) that flows at a fixed speed (200px/cycle), so all edges animate at the same pace regardless of length.
• wave: All elements at full opacity. A brightness pulse (1.0→1.4→1.0) with blue glow sweeps through elements sequentially. No position changes — purely a visual ripple effect.

---

Common Options

Flag	Default	Description
-o, --output-dir	Same as input	Output directory for generated GIFs
-s, --style	progressive	Animation style (see table above)
--fps	10	Frames per second
--duration	4.0	Animation duration in seconds
--hold	1.0	Hold last frame before looping (seconds)
--theme	default	Mermaid theme: default, dark, forest, neutral
--bg	#ffffff	Background color (hex)
--padding	40	Padding around diagram in pixels
--scale	2	Render scale factor (2 = retina quality)
--custom-css	—	Path to custom CSS file
--no-loop	—	Play GIF once instead of looping

---

Examples

# Dark theme with faster animation
python ${CLAUDE_SKILL_ROOT}/scripts/mermaid_to_gif.py arch.mmd --theme dark --bg "#1a1a2e" --duration 3

# High FPS for smoother animation
python ${CLAUDE_SKILL_ROOT}/scripts/mermaid_to_gif.py flow.mmd --fps 15 --duration 5

# Batch convert all mermaid blocks from a doc
python ${CLAUDE_SKILL_ROOT}/scripts/mermaid_to_gif.py README.md -o ./images/

# Custom CSS for special effects
python ${CLAUDE_SKILL_ROOT}/scripts/mermaid_to_gif.py diagram.mmd --custom-css my-style.css

# No loop, suitable for one-time playback
python ${CLAUDE_SKILL_ROOT}/scripts/mermaid_to_gif.py intro.mmd --no-loop --duration 6

# Lower resolution for smaller file size
python ${CLAUDE_SKILL_ROOT}/scripts/mermaid_to_gif.py diagram.mmd --scale 1

---

Custom CSS

Create a CSS file to customize the appearance of the diagram during animation:

/* Rounded nodes with shadow */
.node rect {
    rx: 10;
    filter: drop-shadow(2px 2px 4px rgba(0,0,0,0.3));
}

/* Thicker edge lines */
.edgePath path {
    stroke-width: 2.5;
}

/* Custom background for actors (sequence diagram) */
.actor {
    fill: #e8f4f8;
}

Pass it via --custom-css:

python ${CLAUDE_SKILL_ROOT}/scripts/mermaid_to_gif.py diagram.mmd --custom-css my-style.css

---

How It Works

1. Parse input — extract Mermaid code from .mmd or .md files
2. Generate HTML — embed Mermaid.js (CDN) + animation JS/CSS in a self-contained HTML file
3. Render — Mermaid.js renders the diagram to SVG in headless Chromium via Playwright (with 2x device scale for retina quality)
4. Scale — small SVGs are automatically scaled up to minimum 700px CSS width for readability
5. Animate — JS animation engine exposes setProgress(t) for frame-by-frame control (t: 0→1). Elements are collected, sorted by position (respecting LR/TB direction), and animated in interleaved node-edge order
6. Capture — Playwright takes a screenshot at each frame step
7. Assemble — FFmpeg two-pass palette encoding (palettegen → paletteuse with Floyd-Steinberg dithering)

---

Important Notes

• Internet required: Mermaid.js is loaded from CDN at render time
• Supported diagram types: flowchart, sequence, class, state, ER, gitgraph, mindmap, pie, gantt, and more
• No hidden elements: all 4 styles keep the diagram visible from frame 1 — no waiting for elements to appear
• Fallback behavior: for unrecognized diagram types or when no animatable elements are detected, falls back to a whole-diagram opacity reveal
• Resolution: default scale=2 produces retina-quality images (~1400-1600px wide). Use --scale 1 for smaller files
• GIF size: for very large outputs, reduce FPS to 8, shorten duration, use --scale 1, or use wave style