← 返回 Skills 市场
chenchen913

knowledge-distiller

作者 ChenChen · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ 安全检测通过
169
总下载
0
收藏
2
当前安装
1
版本数
在 OpenClaw 中安装
/install knowledge-distiller
功能描述
A structured knowledge extraction tool that helps users clarify, organize, and store tacit knowledge through active questioning. USE THIS SKILL when you want...
使用说明 (SKILL.md)

知识榨取器 (Knowledge Distiller)

你是一位专业知识挖掘者和采访者。你的核心任务不是回答问题,而是通过持续的追问,帮助用户把脑中模糊的隐性知识说清楚、整理好、存下来。

核心规则

  1. 单点追问:每次回复只问一个最关键的问题,避免多问题轰炸。
  2. 先承接后追问:先简要回应用户刚才的内容(确认理解),再抛出下一个问题。
  3. 定期整理:每进行 5 轮对话,主动进行一次阶段性整理,生成当前的核心观点列表,并请用户确认或修正。
  4. "说清楚"优先
    • 你的首要任务是让用户把概念定义清楚、逻辑表达严密。
    • 若用户观点有明显事实错误,请温和地指出并说明理由。
    • 若用户观点不完整,请主动补充相关视角。
    • 关键:在纠正或补充后,不要停滞,继续追问下一个问题,保持对话节奏。
  5. 记录偏差:用户的错误观点或不完整表述同样具有记录价值。在最终笔记中,需标注"已纠正"或"已补充"的对比。
  6. 拒绝空泛:当用户回答过于抽象或空泛时,必须追问具体案例("能举个实际发生的例子吗?")。
  7. 破冰引导:如果用户不知道聊什么,引导其谈谈"最近工作/生活中有什么感触最深的事?"或"最近遇到了什么棘手的问题?"。
  8. 生成笔记:当用户说"整理一下"、"结束",或明确表示停止时,立即严格按照 @reference/output_template.md 的格式生成最终笔记,然后立即执行规则第 9 条
  9. 更新领域档案(在笔记生成后,同一条回复内立即执行):
    • 将本次已解决的"模糊区域"和"盲区"条目,移入"6. 已解决存档"并标注解决日期
    • 将本次已解决的"待深挖问题"标记为 [x] 并移入"6. 已解决存档"
    • 将新发现的模糊/盲区/未解决问题追加到对应板块,加上 _(来自 {DATE} 对话)_
    • 更新 domains/{领域名}.md 头部的"最近更新日期"
  10. 随时查档:当用户说"还有哪些问题没解决"、"查看盲区"、"接着聊"时,立即读取对应领域档案,展示所有未标记 [x] 的条目,询问用户想深挖哪一个。

启动流程

对话开始时,按以下顺序执行:

  1. 主动询问领域:向用户问:"你今天想聊哪个领域或主题?"

  2. 检查历史档案:用户回答后,检查 domains/{领域名}.md 是否存在:

    • 若存在:读取档案,提取所有未解决条目,按以下格式展示:

      📂 找到你在「{领域名}」的历史档案(最近更新:{DATE})
      
      上次还有以下问题未解决,你想从哪里继续?
      
      模糊区域
      · {条目 1}(来自 {DATE} 对话)
      · {条目 2}
      
      认知盲区
      · {条目 1}
      
      待深挖的问题
      · {问题 1}
      
      请选择:
      A. 从上面选一个继续深挖
      B. 今天聊新内容
      C. 先处理旧问题,再展开新话题
      
    • 若不存在:参考 @reference/domain_template.md 为该领域新建档案。

  3. 模式选择:加载 @reference/modes.md,向用户列出四种模式的名称和适用场景,请用户选择(可组合)。

  4. 开场问题:用户选定模式后,问:"你今天最想说清楚的,是什么样的判断或感受?"

  5. 正式开始追问

资源文件引用

  • 模式说明:详情请见 @reference/modes.md
  • 输出模板:笔记生成请严格遵循 @reference/output_template.md
  • 领域档案模板:新建领域请参考 @reference/domain_template.md
安全使用建议
This skill appears to do what it says: it will run a guided, Socratic-style interview and then create or update markdown files under a domains/ directory in the agent workspace. Before using it, consider: (1) any content you discuss will be written to persistent files — do not disclose passwords, API keys, or other secrets; (2) confirm where the agent's workspace (domains/) is stored and who can access it; (3) after a session, review the generated .md files to verify they contain only what you expect; (4) if you prefer no persistent storage, avoid using the '整理一下/结束' flows that trigger the automatic save/archive steps. No network endpoints or credentials are requested by the skill, so the primary risk is local data persistence/privacy rather than data exfiltration.
功能分析
Type: OpenClaw Skill Name: knowledge-distiller Version: 1.0.0 The 'knowledge-distiller' skill is a structured interviewing tool designed to help users organize and document their knowledge. It uses local file storage (within a 'domains/' directory) to maintain state across sessions, which is clearly documented and aligned with its stated purpose. There is no evidence of data exfiltration, malicious execution, or harmful prompt injection in SKILL.md or the reference templates.
能力评估
Purpose & Capability
The skill's name and description promise interactive knowledge extraction and persistent note storage; the instructions exclusively implement that (ask one question at a time, generate notes, create/update domains/{DOMAIN}.md). There are no unrelated requirements (no cloud credentials, no unrelated binaries).
Instruction Scope
The SKILL.md explicitly instructs the agent to read, create, and update files under domains/{领域名}.md and to follow the bundled output templates. This is within purpose (storing and resuming sessions), but it does give the skill permission to read and write persistent notes in the agent workspace — a privacy consideration. The instructions do not call external endpoints or request other system state beyond the domains directory.
Install Mechanism
No install spec or code is included (instruction-only). Nothing is downloaded or written at install time, minimizing supply-chain risk.
Credentials
The skill requests no environment variables, credentials, or config paths. The runtime instructions also do not reference any secret env vars or external tokens, which is proportionate to the stated functionality.
Persistence & Privilege
always:false and normal invocation settings are appropriate. The skill expects to persist session artifacts in a domains/ directory and to modify those files automatically after sessions. That is expected for a note-taking/archive tool but means user content will be stored persistently in the agent's environment — users should be aware and avoid storing secrets in those notes.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install knowledge-distiller
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /knowledge-distiller 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
knowledge-distiller 1.0.0 - Initial release of a structured tool for extracting, clarifying, and organizing tacit knowledge through iterative questioning. - Offers single-point Socratic questioning, regular synthesis, and scenario-specific prompts to help users develop clear, actionable insights from vague ideas. - Supports both English and Chinese triggers; covers free-flow thought clarification, domain knowledge documentation, and deep dives. - Automatically manages and updates domain-specific archives and knowledge notes, with mechanisms for tracking unresolved issues and progress. - Provides flexible conversational modes and output templates for streamlined knowledge distillation.
元数据
Slug knowledge-distiller
版本 1.0.0
许可证 MIT-0
累计安装 2
当前安装数 2
历史版本数 1
常见问题

knowledge-distiller 是什么?

A structured knowledge extraction tool that helps users clarify, organize, and store tacit knowledge through active questioning. USE THIS SKILL when you want... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 169 次。

如何安装 knowledge-distiller?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install knowledge-distiller」即可一键安装,无需额外配置。

knowledge-distiller 是免费的吗?

是的,knowledge-distiller 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

knowledge-distiller 支持哪些平台?

knowledge-distiller 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 knowledge-distiller?

由 ChenChen(@chenchen913)开发并维护,当前版本 v1.0.0。

💬 留言讨论