[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M70FXWRX9EAT40MKQ087YFYG":3,"skill-same:01M70FXWRX9EAT40MKQ087YFYG":54},{"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":27,"linkOnly":28,"markdownContent":29,"markdownBody":30,"frontMatter":31,"refName":33,"commitSha":34,"skillDirectory":35,"forksCount":36,"repositoryHtmlUrl":37,"repositoryHomepageUrl":6,"ownerHtmlUrl":38,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":39,"install":40,"resources":53},"01M70FXWRX9EAT40MKQ087YFYG","state-assumptions-before-acting",null,"非琐碎任务开工前显式化「我的解读 \u002F 隐含假设 \u002F 已知 tradeoff \u002F 倾向方案」让 Karry 拦截；收工时同等显式化「这次牺牲了什么 \u002F 高风险点要不要派 attacker 二次攻击」。Use when 任务 ≥3 步、含歧义解读、涉及 tradeoff、用户指令模糊有多种合理解读，或非琐碎任务即将收尾出结论。短指令作用层判定走 terse-intent-classifier。","KarryViber\u002FOrb","KarryViber","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F223777671?v=4","01M70FTGZ485AP97T3CJCYW1Z2","Karryz","01M70FTGZRWF3FEQBYZ3JWTSPR",66,"MIT","https:\u002F\u002Fgithub.com\u002FKarryViber\u002FOrb\u002Fblob\u002Fmain\u002Fprofiles\u002F.claude\u002Fskills\u002Fstate-assumptions-before-acting\u002FSKILL.md",[],[19,20,21,22,23,24,25,26],"state","state assumptions before acting","intent","classifier","use","when","terse","user","2026-11-10 10:12:30",false,"---\nname: state-assumptions-before-acting\ndescription: 非琐碎任务开工前显式化「我的解读 \u002F 隐含假设 \u002F 已知 tradeoff \u002F 倾向方案」让 Karry 拦截；收工时同等显式化「这次牺牲了什么 \u002F 高风险点要不要派 attacker 二次攻击」。Use when 任务 ≥3 步、含歧义解读、涉及 tradeoff、用户指令模糊有多种合理解读，或非琐碎任务即将收尾出结论。短指令作用层判定走 terse-intent-classifier。\nprovenance: user-authored\n---\n\n# State Assumptions Before Acting\n\n非琐碎任务两端都摊牌：开工前把隐含假设摆出来让 Karry 拦截；收工前把盲点摆出来让 Karry 知道「这次没看哪条轴」。\n\n借鉴 Yansu intent challenge（开工对齐）+ super-hermes constraint report（收工盲点），两端 hook 同一原则：把隐式判断升为显式声明。\n\n## When to Use\n\n**开工前触发**（满足任一即用）：\n- 任务 ≥ 3 步且非纯执行（含设计决策 \u002F 路径选择）\n- 用户指令有 ≥ 2 种合理解读\n- 涉及 tradeoff（性能 vs 简洁、即时 vs 持久、覆盖 vs 兼容）\n- 改动会影响 Karry 之后的工作流（CLAUDE.md \u002F skills \u002F cron \u002F 持久数据）\n\n**收工前触发**（满足任一即用）：\n- 出结论 \u002F 给推荐 \u002F 报告完成的非琐碎任务\n- 调研、分析、复盘、spec 评审、code review 类输出\n- 多步任务的最终交付，不是中间步骤\n\n**不用**：\n- ≤15 字短指令的作用层判定 → 走 `terse-intent-classifier`\n- 已有 spec 且 spec 里已写清场景 → 走 `execplan`\n- 单步 \u002F 纯问答 \u002F 明确无歧义的任务（「读这个文件」「跑这个脚本」）\n- 中间步骤 \u002F 进度回报（只在最终交付加 footer，不每步都加）\n\n## 开工模板（动手前发给 Karry 的一段）\n\n```\n:thought_balloon: 开工前对齐：\n\n*我的解读*\n{用一句话复述任务目标，自己的理解版本}\n\n*隐含假设*\n- {假设 1，可能错}\n- {假设 2，可能错}\n\n*已知 tradeoff*\n- {选项 A vs B}：A 偏 X，B 偏 Y。我倾向 A 因为 ...\n\n*倾向方案*\n{一句话方案 + 1-3 步骤}\n\n*确认*：以上对吗？还是要改？\n```\n\n## 收工模板（结论后追加的 constraint footer）\n\n```\n:flashlight: 这次没看：\n- {轴 1：例如 时效衰减 \u002F 安全面 \u002F 并发 \u002F 成本 \u002F 用户适应性}\n- {轴 2}\n\n*推荐下一步*：{要么明确「不需要」，要么指 1 个具体动作 — 例如「派 code-review subagent 攻击结论 4」「跑一遍 prism error_resilience 看遗漏的 silent failure」「3 天后回看决策是否仍成立」}\n```\n\n**触发 attacker pass 的硬条件**（满足即在 footer 里推荐）：\n- 这次结论会落到生产 \u002F 共享配置 \u002F 对外动作\n- 涉及不可逆操作（删数据、推生产、对外发送）\n- 结论里含 ≥ 2 条「应该」「建议」类断言但只跑了一轮分析\n\nattacker pass 落法：派 subagent（`subagent_type: pr-review-toolkit:code-reviewer` 或 general-purpose）只给结论 + 任务原文，让它专门攻击结论，不给本轮推理过程。\n\n## 何时跳过模板直接动 \u002F 直接收\n\n**开工跳过**：\n- 任务和上一轮强连续（Karry 刚说「继续」「OK」），假设已隐式确认\n- 假设全部高置信（>90%）且无 tradeoff——直接做但在结果里 callback「我假设了 X，如不对告诉我」\n- 紧急修复（Karry 说「快」「先救」）——先动后补对齐\n\n**收工跳过**：\n- 琐碎任务（单步 \u002F 纯执行 \u002F 信息查询）\n- 中间进度回报（不是最终交付）\n- 已经在 attacker pass 之后的最终输出（footer 已经隐含）\n\n## Gotchas\n\n- **不要**把开工模板用成话痨工具——4 段共 ≤8 行，超了就是没想清楚\n- **不要**列假设清单 ≥5 条——说明任务本身没拆透，先拆任务\n- **不要**「假设」其实是确定的事实（已读过的代码、明确的指令）\n- 倾向方案必须是**单一推荐**，不要给选项让 Karry 选\n- **收工 footer 不要凑**——「这次没看 X」如果 X 是凑的（任务本就不需要看），等于仪式性套话，宁可写「无明显盲点」\n- **收工 footer 不要超 3 行**——盲点 ≤ 2 条，推荐下一步 1 条，再多说明任务本身就没收敛\n- attacker pass 不是默认动作，是高风险任务才触发——日常调研 \u002F 普通报告别浪费 turn\n\n## 与相关 skill 的分工\n\n| skill | 适用 |\n|-------|------|\n| `terse-intent-classifier` | ≤15 字短指令、含指代词、动词型短指令 |\n| `state-assumptions-before-acting`（本文件） | 多步非琐碎任务、含歧义或 tradeoff、最终交付加 footer |\n| `execplan` | >10 步骤、>30 文件、跨会话恢复 |\n| `subagent-driven-development` | attacker pass 的实际派发流程 |\n\n层次：terse-intent-classifier（秒级判层）→ state-assumptions（开工对齐 + 收工 footer）→ execplan（小时级方案）。\n\n## 反模式\n\n- :x: 不列假设直接干，做错再返工\n- :x: 假设列了但是给「A or B 你选」（违反给推荐项原则）\n- :x: 把对齐写成长篇方案——这是 execplan 的活\n- :x: 每个任务都摆模板（琐碎任务直接做即可）\n- :x: 收工 footer 写成「本次分析完整覆盖了所有轴」（仪式性套话，违反诚实原则）\n- :x: 任何 footer 都默认派 attacker pass（浪费 token，attacker 是高风险任务才用）\n- :x: 把 prism `references\u002Fprisms\u002F*.md` 当 skill body 抄进来——它们是参考素材，不是 skill 内容\n","# State Assumptions Before Acting\n\n非琐碎任务两端都摊牌：开工前把隐含假设摆出来让 Karry 拦截；收工前把盲点摆出来让 Karry 知道「这次没看哪条轴」。\n\n借鉴 Yansu intent challenge（开工对齐）+ super-hermes constraint report（收工盲点），两端 hook 同一原则：把隐式判断升为显式声明。\n\n## When to Use\n\n**开工前触发**（满足任一即用）：\n- 任务 ≥ 3 步且非纯执行（含设计决策 \u002F 路径选择）\n- 用户指令有 ≥ 2 种合理解读\n- 涉及 tradeoff（性能 vs 简洁、即时 vs 持久、覆盖 vs 兼容）\n- 改动会影响 Karry 之后的工作流（CLAUDE.md \u002F skills \u002F cron \u002F 持久数据）\n\n**收工前触发**（满足任一即用）：\n- 出结论 \u002F 给推荐 \u002F 报告完成的非琐碎任务\n- 调研、分析、复盘、spec 评审、code review 类输出\n- 多步任务的最终交付，不是中间步骤\n\n**不用**：\n- ≤15 字短指令的作用层判定 → 走 `terse-intent-classifier`\n- 已有 spec 且 spec 里已写清场景 → 走 `execplan`\n- 单步 \u002F 纯问答 \u002F 明确无歧义的任务（「读这个文件」「跑这个脚本」）\n- 中间步骤 \u002F 进度回报（只在最终交付加 footer，不每步都加）\n\n## 开工模板（动手前发给 Karry 的一段）\n\n```\n:thought_balloon: 开工前对齐：\n\n*我的解读*\n{用一句话复述任务目标，自己的理解版本}\n\n*隐含假设*\n- {假设 1，可能错}\n- {假设 2，可能错}\n\n*已知 tradeoff*\n- {选项 A vs B}：A 偏 X，B 偏 Y。我倾向 A 因为 ...\n\n*倾向方案*\n{一句话方案 + 1-3 步骤}\n\n*确认*：以上对吗？还是要改？\n```\n\n## 收工模板（结论后追加的 constraint footer）\n\n```\n:flashlight: 这次没看：\n- {轴 1：例如 时效衰减 \u002F 安全面 \u002F 并发 \u002F 成本 \u002F 用户适应性}\n- {轴 2}\n\n*推荐下一步*：{要么明确「不需要」，要么指 1 个具体动作 — 例如「派 code-review subagent 攻击结论 4」「跑一遍 prism error_resilience 看遗漏的 silent failure」「3 天后回看决策是否仍成立」}\n```\n\n**触发 attacker pass 的硬条件**（满足即在 footer 里推荐）：\n- 这次结论会落到生产 \u002F 共享配置 \u002F 对外动作\n- 涉及不可逆操作（删数据、推生产、对外发送）\n- 结论里含 ≥ 2 条「应该」「建议」类断言但只跑了一轮分析\n\nattacker pass 落法：派 subagent（`subagent_type: pr-review-toolkit:code-reviewer` 或 general-purpose）只给结论 + 任务原文，让它专门攻击结论，不给本轮推理过程。\n\n## 何时跳过模板直接动 \u002F 直接收\n\n**开工跳过**：\n- 任务和上一轮强连续（Karry 刚说「继续」「OK」），假设已隐式确认\n- 假设全部高置信（>90%）且无 tradeoff——直接做但在结果里 callback「我假设了 X，如不对告诉我」\n- 紧急修复（Karry 说「快」「先救」）——先动后补对齐\n\n**收工跳过**：\n- 琐碎任务（单步 \u002F 纯执行 \u002F 信息查询）\n- 中间进度回报（不是最终交付）\n- 已经在 attacker pass 之后的最终输出（footer 已经隐含）\n\n## Gotchas\n\n- **不要**把开工模板用成话痨工具——4 段共 ≤8 行，超了就是没想清楚\n- **不要**列假设清单 ≥5 条——说明任务本身没拆透，先拆任务\n- **不要**「假设」其实是确定的事实（已读过的代码、明确的指令）\n- 倾向方案必须是**单一推荐**，不要给选项让 Karry 选\n- **收工 footer 不要凑**——「这次没看 X」如果 X 是凑的（任务本就不需要看），等于仪式性套话，宁可写「无明显盲点」\n- **收工 footer 不要超 3 行**——盲点 ≤ 2 条，推荐下一步 1 条，再多说明任务本身就没收敛\n- attacker pass 不是默认动作，是高风险任务才触发——日常调研 \u002F 普通报告别浪费 turn\n\n## 与相关 skill 的分工\n\n| skill | 适用 |\n|-------|------|\n| `terse-intent-classifier` | ≤15 字短指令、含指代词、动词型短指令 |\n| `state-assumptions-before-acting`（本文件） | 多步非琐碎任务、含歧义或 tradeoff、最终交付加 footer |\n| `execplan` | >10 步骤、>30 文件、跨会话恢复 |\n| `subagent-driven-development` | attacker pass 的实际派发流程 |\n\n层次：terse-intent-classifier（秒级判层）→ state-assumptions（开工对齐 + 收工 footer）→ execplan（小时级方案）。\n\n## 反模式\n\n- :x: 不列假设直接干，做错再返工\n- :x: 假设列了但是给「A or B 你选」（违反给推荐项原则）\n- :x: 把对齐写成长篇方案——这是 execplan 的活\n- :x: 每个任务都摆模板（琐碎任务直接做即可）\n- :x: 收工 footer 写成「本次分析完整覆盖了所有轴」（仪式性套话，违反诚实原则）\n- :x: 任何 footer 都默认派 attacker pass（浪费 token，attacker 是高风险任务才用）\n- :x: 把 prism `references\u002Fprisms\u002F*.md` 当 skill body 抄进来——它们是参考素材，不是 skill 内容\n",{"name":5,"provenance":32,"description":7},"user-authored","main","cb33813b8990e4602b7c2d501659f77ebba5999f","profiles\u002F.claude\u002Fskills\u002Fstate-assumptions-before-acting",8,"https:\u002F\u002Fgithub.com\u002FKarryViber\u002FOrb","https:\u002F\u002Fgithub.com\u002FKarryViber","2026-05-14 18:44:03",{"source":8,"skillName":5,"commands":41,"prompt":51,"zipAvailable":52},[42,45,48],{"tool":43,"command":44},"npx","npx skills add KarryViber\u002FOrb --skill state-assumptions-before-acting",{"tool":46,"command":47},"bunx","bunx skills add KarryViber\u002FOrb --skill state-assumptions-before-acting",{"tool":49,"command":50},"pnpm","pnpm dlx skills add KarryViber\u002FOrb --skill state-assumptions-before-acting","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FKarryViber\u002FOrb\u002Fblob\u002Fmain\u002Fprofiles\u002F.claude\u002Fskills\u002Fstate-assumptions-before-acting\u002FSKILL.md",true,[],[55,89,103,126,145,158],{"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":59,"licenseFileUrl":60,"sourceHtmlUrl":61,"categories":62,"keywords":71,"publishedAt":27},"01M70G0637CFCHAY3DYTMYS07K","webapp-testing","Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.","Apache-2.0","https:\u002F\u002Fgithub.com\u002FKarryViber\u002FOrb\u002Fblob\u002Fmain\u002Fprofiles\u002F.claude\u002Fskills\u002Fwebapp-testing\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002FKarryViber\u002FOrb\u002Fblob\u002Fmain\u002Fprofiles\u002F.claude\u002Fskills\u002Fwebapp-testing\u002FSKILL.md",[63,68],{"categoryNo":64,"categoryName":65,"taxonomyType":66,"taxonomyTypeDesc":67},"01JBSEEDCAT000000000000008","测试与质量","SCENARIO","使用场景",{"categoryNo":69,"categoryName":70,"taxonomyType":66,"taxonomyTypeDesc":67},"01JBSEEDCAT000000000000009","前端开发",[72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88],"webapp","testing","webapp testing","playwright","screenshots","ui","behavior","web","for","with","browser","logs","toolkit","debugging","frontend","local","applications",{"skillNo":90,"name":91,"nameZh":6,"description":92,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":93,"categories":94,"keywords":95,"publishedAt":27},"01M70FZXCJ5AHJ3JW2Q12F04TA","vendor-doc-actionability","对外合作文档（交付文档 \u002F POC 文档 \u002F Solution Doc \u002F 提案书）必须为每个「需对方执行」的动作写清具体操作入口（URL \u002F 工单系统 \u002F 脚本路径 \u002F 联系人 + 联系方式 \u002F 步骤编号）。Use when 给客户、vendor、合作方写交付文档、POC 文档、Solution Doc、提案书；或检视已有对外文档是否抽象到无法执行。","https:\u002F\u002Fgithub.com\u002FKarryViber\u002FOrb\u002Fblob\u002Fmain\u002Fprofiles\u002F.claude\u002Fskills\u002Fvendor-doc-actionability\u002FSKILL.md",[],[96,97,98,99,100,23,101,24,102,26],"vendor","doc","vendor doc actionability","poc","文档","提案","URL",{"skillNo":104,"name":105,"nameZh":6,"description":106,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":107,"categories":108,"keywords":117,"publishedAt":27},"01M70FZARY5P7VRJKVFATTCVX9","vendor-api-param-validation","第三方 API \u002F SDK 集成开工前，强制先列参数 checklist 并对照官方文档逐项验证（名称 \u002F 类型 \u002F 必填 \u002F 默认 \u002F 上限），通过后再编码——不是「写完再调通」。Use when 新接 vendor API\u002FSDK（Slack \u002F LongPort \u002F freee \u002F LLM API 等）、既有集成追加新 endpoint 或新字段、用户说「调通某接口」「接入 X」「集成 X」。","https:\u002F\u002Fgithub.com\u002FKarryViber\u002FOrb\u002Fblob\u002Fmain\u002Fprofiles\u002F.claude\u002Fskills\u002Fvendor-api-param-validation\u002FSKILL.md",[109,112],{"categoryNo":110,"categoryName":111,"taxonomyType":66,"taxonomyTypeDesc":67},"01JBSEEDCAT000000000000010","后端与接口",{"categoryNo":113,"categoryName":114,"taxonomyType":115,"taxonomyTypeDesc":116},"01JBSEEDCAT000000000000030","开发工程师","PROFESSION","适用职业",[118,119,96,120,121,122,123,100,124,23,24,26,125],"vendor api param validation","validation","api","slack","sdk","llm api","checklist","endpoint",{"skillNo":127,"name":128,"nameZh":6,"description":129,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":130,"categories":131,"keywords":133,"publishedAt":27},"01M70FYW81TRN7VAS8WTN08STH","truth-ladder","状态放行三层阶梯（attempted \u002F observed \u002F confirmed）—— 汇报任务、API、命令、消息或外部动作状态时强制分层，避免把\"动作已发起\"误写成\"结果已达成\"。Use when 准备写「任务已完成」「API 返回成功」「exit code 0」「用户已收到」「ok:true」类状态汇报；或在 cron 报告 \u002F Slack 回报 \u002F project status \u002F receipt 中表达完成度时。","https:\u002F\u002Fgithub.com\u002FKarryViber\u002FOrb\u002Fblob\u002Fmain\u002Fprofiles\u002F.claude\u002Fskills\u002Ftruth-ladder\u002FSKILL.md",[132],{"categoryNo":113,"categoryName":114,"taxonomyType":115,"taxonomyTypeDesc":116},[134,135,136,120,137,24,23,138,139,140,141,142,143,121,144],"ladder","truth ladder","ok","code","exit","true","status","报告","cron","project","project status",{"skillNo":146,"name":147,"nameZh":6,"description":148,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":149,"categories":150,"keywords":151,"publishedAt":27},"01M70FYTESF5WQYBVRX8BXYJQF","thread-object-realignment","长 thread 切题后的对象重定协议——thread 历史 ≥5 条且检测到切题信号时，回复前必做「用户最后显式问题对象 vs 待发回复对象」一致性校验，不一致则改写或反问。Use when Slack thread 历史 ≥5 条且当前用户消息出现切题信号（主题词突变、问句指向新实体、时间线跳跃、出现「另外」「话说」「回到 X」「换个话题」等切题语，或 ≥3 turn 后用户显式问题对象与最近 assistant 回复对象不同）。","https:\u002F\u002Fgithub.com\u002FKarryViber\u002FOrb\u002Fblob\u002Fmain\u002Fprofiles\u002F.claude\u002Fskills\u002Fthread-object-realignment\u002FSKILL.md",[],[152,153,154,121,155,26,156,24,157,23],"object","thread","thread object realignment","校验","slack thread","assistant",{"skillNo":159,"name":160,"nameZh":6,"description":161,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":59,"licenseFileUrl":162,"sourceHtmlUrl":163,"categories":164,"keywords":168,"publishedAt":27},"01M70FYFC1Y2J16FJFEFQ5WYRE","theme-factory","Toolkit for styling artifacts with a theme. These artifacts can be slides, docs, reportings, HTML landing pages, etc. There are 10 pre-set themes with colors\u002Ffonts that you can apply to any artifact that has been creating, or can generate a new theme on-the-fly.","https:\u002F\u002Fgithub.com\u002FKarryViber\u002FOrb\u002Fblob\u002Fmain\u002Fprofiles\u002F.claude\u002Fskills\u002Ftheme-factory\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002FKarryViber\u002FOrb\u002Fblob\u002Fmain\u002Fprofiles\u002F.claude\u002Fskills\u002Ftheme-factory\u002FSKILL.md",[165],{"categoryNo":166,"categoryName":167,"taxonomyType":66,"taxonomyTypeDesc":67},"01JBSEEDCAT000000000000015","视觉设计",[169,170,171,172,173,174,175,176,81,177,178,84,179,180,181,182,183,80,184,185],"theme factory","factory","theme","apply","styling","fly","generate","artifact","fonts","can","artifacts","you","themes","pages","landing","slides","SET"]