Contextual Chinese Quotes Assistant

Problem

When Chinese text appears alongside code, punctuation becomes a parsing boundary, not a style preference. Replacing ASCII quotes with Chinese quotes, changing quote marks in YAML keys and values, or “beautifying” backticks inside Markdown fenced blocks can invalidate JSON, Typst, shell, LaTeX, and prompt examples. zh-quotes targets this mixed-context risk, focusing on quotation marks, punctuation, and symbol choices where natural Chinese typography collides with machine syntax.

How It Works

• Classify the syntax layer first: determine whether the token is pure prose, a code span, a Typst command, a JSON/YAML value, a shell argument, a regex, or a nested prompt example.
• Let syntax outrank typography: preserve language-native quotes in code and data structures, while using “” or ‘’ only inside Chinese natural-language fragments.
• Resolve nesting level by level: for Markdown → Python → Chinese string → Chinese quotation, each outer layer follows its own grammar and the inner Chinese text follows readable Chinese punctuation.
• Validate before emitting: confirm quote legality, parser compatibility, and nested quotation closure instead of normalizing valid syntax into visually “prettier” symbols.

Boundaries

This skill does not handle fonts, spacing, or visual polish, and it should not rewrite business logic. It is useful for documents, prompts, configuration, and generated code examples that contain Chinese, especially in JSON, YAML, Typst, shell, regex, and fenced Markdown blocks where exact symbols matter.