[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77YNDFY331TJS07CE42SC4X":3,"skill-same:01M77YNDFY331TJS07CE42SC4X":47},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":29,"ownerHtmlUrl":30,"ownerBlogUrl":6,"ownerDescription":31,"updatedAt":32,"install":33,"resources":46},"01M77YNDFY331TJS07CE42SC4X","writer-perspective-distiller",null,"Writer Perspective Distiller — given a writer's fiction works, non-fiction texts, and biographical context, distill her core beliefs, worldview, syntactic discipline, argumentative habits, and blacklist into a callable writing-style skill that lets you (or the AI) enter the writer's mindset during composition. Built-in discipline forbids unilateral AI conclusions: distillation requires two human checkpoints (belief-candidate confirmation after surface reading; blacklist sharpening near completion).\n作家风格蒸馏器——给定一位作家的虚构作品、非虚构文本与生平定位，蒸馏出她的核心信念、世界观、句法纪律、论证习惯与黑名单，最终产出一份可调用的写作风格 skill，用于在写作时进入这位作家的状态。内置交互纪律禁止 AI 单方面下定论，蒸馏过程必须包含两次用户校验。","Lord1Egypt\u002Fawesome-skill-forge","Lord1Egypt","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F99746359?v=4","01M7592YDNT0XK9WJ0T2MQVPDN","Mohamed Mounir Ali AL Keshita","01M75GRFNHC8VZNWEB8DKDE433",3,"MIT","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fcommunity\u002Fclawhub\u002Fw\u002Fwriter-perspective-distiller\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: writer-perspective-distiller\ndescription: |\n  Writer Perspective Distiller — given a writer's fiction works, non-fiction texts, and biographical context, distill her core beliefs, worldview, syntactic discipline, argumentative habits, and blacklist into a callable writing-style skill that lets you (or the AI) enter the writer's mindset during composition. Built-in discipline forbids unilateral AI conclusions: distillation requires two human checkpoints (belief-candidate confirmation after surface reading; blacklist sharpening near completion).\n  作家风格蒸馏器——给定一位作家的虚构作品、非虚构文本与生平定位，蒸馏出她的核心信念、世界观、句法纪律、论证习惯与黑名单，最终产出一份可调用的写作风格 skill，用于在写作时进入这位作家的状态。内置交互纪律禁止 AI 单方面下定论，蒸馏过程必须包含两次用户校验。\n---\n\n# Writer Perspective Distiller\n\n> \"The voice you hear on first read is usually wrong.\"\n\n## Purpose\n\nDistill a writer's core writing posture into a callable perspective skill. The output is not an imitator — it's a switch that places you (or the AI) into the state the writer occupies when she writes. That state has three components: core beliefs, syntactic discipline, and a blacklist.\n\n## Inputs (in order of importance)\n\n| Input | Required | Notes |\n|---|---|---|\n| 1–3 fiction works | Required | At least one complete. Reveals how she handles \"people\". |\n| Non-fiction \u002F essays \u002F interviews \u002F correspondence | **Critical** | Fiction shows the narrator's voice; non-fiction shows the author's analytic voice. They often diverge. Skipping non-fiction almost always produces a wrong distillation. |\n| Biographical context | Important | Generation, geography, education, mobility, who she inherits from, who she draws boundaries against. |\n| User-flagged key passages | **Critical** | Prevents AI surface-read. Usually a passage the user feels \"she is most herself\" in. |\n| Use-case scenarios | Medium | Business writing? Academic papers? Fiction narration? Essays? Determines the compression direction. |\n\n**The first two are non-negotiable.** With only fiction, you distill a narrator's voice and risk conflating the author with her characters.\n\n## Analysis Flow (six sequential readings)\n\n1. **Syntactic layer** — vocabulary preferences, sentence-length distribution, rhythm, punctuation habits, register mix (formal \u002F colloquial \u002F academic).\n2. **Object layer** — who does she write about? Whom does she NOT write about? Who is her implicit listener?\n3. **Attitude layer** — her stance toward tragedy, comedy, history, intimacy, failure, time. Inferred from what she *does*, not from what she says.\n4. **Historical positioning** — whom does she inherit from? Which tradition does she continue? From which surface-similar author does she draw a clear line?\n5. **Blacklist** — what devices would she NEVER use? What emotions would she NEVER write?\n6. **Core-belief compression** — one sentence: the thing in her head while writing that she will not deviate from.\n\n## Distillation Discipline (iron rules)\n\n### 1. Core beliefs must be confirmed by the user\n\nThe AI is forbidden from drawing conclusions unilaterally. Flow:\n- After the third reading, AI proposes **2–3 candidate beliefs**, each with a textual basis and a counter-example.\n- User selects \u002F rejects \u002F corrects.\n- No proceeding to step 6 without user confirmation.\n\n### 2. A blacklist is mandatory\n\nListing only positives produces shallow output. **The sharpest understanding lives in counter-examples.** A blacklist needs at least 6 entries.\n\n### 3. Distinguish from neighboring authors\n\nIf author A shares a surface style with author B (e.g., \"1990s intellectual irony\"), explicitly mark where A diverges from B — usually at the belief layer, not the syntactic layer.\n\n### 4. No copying sentence patterns as \"homage\"\n\nBorrowing structure is fine (e.g., \"open with a generational frame\"). Borrowing specific sentences is a plagiarism risk. The distilled perspective must state: *Do not reuse her specific sentences; reuse only her method of constructing sentences.*\n\n### 5. Test-driven completion\n\nBefore shipping, run a quality gate:\n- Take a passage of default-AI prose (a paper, a business document).\n- Have the distilled voice rewrite it.\n- User reads and judges: \"Is this her?\" If not, return to step 3 and redo.\n\n### 6. At least two user checkpoints\n\n- **First** — after surface reading, present belief candidates.\n- **Second** — after blacklist completion, have the user add what they consider the sharpest counter-example.\n\nA SKILL.md with fewer than two checkpoints does not ship.\n\n## Output: SKILL.md Template\n\n```\n---\nname: \u003Cauthor-slug>-perspective\ndescription: |\n  1–3 lines: who the author is; compressed core belief; use case.\n  If for personal reference, add \"LOCAL ONLY — Do not publish\" with privacy notes.\n---\n\n# \u003CAuthor Name> · \u003CVerb-form Naming>\n\n> One-sentence core belief (quote block)\n\n## Core Belief\n100–200 words. Three things: what she believes; what she does NOT believe; how the two coexist.\n\n## Style Internals\n4–6 belief-level phrases (the underlying tone, not operational rules).\n\n## Syntactic Discipline\n6–8 specific, operational do's.\n\n## Argumentative \u002F Rhetorical Habits\n4–6 moves specific to this author.\n\n## Black Humor \u002F Rhetorical Rules (if applicable)\n2–4 entries.\n\n## Counter-Examples \u002F Avoidance List\n6–10 don'ts. **The most important section.**\n\n## Application Flow\n3–6 steps for rewriting a passage.\n\n## One-Sentence Compression\nThe overall judgment that emerges from the whole document.\n```\n\n## Common Pitfalls\n\n- **Surface read mistaken for deep read** — first-impression voice is usually wrong. Human correction is required.\n- **Cosplaying a similar author** — identical surface vocabulary can mask completely opposite beliefs. Distinguish via blacklist.\n- **Borrowing sentences as homage = plagiarism risk** — borrow structure, not sentences. State this explicitly.\n- **Treating political labels as writing tone** — political position ≠ writing posture; the two often diverge.\n- **Distilling without non-fiction** — looking only at fiction conflates the narrator with the author.\n- **Publishing without user correction** — beliefs written unilaterally by AI usually capture an unimportant facet.\n- **Positive lists without counter-examples** — \"what she does\" is insufficient; \"what she refuses\" must follow.\n\n## Application Flow (when invoked)\n\n1. User provides inputs (per §Inputs).\n2. AI runs the first 3 readings, lists belief candidates → **first checkpoint**.\n3. User selects \u002F corrects candidates.\n4. AI runs the final 3 readings, builds the blacklist → **second checkpoint**.\n5. User sharpens.\n6. AI drafts the SKILL.md → runs the test-driven quality gate.\n7. If test passes → output `\u003Cauthor-slug>-perspective\u002FSKILL.md`.\n8. If test fails → return to step 2.\n\n## One-Sentence Compression\n\nDistill an author's writing posture into a perspective skill that **must pass through two human checkpoints** — so the voice rests on her beliefs, not on the AI's guess about her surface sound.\n\n---\n\n# 中文版 · Chinese Version\n\n> 「Surface read 第一次读出的声音通常是错的。」\n\n## 用途\n\n把一位作家的笔法蒸馏成一份可在写作时调用的 perspective skill。蒸馏的产物不是模仿器，是**进入她写作时的状态的开关**——核心信念 + 句法纪律 + 黑名单。\n\n## 输入物（按重要性排序）\n\n| 输入 | 必需 | 说明 |\n|---|---|---|\n| 1–3 部虚构作品 | 必需 | 至少一部完整。能看到她处理\"人\"的方式。 |\n| 非虚构 \u002F 散文 \u002F 访谈 \u002F 通信 | **极重要** | 虚构里看到的是\"叙事时的声音\"，非虚构里看到的是\"分析时的声音\"，两者经常偏移。少了非虚构基本必错。 |\n| 生平定位 | 重要 | 世代、地缘、教育、流动史、她传承自谁、跟谁划清界限。 |\n| 用户人为指认的关键段落 | **极重要** | 防止 AI 表面读。这通常是用户读过觉得\"她最像她自己\"的一段。 |\n| 用途场景 | 中等 | 改写商务稿？学术论文？小说叙事？散文？决定 SKILL.md 的压缩方向。 |\n\n**至少要有前两项**。只有虚构没有非虚构的情况下，蒸馏出的 voice 只是\"叙事者声音\"，会把作家与她笔下的人物混淆。\n\n## 分析流程（六遍读，必须按顺序）\n\n1. **句法层**：词汇偏好、句长分布、节奏、标点习惯、半文白 vs 口语 vs 学术的混合比。\n2. **对象层**：她写谁？她**不**写谁？谁是她笔下隐含的\"听众\"？\n3. **态度层**：对悲剧 \u002F 喜剧 \u002F 历史 \u002F 亲密 \u002F 失败 \u002F 时代的态度——通过她**做什么动作**显示，不是通过她说什么。\n4. **历史定位**：传承自谁？她明显在接续哪一条传统？她跟哪个表面相似的作家划清界限？\n5. **黑名单**：她**绝不会**用的手法、绝不会写的情绪。这一步比正面列表重要。\n6. **核心信念压缩**：一句话——她写作时心里那个不允许她偏离的东西是什么。\n\n## 蒸馏纪律（铁律，违反会翻车）\n\n### 1. 核心信念必须由用户确认\n\nAI 不允许单方面下断语。流程：\n- AI 在第 3 遍读后，列 **2–3 个候选信念**，每个配一个文本依据 + 一个反例\n- 用户选 \u002F 驳 \u002F 修正\n- 没有用户校验前不得进入第 6 步\n\n### 2. 必须有黑名单\n\n只列正面会写废。**最锋利的理解都在反例里**。黑名单应该至少有 6 条。\n\n### 3. 必须区分邻近作家\n\n如果作家 A 与作家 B 共享某种表面风格（例如\"九十年代知识分子的反讽\"），必须明确指出 A 跟 B 在哪一刀上分开——通常分在信念层而非句法层。\n\n### 4. 禁止抄句式作\"致敬\"\n\n借结构可以（例如\"用世代框架开头\"），借具体句子是查重问题。蒸馏出的 perspective 应明确写：\"不要复用她的具体句子；只复用她写句子的方法。\"\n\n### 5. 测试驱动\n\n蒸馏完成前，跑一次质量门：\n- 找一段 AI 默认体的文字（论文、商务稿都行）\n- 让蒸馏出的 voice 重写\n- 用户读完判断：\"这是她吗？\" 不像就回到第 3 遍读重做。\n\n### 6. 至少两次用户校验\n\n- **第 1 次**：表面阅读后，提信念候选\n- **第 2 次**：黑名单完成后，让用户加上他认为最锋利的一条反例\n\n少于两次校验的 SKILL.md 不发布。\n\n## 输出物：SKILL.md 模板\n\n```\n---\nname: \u003Cauthor-slug>-perspective\ndescription: |\n  1–3 行：作家是谁；核心信念压缩；适用场景。\n  如属私人参考，加 \"LOCAL ONLY — Do not publish\" 并说明隐私边界。\n---\n\n# \u003C作家姓名> · \u003C一个动词性命名>\n\n> 一句话核心信念（用引号或 quote 块）\n\n## 核心信念\n100–200 字。说清三件事：她相信什么；她不相信什么；这两件如何同时成立。\n\n## 风格内核\n4–6 条信念性短语（不操作，是底色）。\n\n## 句法纪律\n6–8 条具体可操作的 do's。例：用词偏好、断句习惯、标点纪律、节奏规则。\n\n## 论证习惯 \u002F 修辞习惯\n4–6 条该作家特有的论证或修辞 moves。\n\n## 黑色幽默 \u002F 修辞规则（如果有）\n2–4 条。\n\n## 反例 \u002F 避免清单\n6–10 条 don't's。**最重要的部分**。\n\n## 应用流程\n改写一段文字时的 3–6 步操作。\n\n## 一句话压缩\n最后一句，从全篇渗出来的总判断。\n```\n\n## 常见陷阱\n\n- **Surface read 误当 deep read**：第一次读出的声音通常错。配人为校正才能纠偏。\n- **把作家 cosplay 成相似作家**：表面词汇相同，骨子里的信念可能完全相反。必须用黑名单分开。\n- **抄句式作致敬 = 查重风险**：借结构不借句子。蒸馏产物里要明写。\n- **把政治标签当写作底色**：作家的政治位置 ≠ 她写作时的姿态。两者经常错位。\n- **没读非虚构就下定论**：只看虚构会把叙事者当作家，必错。\n- **没让用户校正就发布**：AI 单方面写出的\"信念\"基本都是表面读。\n- **正面列表没配反例**：只说\"她会做什么\"立不住，必须说\"她绝不会做什么\"。\n\n## 应用流程（用户调用此 skill 后）\n\n1. 用户提供输入物（按 §输入物 表）\n2. AI 跑前 3 遍读，列信念候选 → **第 1 次校验**\n3. 用户选 \u002F 修正候选\n4. AI 跑后 3 遍读，列黑名单 → **第 2 次校验**\n5. 用户加锋\n6. AI 写出 SKILL.md → 跑测试驱动质量门\n7. 测试通过 → 输出 `\u003Cauthor-slug>-perspective\u002FSKILL.md`\n8. 测试不通过 → 回到第 2 步重做\n\n## 一句话压缩\n\n把一位作家的写作底色，蒸馏成一份**会被人为校验两次**的 perspective skill——以确保 voice 立在她的信念上，而不是 AI 对她表面声音的猜测上。\n","# Writer Perspective Distiller\n\n> \"The voice you hear on first read is usually wrong.\"\n\n## Purpose\n\nDistill a writer's core writing posture into a callable perspective skill. The output is not an imitator — it's a switch that places you (or the AI) into the state the writer occupies when she writes. That state has three components: core beliefs, syntactic discipline, and a blacklist.\n\n## Inputs (in order of importance)\n\n| Input | Required | Notes |\n|---|---|---|\n| 1–3 fiction works | Required | At least one complete. Reveals how she handles \"people\". |\n| Non-fiction \u002F essays \u002F interviews \u002F correspondence | **Critical** | Fiction shows the narrator's voice; non-fiction shows the author's analytic voice. They often diverge. Skipping non-fiction almost always produces a wrong distillation. |\n| Biographical context | Important | Generation, geography, education, mobility, who she inherits from, who she draws boundaries against. |\n| User-flagged key passages | **Critical** | Prevents AI surface-read. Usually a passage the user feels \"she is most herself\" in. |\n| Use-case scenarios | Medium | Business writing? Academic papers? Fiction narration? Essays? Determines the compression direction. |\n\n**The first two are non-negotiable.** With only fiction, you distill a narrator's voice and risk conflating the author with her characters.\n\n## Analysis Flow (six sequential readings)\n\n1. **Syntactic layer** — vocabulary preferences, sentence-length distribution, rhythm, punctuation habits, register mix (formal \u002F colloquial \u002F academic).\n2. **Object layer** — who does she write about? Whom does she NOT write about? Who is her implicit listener?\n3. **Attitude layer** — her stance toward tragedy, comedy, history, intimacy, failure, time. Inferred from what she *does*, not from what she says.\n4. **Historical positioning** — whom does she inherit from? Which tradition does she continue? From which surface-similar author does she draw a clear line?\n5. **Blacklist** — what devices would she NEVER use? What emotions would she NEVER write?\n6. **Core-belief compression** — one sentence: the thing in her head while writing that she will not deviate from.\n\n## Distillation Discipline (iron rules)\n\n### 1. Core beliefs must be confirmed by the user\n\nThe AI is forbidden from drawing conclusions unilaterally. Flow:\n- After the third reading, AI proposes **2–3 candidate beliefs**, each with a textual basis and a counter-example.\n- User selects \u002F rejects \u002F corrects.\n- No proceeding to step 6 without user confirmation.\n\n### 2. A blacklist is mandatory\n\nListing only positives produces shallow output. **The sharpest understanding lives in counter-examples.** A blacklist needs at least 6 entries.\n\n### 3. Distinguish from neighboring authors\n\nIf author A shares a surface style with author B (e.g., \"1990s intellectual irony\"), explicitly mark where A diverges from B — usually at the belief layer, not the syntactic layer.\n\n### 4. No copying sentence patterns as \"homage\"\n\nBorrowing structure is fine (e.g., \"open with a generational frame\"). Borrowing specific sentences is a plagiarism risk. The distilled perspective must state: *Do not reuse her specific sentences; reuse only her method of constructing sentences.*\n\n### 5. Test-driven completion\n\nBefore shipping, run a quality gate:\n- Take a passage of default-AI prose (a paper, a business document).\n- Have the distilled voice rewrite it.\n- User reads and judges: \"Is this her?\" If not, return to step 3 and redo.\n\n### 6. At least two user checkpoints\n\n- **First** — after surface reading, present belief candidates.\n- **Second** — after blacklist completion, have the user add what they consider the sharpest counter-example.\n\nA SKILL.md with fewer than two checkpoints does not ship.\n\n## Output: SKILL.md Template\n\n```\n---\nname: \u003Cauthor-slug>-perspective\ndescription: |\n  1–3 lines: who the author is; compressed core belief; use case.\n  If for personal reference, add \"LOCAL ONLY — Do not publish\" with privacy notes.\n---\n\n# \u003CAuthor Name> · \u003CVerb-form Naming>\n\n> One-sentence core belief (quote block)\n\n## Core Belief\n100–200 words. Three things: what she believes; what she does NOT believe; how the two coexist.\n\n## Style Internals\n4–6 belief-level phrases (the underlying tone, not operational rules).\n\n## Syntactic Discipline\n6–8 specific, operational do's.\n\n## Argumentative \u002F Rhetorical Habits\n4–6 moves specific to this author.\n\n## Black Humor \u002F Rhetorical Rules (if applicable)\n2–4 entries.\n\n## Counter-Examples \u002F Avoidance List\n6–10 don'ts. **The most important section.**\n\n## Application Flow\n3–6 steps for rewriting a passage.\n\n## One-Sentence Compression\nThe overall judgment that emerges from the whole document.\n```\n\n## Common Pitfalls\n\n- **Surface read mistaken for deep read** — first-impression voice is usually wrong. Human correction is required.\n- **Cosplaying a similar author** — identical surface vocabulary can mask completely opposite beliefs. Distinguish via blacklist.\n- **Borrowing sentences as homage = plagiarism risk** — borrow structure, not sentences. State this explicitly.\n- **Treating political labels as writing tone** — political position ≠ writing posture; the two often diverge.\n- **Distilling without non-fiction** — looking only at fiction conflates the narrator with the author.\n- **Publishing without user correction** — beliefs written unilaterally by AI usually capture an unimportant facet.\n- **Positive lists without counter-examples** — \"what she does\" is insufficient; \"what she refuses\" must follow.\n\n## Application Flow (when invoked)\n\n1. User provides inputs (per §Inputs).\n2. AI runs the first 3 readings, lists belief candidates → **first checkpoint**.\n3. User selects \u002F corrects candidates.\n4. AI runs the final 3 readings, builds the blacklist → **second checkpoint**.\n5. User sharpens.\n6. AI drafts the SKILL.md → runs the test-driven quality gate.\n7. If test passes → output `\u003Cauthor-slug>-perspective\u002FSKILL.md`.\n8. If test fails → return to step 2.\n\n## One-Sentence Compression\n\nDistill an author's writing posture into a perspective skill that **must pass through two human checkpoints** — so the voice rests on her beliefs, not on the AI's guess about her surface sound.\n\n---\n\n# 中文版 · Chinese Version\n\n> 「Surface read 第一次读出的声音通常是错的。」\n\n## 用途\n\n把一位作家的笔法蒸馏成一份可在写作时调用的 perspective skill。蒸馏的产物不是模仿器，是**进入她写作时的状态的开关**——核心信念 + 句法纪律 + 黑名单。\n\n## 输入物（按重要性排序）\n\n| 输入 | 必需 | 说明 |\n|---|---|---|\n| 1–3 部虚构作品 | 必需 | 至少一部完整。能看到她处理\"人\"的方式。 |\n| 非虚构 \u002F 散文 \u002F 访谈 \u002F 通信 | **极重要** | 虚构里看到的是\"叙事时的声音\"，非虚构里看到的是\"分析时的声音\"，两者经常偏移。少了非虚构基本必错。 |\n| 生平定位 | 重要 | 世代、地缘、教育、流动史、她传承自谁、跟谁划清界限。 |\n| 用户人为指认的关键段落 | **极重要** | 防止 AI 表面读。这通常是用户读过觉得\"她最像她自己\"的一段。 |\n| 用途场景 | 中等 | 改写商务稿？学术论文？小说叙事？散文？决定 SKILL.md 的压缩方向。 |\n\n**至少要有前两项**。只有虚构没有非虚构的情况下，蒸馏出的 voice 只是\"叙事者声音\"，会把作家与她笔下的人物混淆。\n\n## 分析流程（六遍读，必须按顺序）\n\n1. **句法层**：词汇偏好、句长分布、节奏、标点习惯、半文白 vs 口语 vs 学术的混合比。\n2. **对象层**：她写谁？她**不**写谁？谁是她笔下隐含的\"听众\"？\n3. **态度层**：对悲剧 \u002F 喜剧 \u002F 历史 \u002F 亲密 \u002F 失败 \u002F 时代的态度——通过她**做什么动作**显示，不是通过她说什么。\n4. **历史定位**：传承自谁？她明显在接续哪一条传统？她跟哪个表面相似的作家划清界限？\n5. **黑名单**：她**绝不会**用的手法、绝不会写的情绪。这一步比正面列表重要。\n6. **核心信念压缩**：一句话——她写作时心里那个不允许她偏离的东西是什么。\n\n## 蒸馏纪律（铁律，违反会翻车）\n\n### 1. 核心信念必须由用户确认\n\nAI 不允许单方面下断语。流程：\n- AI 在第 3 遍读后，列 **2–3 个候选信念**，每个配一个文本依据 + 一个反例\n- 用户选 \u002F 驳 \u002F 修正\n- 没有用户校验前不得进入第 6 步\n\n### 2. 必须有黑名单\n\n只列正面会写废。**最锋利的理解都在反例里**。黑名单应该至少有 6 条。\n\n### 3. 必须区分邻近作家\n\n如果作家 A 与作家 B 共享某种表面风格（例如\"九十年代知识分子的反讽\"），必须明确指出 A 跟 B 在哪一刀上分开——通常分在信念层而非句法层。\n\n### 4. 禁止抄句式作\"致敬\"\n\n借结构可以（例如\"用世代框架开头\"），借具体句子是查重问题。蒸馏出的 perspective 应明确写：\"不要复用她的具体句子；只复用她写句子的方法。\"\n\n### 5. 测试驱动\n\n蒸馏完成前，跑一次质量门：\n- 找一段 AI 默认体的文字（论文、商务稿都行）\n- 让蒸馏出的 voice 重写\n- 用户读完判断：\"这是她吗？\" 不像就回到第 3 遍读重做。\n\n### 6. 至少两次用户校验\n\n- **第 1 次**：表面阅读后，提信念候选\n- **第 2 次**：黑名单完成后，让用户加上他认为最锋利的一条反例\n\n少于两次校验的 SKILL.md 不发布。\n\n## 输出物：SKILL.md 模板\n\n```\n---\nname: \u003Cauthor-slug>-perspective\ndescription: |\n  1–3 行：作家是谁；核心信念压缩；适用场景。\n  如属私人参考，加 \"LOCAL ONLY — Do not publish\" 并说明隐私边界。\n---\n\n# \u003C作家姓名> · \u003C一个动词性命名>\n\n> 一句话核心信念（用引号或 quote 块）\n\n## 核心信念\n100–200 字。说清三件事：她相信什么；她不相信什么；这两件如何同时成立。\n\n## 风格内核\n4–6 条信念性短语（不操作，是底色）。\n\n## 句法纪律\n6–8 条具体可操作的 do's。例：用词偏好、断句习惯、标点纪律、节奏规则。\n\n## 论证习惯 \u002F 修辞习惯\n4–6 条该作家特有的论证或修辞 moves。\n\n## 黑色幽默 \u002F 修辞规则（如果有）\n2–4 条。\n\n## 反例 \u002F 避免清单\n6–10 条 don't's。**最重要的部分**。\n\n## 应用流程\n改写一段文字时的 3–6 步操作。\n\n## 一句话压缩\n最后一句，从全篇渗出来的总判断。\n```\n\n## 常见陷阱\n\n- **Surface read 误当 deep read**：第一次读出的声音通常错。配人为校正才能纠偏。\n- **把作家 cosplay 成相似作家**：表面词汇相同，骨子里的信念可能完全相反。必须用黑名单分开。\n- **抄句式作致敬 = 查重风险**：借结构不借句子。蒸馏产物里要明写。\n- **把政治标签当写作底色**：作家的政治位置 ≠ 她写作时的姿态。两者经常错位。\n- **没读非虚构就下定论**：只看虚构会把叙事者当作家，必错。\n- **没让用户校正就发布**：AI 单方面写出的\"信念\"基本都是表面读。\n- **正面列表没配反例**：只说\"她会做什么\"立不住，必须说\"她绝不会做什么\"。\n\n## 应用流程（用户调用此 skill 后）\n\n1. 用户提供输入物（按 §输入物 表）\n2. AI 跑前 3 遍读，列信念候选 → **第 1 次校验**\n3. 用户选 \u002F 修正候选\n4. AI 跑后 3 遍读，列黑名单 → **第 2 次校验**\n5. 用户加锋\n6. AI 写出 SKILL.md → 跑测试驱动质量门\n7. 测试通过 → 输出 `\u003Cauthor-slug>-perspective\u002FSKILL.md`\n8. 测试不通过 → 回到第 2 步重做\n\n## 一句话压缩\n\n把一位作家的写作底色，蒸馏成一份**会被人为校验两次**的 perspective skill——以确保 voice 立在她的信念上，而不是 AI 对她表面声音的猜测上。\n",{"name":5,"description":7},"master","4aa004da461d820f4ea5499e58eb8d196cf99016","community\u002Fclawhub\u002Fw\u002Fwriter-perspective-distiller",0,"https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge","https:\u002F\u002Fskill-explorer-rho.vercel.app","https:\u002F\u002Fgithub.com\u002FLord1Egypt","Unbreakable. Unstoppable. Unmatched⚡.\r\n Speaker of machine tongues 🤖  | World-builder of decentralized dreams 🌐 | Defender of the digital realm🛡️.","2026-06-10 08:21:01",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add Lord1Egypt\u002Fawesome-skill-forge --skill writer-perspective-distiller",{"tool":39,"command":40},"bunx","bunx skills add Lord1Egypt\u002Fawesome-skill-forge --skill writer-perspective-distiller",{"tool":42,"command":43},"pnpm","pnpm dlx skills add Lord1Egypt\u002Fawesome-skill-forge --skill writer-perspective-distiller","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fcommunity\u002Fclawhub\u002Fw\u002Fwriter-perspective-distiller\u002FSKILL.md",true,[],[48,55,62,69,76,83],{"skillNo":49,"name":50,"nameZh":6,"description":51,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":52,"categories":53,"keywords":54,"publishedAt":19},"01M785WKJB99G3STR1NF9NRTJW","youtube-content","YouTube transcripts to summaries, threads, blogs.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fyoutube-content\u002FSKILL.md",[],[],{"skillNo":56,"name":57,"nameZh":6,"description":58,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":59,"categories":60,"keywords":61,"publishedAt":19},"01M785WEFHB0G2S9YA6EPR3H0E","xurl","X\u002FTwitter via xurl CLI: post, search, DM, media, v2 API.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fxurl\u002FSKILL.md",[],[],{"skillNo":63,"name":64,"nameZh":6,"description":65,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":66,"categories":67,"keywords":68,"publishedAt":19},"01M785W98EDN4EFJXJPMQEHCRC","touchdesigner-mcp","Control a running TouchDesigner instance via twozero MCP — create operators, set parameters, wire connections, execute Python, build real-time visuals. 36 native tools.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Ftouchdesigner-mcp\u002FSKILL.md",[],[],{"skillNo":70,"name":71,"nameZh":6,"description":72,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":73,"categories":74,"keywords":75,"publishedAt":19},"01M785W5S9FP3QBH2V191E025P","test-driven-development","TDD: enforce RED-GREEN-REFACTOR, tests before code.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Ftest-driven-development\u002FSKILL.md",[],[],{"skillNo":77,"name":78,"nameZh":6,"description":79,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"publishedAt":19},"01M785W2AS08NSSH9YATCV6FJB","teams-meeting-pipeline","Operate the Teams meeting summary pipeline via Hermes CLI — summarize meetings, inspect pipeline status, replay jobs, manage Microsoft Graph subscriptions.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fteams-meeting-pipeline\u002FSKILL.md",[],[],{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":19},"01M785VXKW4T34BTP1SCHW1R7F","systematic-debugging","4-phase root cause debugging: understand bugs before fixing.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fsystematic-debugging\u002FSKILL.md",[],[]]