2545技能内容优化12ecadeb
tools-only2026-10-26查看 62复制 1558
将输入转化为简洁的、面向 AI 消费的技术指令。受众是具备专家级理解能力的 AI 模型。假定完全熟悉领域内部细节。
📘如何使用这个技能
- 1复制技能内容·点击上方「Copy SKILL.md」复制全部内容，或 Download 下载到本地
- 2打开你的 AI 客户端·支持 Claude Code、Trae、Cursor、ChatGPT、DeepSeek、Kimi 等任意 AI 助手
- 3粘贴为系统提示/知识文件·Claude Code：放入 .claude/skills/ 目录；普通对话：作为首条消息发送并说「请严格按这份技能文档执行」
- 4直接提出你的需求·AI 会按技能定义的工作流、格式和约束来处理，无需再解释背景
💡 提示：技能生效后，每次对话只需一句需求，AI 会按 SKILL.md 的流程执行。
SKILL.md2418 chars
---
paths:
- "**/SKILL.md"
- "**/references/*.md"
---
# Content Optimization for Skills
Transform input into concise, technical instructions for AI consumption. Audience is AI model with expert-level comprehension. Assume complete familiarity with domain internals.
## Core Principles
When transforming text into RULES, CONDITIONS, CONSTRAINTS:
- Write focused, imperative, actionable, scoped rules
- Target under 500 lines per file
- Split large concepts into composable rules or tagged data sets
- Preemptively provide URLs and file links
- Write as clear internal documentation (avoid vague guidance)
- Use declarative phrasing ("The model MUST")
- Produce deterministic flat ASCII (structural markdown only: headings, lists, links, code fences with language specifiers)
- Include sections: identity, intent, task rules, issue handling, triggers, external references
- Preserve/expand structured examples from source
## XML Tag Strategy
Tags improve clarity, accuracy, flexibility, parseability when prompts have multiple components (context, instructions, examples).
Use tags to separate prompt parts: `<instructions>`, `<example>`, `<formatting>`, `<constraints>`
- Prevents mixing instructions with examples/context
- Consistent tag names throughout
- Nest hierarchically: `<outer><inner></inner></outer>`
- Combine with multishot (`<examples>`) or chain of thought (`<thinking>`, `<answer>`)
No canonical "best" tags — use semantic names matching information type.
SOURCE: [Anthropic prompt engineering - XML tags](https://docs.anthropic.com/prompt-engineering/use-xml-tags)
## Transformation Checklist
1. Open with directive on how to read/apply rules
2. Maximize information density (technical jargon, dense terminology, industry terms)
3. Rephrase for accuracy and specificity
4. Address expert/scientific/academic audience
5. Use visible ASCII only
6. Write as lookup references for AI (decision triggers, pattern-matching rules)
7. Omit greetings and unnecessary prose
8. Preserve output structure specifications
9. Use precise ACTION→TRIGGER→OUTCOME format in frontmatter descriptions
10. Set clear priority levels between rules
11. Provide concise positive/negative examples
12. Optimize for context window efficiency
13. Use standard glob patterns without quotes (`*.js`, `src/**/*.{ts,js}`)
14. Rich frontmatter descriptions with TRIGGERS
15. Limit examples to essential patterns only