refine-skill-design

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Skill: Refine Skill Design

Skill：Skill设计优化

Purpose

用途

As a "Skill for Skills," this skill audits and refactors AI capability definitions in draft form. It applies a senior prompt-engineering perspective to improve logic robustness, scenario coverage, and instruction adherence so each capability meets LLM best practices.

作为一款“面向Skill的Skill”，本Skill负责审核并重构草稿形式的AI能力定义。它以资深提示工程的视角，提升逻辑健壮性、场景覆盖度以及对指令的遵循度，确保每项能力符合LLM最佳实践。

Use Cases

适用场景

New skill onboarding: Expert review after an Agent has generated a new Skill draft.
Quality fixes: When a Skill behaves inconsistently on newer models, align logic and strengthen examples.
Consistency audit: Check that a new Skill matches the tagging system and naming in
```
INDEX.md
```
.
Upgrade: Turn a simple "formatting tool" into a full Agent capability with interaction policy and error handling.

Scope: This skill is for auditing and refactoring existing SKILLs, not creating from scratch. For learning how to create new skills, plan scripts/references/assets, or run init/package scripts, use skills.sh's

skill-creator

(e.g. anthropics/skills).

新Skill入门：Agent生成新Skill草稿后的专家评审。
质量修复：当Skill在较新模型上表现不一致时，对齐逻辑并强化示例。
一致性审核：检查新Skill是否与
```
INDEX.md
```
中的标签系统和命名规则匹配。
升级优化：将简单的“格式化工具”升级为具备交互策略和错误处理的完整Agent能力。

适用范围：本Skill仅用于审核并重构现有SKILL，不支持从零创建。如需学习如何创建新技能、规划脚本/参考资料/资源，或运行初始化/打包脚本，请使用skills.sh的

skill-creator

（例如anthropics/skills）。

Behavior

行为规范

Meta-audit model

元审核模型

Intent: Is Purpose specific enough? Avoid vague terms like "Helper," "Utilities."
Logic: Does Input → Behavior → Output form a clear chain?
Constraints: Do Restrictions cover the most common failure modes in the domain?
Examples: Do Examples progress from simple to complex and include at least one edge case?

意图：Purpose是否足够具体？避免使用“助手”“工具集”等模糊表述。
逻辑：输入→行为→输出是否形成清晰的链路？
约束：Restrictions是否覆盖了该领域最常见的失败场景？
示例：Examples是否从简单到复杂递进，且至少包含一个边缘案例？

Optimization flow

优化流程

Structure: Apply the standard template (YAML, Purpose, Use cases, Behavior, I/O, Restrictions, Self-Check, Examples).
Verbs: Use precise, unambiguous verbs (e.g. "handle" → "parse," "transform," "trim").
Interaction: For complex logic, add "confirm before proceed" or "choose among options."
Metadata: Align tags with
```
INDEX.md
```
and suggest a sensible SemVer version.

结构：应用标准模板（YAML、Purpose、Use cases、Behavior、I/O、Restrictions、Self-Check、Examples）。
动词：使用精准、无歧义的动词（例如将“处理”替换为“解析”“转换”“修剪”）。
交互：针对复杂逻辑，添加“执行前确认”或“选择选项”机制。
元数据：与
```
INDEX.md
```
中的标签对齐，并建议合理的SemVer版本。

Input & Output

输入与输出

Input

输入

A SKILL Markdown document to optimize, or a rough draft.

待优化的SKILL Markdown文档或草稿。

Output

输出

Optimized SKILL: Production-grade Markdown that satisfies the spec.
Diff summary: What was changed and why.
Version suggestion: SemVer recommendation.

优化后的SKILL：符合规范的生产级Markdown内容。
差异总结：变更内容及原因说明。
版本建议：SemVer版本推荐。

Restrictions

约束条件

Do not change intent: Optimization must preserve the skill's core purpose.
Minimize prose: Prefer lists and tables over long README-style text.
Multiple examples: Do not keep only one "happy path" example; include at least one challenging or edge-case example.

不得修改核心意图：优化必须保留Skill的核心用途。
精简文字：优先使用列表和表格，而非长篇README式文本。
多示例要求：不得仅保留“常规路径”示例；至少包含一个具有挑战性的边缘案例示例。

Self-Check

自我检查

Self-apply: Could this skill successfully refine itself?
Clarity: Could an Agent without domain background reproduce results from Behavior?
Compliance: Are all required sections and metadata fields present?

自我验证：本Skill能否成功优化自身？
清晰度：无领域背景的Agent能否根据Behavior部分复现结果？
合规性：是否包含所有必填章节和元数据字段？

Examples

示例

Before

优化前

name: spell-check This skill checks spelling. Input: multilingual text. Output: corrected text.

name: spell-check 本技能用于检查拼写。输入：多语言文本。输出：修正后的文本。

After

优化后

name: polish-text-spelling description: Context-aware spelling and terminology correction for multilingual documents. tags: [writing, quality-control] version: 1.1.0

Skill: Spelling and terminology

Purpose: Identify and fix low-level spelling errors and terminology inconsistency without changing the author's intent or tone.

Behavior:

Detect language.

Build a term list if the text is long.

Distinguish "typos" from "intentional style."

Restrictions: Do not change proper nouns or specific abbreviations unless clearly wrong.

name: polish-text-spelling description: 针对多语言文档的上下文感知拼写与术语修正。 tags: [writing, quality-control] version: 1.1.0

Skill：拼写与术语修正

用途：识别并修复低级拼写错误和术语不一致问题，不改变作者的意图或语气。

行为：

检测语言类型。

若文本较长，构建术语列表。

区分“拼写错误”与“有意的风格表达”。

约束条件：不得修改专有名词或特定缩写，除非明显错误。

Example 2: Edge case — ambiguous draft

示例2：边缘案例——模糊草稿

Input: A Skill draft whose Purpose is "help users handle files," with no Use Cases or Restrictions.
Expected: Pin down intent (replace "handle" with concrete verbs: parse, convert, merge, etc.); add Use Cases and Restrictions (e.g. do not overwrite source; do not modify binaries); add at least one edge example (e.g. empty file, very large file, permission denied).

输入：一款Purpose为“帮助用户处理文件”的Skill草稿，无Use Cases和Restrictions章节。
预期输出：明确核心意图（将“处理”替换为具体动词：解析、转换、合并等）；添加Use Cases和Restrictions（例如：不得覆盖源文件；不得修改二进制文件）；至少添加一个边缘示例（例如：空文件、超大文件、权限不足）。

Appendix: Output contract

附录：输出契约

When this skill produces a refinement, the output MUST satisfy this contract so that Agents and downstream tools can consume it consistently:

Element	Requirement
Optimized SKILL	Full Markdown content (or path to file). MUST satisfy spec/skill.md: YAML front matter, Purpose, Use cases, Behavior, I/O, Restrictions, Self-Check, and at least one Example.
Diff summary	List of changes. Each entry MUST include Section (e.g. `Purpose` , `Behavior` , `metadata` ), Change (short description of what was changed), and Reason (why the change was made).
Version suggestion	SemVer string `major.minor.patch` . Optionally Rationale (e.g. `minor: added Restrictions` ).

当本Skill生成优化结果时，输出必须满足以下契约，以便Agent和下游工具能够一致地使用：

元素	要求
优化后的SKILL	完整的Markdown内容（或文件路径）。必须符合spec/skill.md规范：包含YAML前置元数据、Purpose、Use cases、Behavior、I/O、Restrictions、Self-Check，以及至少一个Example。
差异总结	变更列表。每个条目必须包含章节（例如 `Purpose` 、 `Behavior` 、 `metadata` ）、变更内容（简短描述变更点）和原因（说明变更的理由）。
版本建议	SemVer格式字符串 `major.minor.patch` 。可选择性添加理由（例如 `minor: 新增Restrictions章节` ）。