AI Agent Hub
Back to plugins
🎮

dsh-chinese-thinking

Fun & Dress-up Updated 2026.08.25

Run the following command in DeepSeek Harness:

dsh plugin install GongYuanCaiJi/dsh-chinese-thinking

Paste the following prompt into your AI chat to install this plugin:

Clone the repository at https://github.com/GongYuanCaiJi/dsh-chinese-thinking into your local workspace, then run dsh plugin install GongYuanCaiJi/dsh-chinese-thinking in DeepSeek Harness; restart dsh and start a new conversation to activate it.

About this plugin

The most jarring moment in any Chinese-language conversation with a DeepSeek Harness agent is watching it drift into English mid-reply. dsh-chinese-thinking injects a configurable "Chinese Thinking" section into the agent system prompt so that when you write in Chinese, the agent thinks and answers in Chinese end to end, while code, commands, and proper nouns remain in English.

The preset goes well beyond a single language directive. It ports verbatim five convention sets from superpowers-zh: CJK-Latin spacing and full-width punctuation rules, bilingual writing guidelines covering term retention, first-mention annotation, and avoiding over-translation, criteria for deciding when code comments should be Chinese versus English, commit-message conventions that keep the type in English while using Chinese for scope, description, and body, and core principles for domestic Git workflow. The prompt body is fully overridable via config.text, the injection order is adjustable via the order key, and the entire section can be toggled off with enabled, so you can tailor every detail to your team style.

If Chinese is your primary working language, you want consistent output language from your agent, or your team follows unified conventions for Chinese code comments and commit messages, this plugin removes the need to remind the agent in every single conversation and lets you focus on what actually matters.

Use Cases

  • Chatting with an agent in Chinese and wanting consistent Chinese thinking and replies without mid-conversation drift
  • Enforcing uniform Chinese code-comment and commit-message conventions across a development team
  • Setting up CJK-Latin typography rules once so you stop reminding the agent in every conversation

Best For

  • Developers who work primarily in Chinese and interact with agents daily at high frequency
  • Engineering leads who need to standardize Chinese output conventions across their team
  • Technical writers who care about CJK-Latin typography in agent-generated content