translate-content
关于
This skill translates technical documentation and code resources between locales while preserving code blocks, IDs, and file structure. It is used to localize content for a new language or update existing translations after source changes. Key features include automated scaffolding, frontmatter management, and freshness tracking for batch operations.
快速安装
Claude Code
推荐npx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/translate-content在 Claude Code 中复制并粘贴此命令以安装该技能
技能文档
Translate Content
EN src → target locale, preserve tech accuracy + structural integrity.
Use When
- Localize skill, agent, team, guide → supported lang
- Update stale translation after src changes
- Batch-translate multi items in domain | content type
- Initial translations for new locale
In
- Required: Content type —
skills,agents,teams,guides - Required: Item ID — name/identifier (e.g.
create-r-package) - Required: Target locale — IETF BCP 47 (e.g.
de,zh-CN,ja,es) - Optional: Batch list — multiple IDs in sequence
Do
Step 1: Read EN Src
1.1. Determine src path:
- Skills:
skills/<id>/SKILL.md - Agents:
agents/<id>.md - Teams:
teams/<id>.md - Guides:
guides/<id>.md
1.2. Read entire src to understand context, structure, content.
1.3. ID sections that stay in EN:
- All code blocks (fenced w/ triple backticks)
- Inline code (backtick-wrapped)
- YAML frontmatter field names + tech values (
name,tools,model,priority,skillslist entries,allowed-tools,tags,domain,language) - File paths, URLs, cmd examples
<!-- CONFIG:START -->/<!-- CONFIG:END -->blocks in teams
Got: Full understanding w/ clear mental separation translatable prose vs preserved tech.
If err: Src not found → verify ID exists in registry. Check typos in content type | ID.
Step 2: Scaffold Translation File
2.1. Run scaffolding script:
npm run translate:scaffold -- <content-type> <id> <locale>
2.2. File exists → read to check stale vs current.
2.3. Verify scaffolded has translation frontmatter:
locale— matches targetsource_locale—ensource_commit— current git short hashtranslator— attributiontranslation_date— today
Got: Scaffolded at i18n/<locale>/<content-type>/<id>/SKILL.md (or .md for others) w/ correct frontmatter.
If err: Scaffold script fails → create dir manually w/ mkdir -p + copy src. Add frontmatter manually.
Step 3: Translate Description
3.1. Translate description field in YAML frontmatter → target locale.
3.2. Skills: top-level frontmatter. Agents/teams/guides: also top-level.
3.3. Concise — match length + style of original.
Got: Description field has idiomatic translation accurate to original meaning.
If err: Ambiguous → keep closer to literal vs risk misinterpretation.
Step 4: Translate Prose Sections
4.1. Translate prose section by section:
- Headings ("## When to Use" → "## Wann verwenden" in DE)
- Paragraph text
- List item text (NOT list item code/paths)
- Table cell text (NOT table cell code/values)
4.2. Preserve exactly:
- Code blocks (``` fenced + indented)
- Inline code (
backtick-wrapped) - File paths + URLs
- Skill/agent/team IDs in cross-refs
- YAML/JSON config examples
- Cmd-line examples
**Expected:**+**On failure:**markers (translate label, keep structure)
4.3. Skills: translate std section names:
- "When to Use" → locale equiv
- "Inputs" → locale equiv
- "Procedure" → locale equiv
- "Validation" → locale equiv
- "Common Pitfalls" → locale equiv
- "Related Skills" → locale equiv
4.4. Agents: translate:
- Purpose, Capabilities, Available Skills (section name only — skill IDs stay EN), Usage Scenarios, Best Practices, Examples, Limitations, See Also
4.5. Teams: translate:
- Purpose, Team Composition (prose only — IDs stay EN), Coordination Pattern, Task Decomposition, Usage Scenarios, Limitations
4.6. Guides: translate:
- All prose, troubleshooting, table descriptions
- Keep cmd examples, code blocks, config snippets in EN
Got: All prose translated idiomatically. Code blocks identical to EN src. Cross-refs use EN IDs.
If err: Uncertain tech term → keep EN w/ parenthetical translation. e.g. "Staging-Bereich (Staging Area)" in DE.
Step 5: Verify Structural Integrity
5.1. Same # sections as src.
5.2. Skills: verify all req sections:
- YAML frontmatter w/
name,description,allowed-tools,metadata - When to Use, Inputs, Procedure, Validation, Common Pitfalls, Related Skills
5.3. Verify code blocks identical to EN src (diff fenced blocks).
5.4. Check line count: skills must be ≤ 500.
5.5. Verify name field matches EN src exact (= ID, never translated).
Got: Structurally valid file passes validation.
If err: Cmp section-by-section w/ EN src. Restore missing.
Step 5.5: Verify Prose is Translated
5.5.1. Sample 3 prose paragraphs from body. Diff sections — not headings, not code blocks, not frontmatter.
5.5.2. Confirm each in target lang, not EN.
5.5.3. Any in EN → translation incomplete. Return Step 4, translate remaining EN before continuing.
Got: All 3 sampled in target lang, body text translated — not just headings + frontmatter.
If err: ID which sections still EN prose. Translate before Step 6.
Step 6: Write Translated File
6.1. Write complete translated content to target path via Write | Edit.
6.2. Verify file at expected path:
- Skills:
i18n/<locale>/skills/<id>/SKILL.md - Agents:
i18n/<locale>/agents/<id>.md - Teams:
i18n/<locale>/teams/<id>.md - Guides:
i18n/<locale>/guides/<id>.md
Got: Translated file written to disk at correct path.
If err: Check dir exists. Create w/ mkdir -p if needed.
Check
- Translated file at
i18n/<locale>/<type>/<id> -
namefield matches EN src exact -
localefield matches target -
source_commitset to valid git short hash - All code blocks identical to EN src
- All cross-ref IDs (skills, agents, teams) in EN
- File ≤ 500 lines (skills)
-
npm run validate:translationsreports no issues - Prose reads idiomatically in target lang
Traps
- Translate code blocks: Code, cmds, config stay EN. Only surrounding prose.
- Translate
namefield:name= canonical ID. Never translate. - Translate tag values: Tags in
metadata.tagsstay EN for cross-locale consistency. - Inconsistent terminology: Same translation for tech term throughout file + across files in same locale.
- Literal translation of idioms: Meaning, not words. "Common Pitfalls" → locale's natural equiv, not word-for-word.
- Missing
source_commit: Without → freshness tracking breaks. Always include. - Batch throughput > quality: Scaffolding-only (headings translated, body still EN) ≠ valid translation. Fewer complete > many partial.
- Exceeding 500 lines: Translations may expand ~10-20% vs EN. Near limit → tighten prose vs remove content.
→
- create-skill — understand SKILL.md structure being translated
- review-skill-format — validate translated skill structure
- evolve-skill — update skills that changed since translation
GitHub 仓库
相关推荐技能
content-collections
元Content Collections 是一个 TypeScript 优先的构建工具,可将本地 Markdown/MDX 文件转换为类型安全的数据集合。它专为构建博客、文档站和内容密集型 Vite+React 应用而设计,提供基于 Zod 的自动模式验证。该工具涵盖从 Vite 插件配置、MDX 编译到生产环境部署的完整工作流。
polymarket
元这个Claude Skill为开发者提供完整的Polymarket预测市场开发支持,涵盖API调用、交易执行和市场数据分析。关键特性包括实时WebSocket数据流,可监控实时交易、订单和市场动态。开发者可用它构建预测市场应用、实施交易策略并集成实时市场预测功能。
creating-opencode-plugins
元该Skill帮助开发者创建OpenCode插件,用于接入命令、文件、LSP等25+种事件。它提供了插件结构、事件API规范和JavaScript/TypeScript实现模式,适合需要拦截操作、扩展功能或自定义事件处理的场景。开发者可通过它快速构建响应式模块来增强OpenCode AI助手的能力。
sglang
元SGLang是一个专为LLM设计的高性能推理框架,特别适用于需要结构化输出的场景。它通过RadixAttention前缀缓存技术,在处理JSON、正则表达式、工具调用等具有重复前缀的复杂工作流时,能实现极速生成。如果你正在构建智能体或多轮对话系统,并追求远超vLLM的推理性能,SGLang是理想选择。
