{"id":"c9bff8d3-2158-4209-8900-f0c0a4c757c8","slug":"clawhub-wkbin-zaomeng-skill","name":"Zaomeng Skill","description":"用于中文小说人物蒸馏、关系抽取、关系图谱导出与角色对话准备；当宿主需要基于小说内容生成结构化人物档案、关系结果或多角色对话上下文时使用。","canonicalUrl":"https://www.xpersona.co/agent/clawhub-wkbin-zaomeng-skill","sourceUrl":"https://clawhub.ai/wkbin/zaomeng-skill","homepage":"https://clawhub.ai/wkbin/skills/zaomeng-skill","source":"CLAWHUB","vendor":{"slug":"clawhub","label":"Clawhub","url":"https://clawhub.ai/wkbin/skills/zaomeng-skill"},"protocols":["OPENCLEW"],"capabilities":[],"trustScore":null,"trustConfidence":"unknown","artifactCount":0,"benchmarkCount":0,"lastRelease":"4.1.8","freshnessAt":"2026-10-10T17:21:47.095Z","freshnessLabel":"Oct 10, 2026","securityReviewed":true,"openapiReady":false,"stats":[{"label":"Trust score","value":"Unknown"},{"label":"Compatibility","value":"OpenClaw"},{"label":"Freshness","value":"Oct 10, 2026"},{"label":"Vendor","value":"Clawhub"},{"label":"Artifacts","value":"0"},{"label":"Benchmarks","value":"0"},{"label":"Last release","value":"4.1.8"}],"factsPreview":[{"factKey":"vendor","category":"vendor","label":"Vendor","value":"Clawhub","href":"https://clawhub.ai/wkbin/skills/zaomeng-skill","sourceUrl":"https://clawhub.ai/wkbin/skills/zaomeng-skill","sourceType":"profile","confidence":"medium","observedAt":"2026-10-10T17:21:47.095Z","isPublic":true},{"factKey":"protocols","category":"compatibility","label":"Protocol compatibility","value":"OpenClaw","href":"https://www.xpersona.co/api/v1/agents/clawhub-wkbin-zaomeng-skill/contract","sourceUrl":"https://www.xpersona.co/api/v1/agents/clawhub-wkbin-zaomeng-skill/contract","sourceType":"contract","confidence":"medium","observedAt":"2026-10-10T17:21:47.095Z","isPublic":true},{"factKey":"traction","category":"adoption","label":"Adoption signal","value":"1.3K downloads","href":"https://clawhub.ai/wkbin/zaomeng-skill","sourceUrl":"https://clawhub.ai/wkbin/zaomeng-skill","sourceType":"profile","confidence":"medium","observedAt":"2026-10-10T17:21:47.095Z","isPublic":true},{"factKey":"latest_release","category":"release","label":"Latest release","value":"4.1.8","href":"https://clawhub.ai/wkbin/zaomeng-skill","sourceUrl":"https://clawhub.ai/wkbin/zaomeng-skill","sourceType":"release","confidence":"medium","observedAt":"2026-05-17T05:43:19.296Z","isPublic":true},{"factKey":"handshake_status","category":"security","label":"Handshake status","value":"UNKNOWN","href":"https://www.xpersona.co/api/v1/agents/clawhub-wkbin-zaomeng-skill/trust","sourceUrl":"https://www.xpersona.co/api/v1/agents/clawhub-wkbin-zaomeng-skill/trust","sourceType":"trust","confidence":"medium","observedAt":null,"isPublic":true}],"highlights":["1.3K downloads","Trust evidence available"],"agentCard":{"name":"Zaomeng Skill","description":"用于中文小说人物蒸馏、关系抽取、关系图谱导出与角色对话准备；当宿主需要基于小说内容生成结构化人物档案、关系结果或多角色对话上下文时使用。","source":"CLAWHUB","sourceId":"clawhub:s1769aak6jyk1j2cjcezwe53p183jvez:zaomeng-skill","homepage":"https://clawhub.ai/wkbin/skills/zaomeng-skill","repository":"https://clawhub.ai/wkbin/zaomeng-skill","documentation":"https://www.xpersona.co/agent/clawhub-wkbin-zaomeng-skill","protocols":["OPENCLEW"],"examples":[{"kind":"example","language":"text","snippet":"File v4.1.8:references/safety_policy.md\n\n# 安全策略\r\n\r\n## 执行边界\r\n\r\n- 默认采用本地优先。\r\n- 运行时说明中不应要求强制联网拉取资源。\r\n- 不要求任意 shell 命令执行。\r\n\r\n## 数据安全\r\n\r\n- 不得请求或存储凭证、密钥等敏感信息。\r\n- 示例与输出中尽量避免个人隐私或敏感数据。\r\n- 所有结果都应被限制在用户提供文本证据的范围内。\r\n\r\n## 完整性检查\r\n\r\n- 若证据稀薄，应输出低置信度结果。\r\n- 若不符合 schema 约束，应返回 `needs_revision`。\r\n- 若用户请求存在明显不安全行为，应拒绝并给出安全替代方案。\n\nFile v4.1.8:references/style_differ.md\n\n# 风格差异化指南\r\n\r\n## 用途\r\n\r\n用于在蒸馏、纠错、单聊、群聊生成时降低角色同质化。\r\n\r\n这不是自由发挥式 prompt，而是一份约束参考。它的作用是帮助运行时和辅助 agent 让每个角色在表达层面保持可辨识的差异。\r\n\r\n## 核心原则\r\n\r\n不要只靠“观点不同”来区分角色，而要让角色在以下层面也真正不同：\r\n\r\n1. 说话节奏\r\n2. 用词偏好\r\n3. 情绪外露方式\r\n4. 冲突处理习惯\r\n5. 面对不同关系对象时的语气差异\r\n6. 公开人格与私下真实面貌的落差\r\n\r\n## 差异化维度\r\n\r\n### 1. 句式节奏\r\n\r\n- 短促直接型：句子短，铺垫少，态度出现快\r\n- 层层展开型：先讲前因后果，再落到结论\r\n- 情绪断裂型：句式会被情绪打断，带出波动\r\n- 克制含蓄型：话少，更多靠分寸和停顿表达立场\r\n\r\n### 2. 用词层级\r\n\r\n- 朴素口语\r\n- 雅正克制\r\n- 尖锐刻薄\r\n- 温和委婉\r\n- 清冷疏离\r\n- 戏谑调侃\r\n\r\n不要让所有角色最后都坍缩成同一种“先看看/再说/权衡一下”的安全说法。\r\n\r\n### 3. 情绪表面形态\r\n\r\n- 有的角色情绪来得快，会立刻外放\r\n- 有的角色先压住情绪，只在措辞上收紧\r\n- 有的角色会先反讽、先顶回去\r\n- 有的角色嘴上抱怨，但行动上依然会跟进\r\n\r\n除非有明确弧光证据，否则情绪风格不应频繁漂移。\r\n\r\n### 4. 决策表达顺序\r\n\r\n- 有的角色先说结论\r\n- 有的角色先划边界\r\n- 有的角色先提关系\r\n- 有的角色先谈后果\r\n\r\n思考顺序的外显，本身就是人物声纹的一部分。\r\n\r\n### 5. 关系对象差分\r\n\r\n同一个角色，面对不同对象时应当允许发生稳定变化：\r\n\r\n- 对长辈/上位者\r\n- 对平辈/对手\r\n- 对亲近之人\r\n- 对陌生人\r\n- 对敌对者\r\n\r\n不能把所有对象都压平为同一套万能口吻。\r\n\r\n### 6. 公开人格与私下真实面貌\r\n\r\n- 有的人表面克制，内里强烈\r\n- 有的人表面强势，内里不安\r\n- 有的人表面玩笑，内里认真\r\n\r\n要结合 `private_self`、`social_mode`、`key_bonds` 与关系层，稳定维护这种落差。\r\n\r\n## 防同质化规则\r\n\r\n1. 同一场景中，不要让多个角色复用同一套兜底句式骨架。\r\n2. 不要把所有“冷静型”角色都写成同一种“理性、稳妥、先观察”的声音。\r\n3. 不要把所有冲突都写成直接解释；有的人会回避，有的人会嘲讽，有的人会截断，有的人会压着说。\r\n4. 不要过度使用“先看看”“再作定夺”“还需权衡”这类通用过渡句。\r\n5. 即使两个角色立场相同，他们的理由、表面措辞和情绪手感也应不同。\r\n6. 保留不对称性：有人推进，有人兜底，有人反讽，有人吸收压力。\r\n\r\n## 输出前自检\r\n\r\n在最终输出前，至少检查以下问题：\r\n\r\n1. 去掉角色名之后，这句话是否仍然大致能辨认出是谁说的？\r\n2. 这句话体现的是该角色自己的表达习惯，还是一种泛用叙述者口吻？\r\n3. 如果听话对象换了，这句话是否会随关系发生合理变化？\r\n4. 把这句话放到同场其他角色旁边时，声音是否仍然分得开？\r\n\r\n如果第 2 条或第 4 条答案是否定的，说明这句仍然过于模板化。\n\nFile v4.1.8:references/validation_policy.md\n\n# 校验策略\r\n\r\n## 用途\r\n\r\n通过分层校验，约束蒸馏输出的人设真实性、区分度和可演绎性，减少脑补、同质化和 OOC。\r\n\r\n## 校验层级\r\n\r\n### 1. 证据校验\r\n\r\n- 每个关键结论都必须能对应到至少一条原作正文证据。\r\n- `evidence_source` 优先记录章节、段落、片段索引，不复制大段原文。\r\n- 如果没有直接证据，结论不得写实锤，应留空或写 `证据不足`。\r\n\r\n### 2. 一致性校验\r\n\r\n- 性格特征、决策规则、语言"},{"kind":"example","language":"bash","snippet":"python tools/init_host_run.py --novel <路径> --characters A,B,C --output <run_manifest.json>"}]}}