Skill: Mermaid to Image

Convert ```mermaid code blocks in Markdown (or other text) files into PNG images, and replace the code blocks with image references. Useful for platforms that don't render Mermaid natively (GitHub Pages/Jekyll, Dev.to, etc.).

When to Use

• The user asks to convert Mermaid diagrams in a file to images
• The user wants to render specific Mermaid code blocks as PNG
• A publishing workflow requires static images instead of Mermaid code blocks

Workflow

Step 1: Identify target files

The user may specify:

• A single file: convert mermaid blocks in docs/architecture.md
• Multiple files: convert mermaid in all files under docs/
• A specific code block: convert the second mermaid block in README.md

Scan the target file(s) for ```mermaid code blocks. Report how many blocks were found and in which files before proceeding.

Step 2: Determine the image output directory

Check the project structure to find where images are typically stored:

# Look for common image directories
ls -d images/ img/ assets/ assets/images/ static/images/ docs/images/ 2>/dev/null

If a clear image directory exists (e.g., images/, assets/images/), use it. Create a subdirectory by topic if appropriate (e.g., images/<topic>/).

If no image directory is obvious or multiple candidates exist, ask the user:

Where should I save the rendered Mermaid images?

1. images/ (create new)
2. assets/images/
3. docs/figures/
4. Custom — enter a path

Tip: add "remember" to save this choice to CLAUDE.local.md.

If the user says "remember", save the choice to the project's CLAUDE.local.md:

## Mermaid Image Output

- **Image directory**: `<chosen-path>`

On subsequent runs, check CLAUDE.local.md for a ## Mermaid Image Output section and use it directly.

Step 3: Render each diagram to PNG

Use the mermaid.ink API to render diagrams. Run this Python snippet for each block:

import base64, urllib.request

def render_mermaid(code: str, output_path: str):
    """Render a Mermaid diagram to PNG via mermaid.ink API."""
    encoded = base64.urlsafe_b64encode(code.encode()).decode()
    url = f"https://mermaid.ink/img/{encoded}?bgColor=white"
    req = urllib.request.Request(url, headers={"User-Agent": "Mozilla/5.0"})
    resp = urllib.request.urlopen(req, timeout=30)
    with open(output_path, "wb") as f:
        f.write(resp.read())

Important: The User-Agent header is required — mermaid.ink returns 403 without it.

Naming convention

Use descriptive filenames based on the diagram content, not generic names:

• GOOD: architecture-overview.png, data-flow.png, heartbeat-sequence.png
• BAD: mermaid-1.png, diagram.png, image1.png

Step 4: Replace code blocks with image references

Replace each ```mermaid ... ``` block with a Markdown image reference using a relative path from the file to the image:

![Architecture overview](images/topic/architecture-overview.png)

If the project uses absolute URLs (e.g., GitHub Pages), use those instead:

![Architecture overview](https://example.github.io/images/topic/architecture-overview.png)

Choose the link style that matches the project's existing image references. If unsure, use relative paths.

Step 5: Report results

After processing, summarize:

• How many diagrams were converted
• Where the images were saved
• Which files were modified

Edge Cases

• Large diagrams: mermaid.ink may time out on very complex diagrams. If a render fails, report the error and suggest the user simplify the diagram or try an alternative renderer.
• Multiple blocks in one file: process all blocks in order, give each a unique descriptive filename.
• Already-rendered blocks: if a mermaid block already has a corresponding image (commented out or adjacent), skip it or ask the user.
• Non-Markdown files: the same approach works for any text file containing mermaid code blocks (e.g., .rst, .txt).