mcpField
M
md-to-feishu

md-to-feishu

github.com/zc277584121Documents03.8k installsUpdated 9w ago
Repository

About md-to-feishu

A skill for turning local Markdown files into Feishu documents, for users publishing written materials there. It can infer a title from the first heading and converts Mermaid diagrams to images before upload.

Key features

  1. 01Infer title from the first heading
  2. 02Upload images automatically
  3. 03Convert Mermaid diagrams to images before upload
  4. 04Prefer uvx or an installed feishu-docx

Use cases

  1. 01Publish Markdown notes as Feishu documents
  2. 02Convert Markdown with Mermaid diagrams into a Feishu document

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/zc277584121/marketing-skills --skill md-to-feishu

After installing, type/md-to-feishuin Claude Code to confirm it loaded.

Details

From the project README

Markdown to Feishu Document

Convert a local Markdown file into a Feishu document, with automatic image upload.

User Input

The user only needs to provide a Markdown file path. Title is optional — if not provided, extract it automatically (see below).

Step 1: Determine the Document Title

  1. Read the Markdown file and look for the first # heading — use that as the title.
  2. If no # heading exists, scan the content and generate a concise, descriptive title based on the topic.
  3. If the user explicitly provides a title, use that instead.

Step 2: Check the Runtime Environment

Try each option in order. Use the first one that works.

Option A: uvx (preferred)
which uvx

If uvx is available, the run command is:

uvx feishu-docx create "<TITLE>" -f <MARKDOWN_FILE_PATH>

If uvx runs with Python < 3.11, add --python 3.11:

uvx --python 3.11 feishu-docx create "<TITLE>" -f <MARKDOWN_FILE_PATH>
Option B: feishu-docx already installed
which feishu-docx

If found, check Python version:

python3 --version

If Python >= 3.11, the run command is:

feishu-docx create "<TITLE>" -f <MARKDOWN_FILE_PATH>
Option C: Nothing available — install guidance

If neither uvx nor feishu-docx is found, tell the user:

feishu-docx requires Python >= 3.11. Install with one of:

# Recommended: install uv, then run directly without global install
curl -LsSf https://astral.sh/uv/install.sh | sh
uvx feishu-docx create "Title" -f file.md

# Or: install globally with pip (Python >= 3.11 required)
pip install feishu-docx

Feishu credentials must be configured first:

feishu-docx config set --app-id <APP_ID> --app-secret <APP_SECRET>

Then stop and wait for the user to set up the environment.

Step 3: Pre-process Mermaid Blocks

The feishu-docx tool cannot handle Mermaid code blocks. Before uploading, check if the Markdown contains any ```mermaid blocks and convert them to images first.

3a: Scan for Mermaid blocks

Read the Markdown file and check if it contains any ```mermaid fenced code blocks. If none are found, skip to Step 4.

3b: Create a temporary copy

Copy the original Markdown file to a temp file in the same directory (so relative image paths still work):

<original-name>.feishu-tmp.md

For example: blog_post.md → blog_post.feishu-tmp.md

All subsequent modifications happen on this temp copy. The original file is never modified.

3c: Render Mermaid diagrams to PNG

For each ```mermaid ... ``` block in the temp file, render it to a PNG image using the mermaid.ink API:

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.

Save rendered images to the same directory as the Markdown file, using descriptive filenames based on diagram content:

  • GOOD: mermaid-architecture-overview.png, mermaid-data-flow.png
  • BAD: mermaid-1.png, diagram.png
3d: Replace Mermaid blocks with image references

In the temp copy, replace each ```mermaid ... ``` block with a Markdown image reference:

![Architecture overview](mermaid-architecture-overview.png)

Use relative paths from the temp file to the rendered images.

3e: Use the temp file for upload

From this point, the temp file becomes the <MARKDOWN_FILE_PATH> used in Step 4.

Step 4: Run the Command

  • Run from the directory where the Markdown file lives (or where its relative image paths resolve correctly), so that local image references work.
  • The tool will:
    • Convert Markdown blocks to Feishu format
    • Automatically upload local images referenced in the Markdown
    • Wait ~10s for block consistency before uploading images

Step 5: Clean Up

If a temp file was created in Step 3:

  • Delete the temp Markdown file (*.feishu-tmp.md)
  • Delete all rendered Mermaid PNG files created in Step 3c (they were only needed for the upload)

Step 6: Report Result

Show the user:

  • Number of blocks converted and images uploaded
  • The created document ID
  • Success or failure status

If it fails with authentication errors, remind the user to configure credentials:

feishu-docx config set --app-id <APP_ID> --app-secret <APP_SECRET>

Comments

Markdown supported