ljg-analogy
Finds apt analogies and relational metaphors for a given scenario or differences between two objects based on Structure-Mapping Theory, preserving core relationships, explanatory increments, and mapping boundaries. USE WHEN the user asks to "find an analogy", "make a comparison", "come up with a clever metaphor", "clarify the difference between the two using an analogy", "understand it from a different perspective", or calls ljg-analogy. NOT FOR merely explaining existing metaphors, pure rhetorical polishing, theoretical definitions, or treating analogies as factual proof.
1How It Works & Mechanism
Grounded in Gentner's (1983) Structure-Mapping Theory: mapping prioritizes preserving relations between objects over superficial attributes, favoring systems of relations organized by higher-order relations.
对象属性与对象间关系区别处理；映射优先保留关系，并选择关系系统
Fix terminology direction: the scenario the user wants to understand is the target domain, and the scenario borrowed to aid understanding is the base domain; search direction is opposite to knowledge transfer direction.
术语方向固定：**用户想理解的场景是目标域；借来帮助理解的场景是基域。**
Route by request type: four distinct handling paths—finding an analogy for a single scenario, comparing two objects, adapting to audience/style, and comparing multiple candidates.
| 给一个场景，寻找类比或隐喻 | 按下文的成品标准交付一个主类比 |
Evaluate input sufficiency first: when only a single word is provided without context, ask a question that can actively shift the analogy selection.
只有一个词而无可辨认情境时，问一个能改变类比选择的问题
Filter out generic commonalities: overly broad claims such as "both are very complex" or "both are networks" cannot independently support an analogy.
「都很复杂」「都是网络」「都需要平衡」适用面太宽，不能单独支撑类比。
The primary criterion for a finished analogy is task alignment: the mapping must preserve the specific relationship the reader needs to understand this time.
**任务吻合。** 映射保留的是读者这次需要理解的关系。
Perform a discriminative power check when comparing two objects: if swapping the base vehicles still holds without altering the load-bearing relations, it indicates only shared attributes were captured.
交换喻体以后，如果不用改变承重关系仍能轻易圆回来，它们可能只命中了共同属性。
Deliver directly in conversation by default, explaining a single primary analogy thoroughly in a three-stage structure without automatically creating notes, images, or files.
默认直接在对话交付，不自动创建笔记、图片或文件。
Maintain honest theoretical boundaries: only Gentner's mapping and systematicity principles are borrowed; understanding tasks, candidate elimination, and boundaries are novel authoring criteria.
本技能借用 Gentner（1983）的关系映射与系统性原则。理解任务、双对象对照、候选淘汰、受众熟悉度、认知增量与具体边界，是为创作任务增加的设计标准，不宣称出自论文或已被论文验证。
EvalCases.json serves only as a behavioral acceptance reference, defining inputs and acceptance criteria rather than prescribing a single correct analogy.
这些用例只定义输入和验收依据，不规定唯一的正确比方。
2Core Capabilities
4Recommended Governance & Evidence
- SKILL.md: Pure prompt instructions and rule text, without script execution or external side effects
5Second-Pass Verification
- [ok] 功能声明与实现一致 — frontmatter 声明『基于结构映射理论…寻找精妙类比与关系隐喻』，正文给出目标域/基域定义、成品标准与 Theory.md 理论依据，未发现夸大。
- [ok] 外部依赖核验：无运行时依赖 — 目录四个文件均为文本（md/json/yaml）；风险扫描四类（执行外部程序、网络请求、凭证、写盘）全部无命中。
- [ok] 安全结论反例检查 — 逐行重核无 curl/wget/fetch、child_process 或 fs 写入/读取凭据；无脚本文件，故 scripts_executed 为空。
- [ok] 引用可定位性 — 所有 quote 均可在 SKILL.md／Theory.md／EvalCases.json 原文逐字匹配，长度均大于 12 字符。
- [ok] 判级口径复核 — 按判级六档，纯指令无脚本无外发，落 green；Theory.md 的外链属来源标注，不改档位。
- [ok] 理论归属表述核对 — SKILL.md 明写新增标准『不宣称出自论文或已被论文验证』，分析据此未把创作规则写成论文结论。
- [ok] 用例集性质核对 — EvalCases.json purpose 声明不按固定措辞或唯一基域打分，分析未把单一基域当作正确解。
6Verdict & Scenarios
f6af0a46838494b1…