name

keyword-match-optimization

description

This skill should be used when a user submits a request in Chinese that does not directly match any Chinese-language skill. It translates Chinese user input into concise, standard English instructions to improve skill matching accuracy, while preserving native Chinese skill matching priority. All final responses to the user are delivered in Simplified Chinese (code, file paths, and technical terms excepted). Use this skill when Chinese input fails to trigger an English skill, when a Chinese-to-English instruction translation is needed, or when the user's Chinese request should be routed to an English-language skill.

Keyword Match Optimization

Overview

This skill acts as a translation bridge between Chinese user input and English skill triggers. It solves a core pain point: most skills in the Claude Code ecosystem use English trigger keywords, but Chinese-speaking users naturally describe their needs in Chinese, resulting in low skill matching rates. This skill translates Chinese requests into standard English instructions, triggers the corresponding English skill, and ensures the final response is presented in Simplified Chinese.

Workflow

Chinese user input
  │
  ▼
Check: Does input directly match a Chinese skill?
  ├── YES → Route to that Chinese skill (no intervention)
  └── NO  ↓
Translate to concise English instruction
  │
  ▼
Match and invoke corresponding English skill
  │
  ▼
Force Simplified Chinese output (except code/paths/terms)
  │
  ▼
Return natural, fluent Simplified Chinese response

Step 1: Receive and Analyze Input

Receive the user's Chinese input and determine the intent.

1.1 Check Chinese skill keywords first

Read references/chinese-skill-keywords.md for the mapping of known Chinese-language skills.

If the input directly matches a Chinese skill's trigger keyword:

Do not translate. Route to that skill immediately.
End workflow.

1.2 If no Chinese skill match

Proceed to translation.

Step 2: Translate to English Instruction

Convert the Chinese request into a concise, standard English instruction optimized for skill triggering.

Translation rules:

Rule	Detail
Simplify	Remove filler words, politeness markers, modifiers — keep only core intent
Verb-first	Use imperative form: Create, Review, Fix, Generate, Deploy, Optimize...
Match skill names	Prioritize keywords that appear in known skill names and descriptions
Preserve technical terms	Framework names, tool names, protocol names stay in English
Remove noise	Only translate the instruction-relevant portion of the input

Translation examples:

Chinese input	English instruction
帮我做一下代码审查	review code
我想部署到 Vercel	deploy to Vercel
写一个 React 组件	create React component
帮我优化数据库查询	optimize database queries
这个 bug 怎么修	debug and fix
帮我设计系统架构	design system architecture
写个单元测试	write unit tests
帮我创建 PR	create pull request
做个无障碍审查	accessibility audit
帮我配置 CI/CD	setup deployment pipeline

For the full mapping table, read references/translation-patterns.md.

Step 3: Invoke the Matched Skill

3.1 Exact match: If the translated instruction precisely matches a skill, invoke it directly.

3.2 Fuzzy match: If no exact match, attempt in order:

Skill name contains key terms from the translated instruction
Skill description contains relevant keywords
Skill tags contain relevant keywords

3.3 No match: If no skill matches:

Inform the user in Simplified Chinese that no specialized skill was found
Ask whether to proceed with general-purpose handling

Step 4: Force Chinese Output

Process the skill's response for language compliance:

Must be Simplified Chinese:

All explanatory text, guidance, and summaries for the user
Question text and option labels (AskUserQuestion)
Error messages and prompts

Keep in English (do not translate):

Code inside code blocks
File paths and filenames
Command-line instructions
Technical proper nouns (React, API, JSON, GraphQL, etc.)
Code comments (preserve as-is from source)

Critical Constraints

Never answer the user's question directly — this skill only translates and routes; the actual answer comes from the matched skill or general capability
Never interfere with existing Chinese skill matching — if input already matches a Chinese skill, do not intervene
Never expose the translation process — the user should perceive a seamless Chinese-in, Chinese-out experience
Never translate code content — code, paths, and commands remain as-is
Always ensure Chinese output reads naturally — like a native speaker wrote it, not machine-translated
Always prioritize known skill name keywords when choosing English translation terms

Resources

references/

translation-patterns.md — Comprehensive Chinese-to-English translation mapping table covering development, deployment, frontend, backend, AI/ML, security, and more
chinese-skill-keywords.md — Registry of known Chinese-language skills and their trigger keywords, used to determine when NOT to translate