chinese-quote-normalizer

name

chinese-quote-normalizer

description

Convert English straight quote marks to Chinese typography in Chinese-language writing. Use when the user asks to replace English-state quotes, straight quotes, ASCII quotes, or Markdown/Feishu/Lark document quotes with Chinese quote marks; especially for .md, .markdown, .mdx, .txt files, pasted Chinese text, or Feishu/Lark docs where visible text should use “...” instead of "..." while preserving code, attributes, and document structure.

Chinese Quote Normalizer

Goal

Replace visible English straight double quotes with Chinese curved quotes:

• "内容" -> “内容”

Optionally replace straight single quotes when the user explicitly asks:

• '内容' -> ‘内容’

Default to double quotes only. Preserve code syntax, Markdown structure, XML/HTML tag attributes, and Feishu/Lark document block structure.

Local Markdown/Text Files

Use the bundled script for local files. Paths below are relative to this skill's own directory; run them from the skill folder, or prefix with the skill's absolute path on the current machine.

python3 scripts/normalize_quotes.py --write <file-or-directory>

Useful modes:

# Preview only
python3 scripts/normalize_quotes.py <file-or-directory>

# Fail if changes would be needed
python3 scripts/normalize_quotes.py --check <file-or-directory>

# Also convert likely paired single quotes
python3 scripts/normalize_quotes.py --write --single <file-or-directory>

The script recursively processes Markdown-like files in directories (.md, .markdown, .mdx, .txt) and skips:

• YAML frontmatter at the top of Markdown files
• fenced code blocks
• inline code spans
• HTML/XML tags and their attributes

After writing, run --check or re-run preview to confirm no remaining visible straight double quotes.

Feishu/Lark Docs

Use lark-doc first for Feishu/Lark URLs. Do not overwrite the whole document just to replace quotes.

Workflow:

1. Fetch the document as full XML:

   lark-cli docs +fetch --api-version v2 --doc "<url-or-token>" --detail full --format json
   

2. Extract only visible text segments outside XML tags. Ignore id="...", width attributes, URLs, tokens, and other tag attributes.

3. Find paired straight double quote phrases in visible text with a pattern equivalent to:

   "([^"]*)"
   

4. For each unique phrase, replace it with Chinese quotes using str_replace:

   lark-cli docs +update --api-version v2 --doc "<url-or-token>" \
     --command str_replace \
     --pattern "<regex-escaped-original-phrase>" \
     --content "“内容”"
   

   Regex-escape metacharacters in the original phrase, including +, ., ?, (, ), [, ], {, }, |, ^, $, *, and \.

5. If a full phrase fails to change because of regex or document matching behavior, replace the left and right halves in sequence, for example:

   lark-cli docs +update --api-version v2 --doc "<url-or-token>" \
     --command str_replace --pattern '"关键词开头' --content '“关键词开头'
   
   lark-cli docs +update --api-version v2 --doc "<url-or-token>" \
     --command str_replace --pattern '关键词结尾"' --content '关键词结尾”'
   

6. Fetch again and validate only visible text:

   • remaining visible " count is 0
   • “ count equals ” count

Pasted Text

For pasted Chinese prose, convert paired straight double quotes directly:

1. Scan left to right outside code blocks/spans if Markdown is present.
2. Alternate opening and closing marks: first " -> “, next " -> ”.
3. Keep unmatched quotes visible in the result only if the source text has an odd quote count; mention the ambiguity.

Safety Notes

• Do not convert quotes inside programming code, JSON, shell commands, YAML frontmatter, XML/HTML attributes, or Markdown inline code unless the user explicitly asks to normalize code too.
• Treat apostrophes in English words (don't, user's) as apostrophes, not Chinese single quotes.
• For live Feishu/Lark documents, always refetch and validate after edits.