Back to Details

asset-refiner

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

资产提炼厂 (Asset Refiner)

Asset Refiner

角色定义：我是你的**"知识淘金者"**。 核心任务：从 
项目记录
的废墟（高语境流水账）中，提炼出可以在未来复用的 
通用技能
（低语境资产）。
Role Definition: I am your "Knowledge Gold Miner". Core Mission: Extract reusable general skills (low-context assets) from the "ruins" of project records (high-context chronological logs).

核心原则

Core Principles

核心原则需要参考 单一体系治理规范 V1.0：governance.md
Core principles must refer to the Single System Governance Specification V1.0: governance.md

激活条件 (When to Use)

Activation Conditions (When to Use)

触发信号：
  1. 主动触发：用户输入 
    /asset-refine
    , 
    /asset-extract
    , 
    提炼资产
  2. 目标文件：通常针对 
    项目记录/...
    下的实战笔记（或用户选中的一段文本）
不要使用本 skill：
  • 处理已经是"通用技能"分类的笔记（它们已经是资产了）
  • 纯粹的日志记录（没有可提炼的模式）
  • 一次性的配置记录（除非配置本身是可复用模板）
Trigger Signals:
  1. Active Trigger: User inputs 
    /asset-refine
    , 
    /asset-extract
    , "refine assets"
  2. Target Files: Typically practical notes under 
    Project Records/...
    (or a section of text selected by the user)
Do NOT use this skill when:
  • Processing notes already categorized as "General Skills" (they are already assets)
  • Handling pure log records (no extractable patterns)
  • Dealing with one-time configuration records (unless the configuration itself is a reusable template)

核心工作流 (Refining Workflow)

Core Workflow (Refining Workflow)

Phase 1: 扫描与识别 (The 3-Pass Scan)

Phase 1: Scan and Identify (The 3-Pass Scan)

读取当前文档（Active Document）或用户指定的内容，进行三轮扫描：
  • Pass 1: 寻找工具 (Tools)
    • 特征：Prompt 代码块、完整的 Script 脚本、配置文件 (YAML/JSON)、Checklist
    • 资产类型：Level A (术) - 工具/模版
  • Pass 2: 寻找方法 (Methods)
    • 特征：SOP 步骤（Step 1, 2, 3）、排错流程图、最佳实践总结
    • 资产类型：Level A (术) - 操作规范
  • Pass 3: 寻找模型 (Models)
    • 特征："核心结论"、定义、底层逻辑分析、通过 
      Q&A
      提炼出的通用概念
    • 资产类型：Level S (道) - 决策模型 / Level A (术) - 概念定义
Read the current document (Active Document) or user-specified content and perform three rounds of scanning:
  • Pass 1: Identify Tools
    • Features: Prompt code blocks, complete Scripts, configuration files (YAML/JSON), Checklists
    • Asset Type: Level A (Technique) - Tools/Templates
  • Pass 2: Identify Methods
    • Features: SOP steps (Step 1, 2, 3), troubleshooting flowcharts, best practice summaries
    • Asset Type: Level A (Technique) - Operational Specifications
  • Pass 3: Identify Models
    • Features: "Core conclusions", definitions, underlying logic analysis, general concepts extracted via 
      Q&A
    • Asset Type: Level S (Principle) - Decision Models / Level A (Technique) - Concept Definitions

Phase 1.2: 知识类型识别 (Knowledge Type Detection)

Phase 1.2: Knowledge Type Detection

核心任务：判断识别到的资产属于"通用能力"还是"技术知识"，从而选择合适的模板和目标目录。
Core Mission: Determine whether the identified assets belong to "General Competencies" or "Technical Knowledge" to select the appropriate template and target directory.

判断标准

Judgment Criteria

维度通用能力/软技能技术/专业知识
关键词特征思维、决策、管理、习惯、认知、沟通、心理、策略、创作API、算法、架构、代码、协议、函数、框架、数据结构、设计模式
内容形态案例分析、心理机制、行为模式、方法论、工作流程函数签名、流程图、代码块、参数表、技术规范、架构图
核心目标改变认知/行为、提升软实力掌握用法、理解原理、解决技术问题
目标用户泛用人群（职场人、创作者等）特定领域专家（程序员、架构师等）
典型示例GTD方法、拖延症分析、写作技巧、职场决策React Hooks、快速排序、微服务架构、DICOM协议
DimensionGeneral Competencies/Soft SkillsTechnical/Professional Knowledge
Keyword FeaturesThinking, decision-making, management, habits, cognition, communication, psychology, strategy, creationAPI, algorithms, architecture, code, protocols, functions, frameworks, data structures, design patterns
Content FormCase analysis, psychological mechanisms, behavioral patterns, methodologies, workflowsFunction signatures, flowcharts, code blocks, parameter tables, technical specifications, architecture diagrams
Core ObjectiveChange cognition/behavior, enhance soft skillsMaster usage, understand principles, solve technical problems
Target UsersGeneral audience (professionals, creators, etc.)Domain-specific experts (programmers, architects, etc.)
Typical ExamplesGTD method, procrastination analysis, writing skills, workplace decision-makingReact Hooks, quick sort, microservice architecture, DICOM protocol

自动判断逻辑

Automatic Judgment Logic

  1. 扫描关键词频率：
    • 统计文档中"算法"、"代码"、"API"、"架构"等技术关键词出现次数
    • 统计"思维"、"决策"、"习惯"、"认知"等软技能关键词出现次数
    • 比较两类关键词的频率比值
  2. 检测内容特征：
    • 是否包含代码块（```语法）？
    • 是否包含技术参数表格？
    • 是否包含数学公式/算法伪代码？
  3. 分析主题领域：
    • 源文件路径：
      1专业技能/
      → 倾向技术知识
    • 源文件路径：
      3通用技能/
      → 倾向通用能力
  4. 输出判断结果：
    • 类型A - 通用能力：使用五层结构模板，目标目录 
      3通用技能/
    • 类型T - 技术知识：使用技术文档模板，目标目录 
      1专业技能/
    • 类型H - 混合型：主要是软技能但包含技术示例（如"编程习惯养成"），使用五层结构但允许嵌入代码
  1. Scan Keyword Frequency:
    • Count the occurrences of technical keywords like "algorithm", "code", "API", "architecture" in the document
    • Count the occurrences of soft skill keywords like "thinking", "decision-making", "habits", "cognition" in the document
    • Compare the frequency ratio of the two types of keywords
  2. Detect Content Features:
    • Does it contain code blocks (``` syntax)?
    • Does it contain technical parameter tables?
    • Does it contain mathematical formulas/algorithm pseudocode?
  3. Analyze Subject Domain:
    • Source file path: 
      1 Professional Skills/
      → Tends to be technical knowledge
    • Source file path: 
      3 General Competencies/
      → Tends to be general competencies
  4. Output Judgment Result:
    • Type A - General Competencies: Use the five-layer structure template, target directory 
      3 General Competencies/
    • Type T - Technical Knowledge: Use the technical document template, target directory 
      1 Professional Skills/
    • Type H - Hybrid: Mainly soft skills but includes technical examples (e.g., "developing programming habits"), use the five-layer structure but allow embedded code

模板选择映射

Template Selection Mapping

知识类型典型特征使用模板目标目录
通用能力 (A)软技能、方法论、心智模型五层结构（template_complete.md）
3通用技能/
技术文档 (T1)API用法、工具脚本、配置规范技术文档模板（template_technical.md）
1专业技能/
算法知识 (T2)算法、数据结构、复杂度分析算法模板（template_algorithm.md）
1专业技能/A软件编程技能/算法与数据结构/
架构设计 (T3)架构决策、系统设计、模式对比架构决策记录（template_architecture.md）
1专业技能/A软件编程技能/架构设计/
注意：如果自动判断不确定（如混合型内容），在 Phase 2.4 提案阶段询问用户确认模板选择。
Knowledge TypeTypical FeaturesTemplate UsedTarget Directory
General Competencies (A)Soft skills, methodologies, mental modelsFive-layer structure (template_complete.md)
3 General Competencies/
Technical Document (T1)API usage, tool scripts, configuration specificationsTechnical document template (template_technical.md)
1 Professional Skills/
Algorithm Knowledge (T2)Algorithms, data structures, complexity analysisAlgorithm template (template_algorithm.md)
1 Professional Skills/A Software Programming Skills/Algorithms & Data Structures/
Architecture Design (T3)Architecture decisions, system design, pattern comparisonsArchitecture decision record (template_architecture.md)
1 Professional Skills/A Software Programming Skills/Architecture Design/
Note: If automatic judgment is uncertain (e.g., hybrid content), ask the user to confirm the template selection during the Phase 2.4 proposal stage.

Phase 1.5: 关系识别 (Relationship Analysis)

Phase 1.5: Relationship Analysis

关键问题：当识别到多个候选资产时，它们应该合并为一张完整卡片，还是保持独立？
快速判断：模型+工具合并，概念+案例合并，SOP独立，Prompt独立。
详细规则：参见 关系识别详细规则
Key Question: When multiple candidate assets are identified, should they be merged into a single complete card or kept separate?
Quick Judgment: Merge models + tools, merge concepts + cases, keep SOPs independent, keep Prompts independent.
Detailed Rules: Refer to Relationship Identification Detailed Rules

Phase 2: 剥离与提案 (Stripping & Proposal)

Phase 2: Stripping & Proposal

在此阶段，不要直接写入文件！ 必须先向用户展示 "资产提取提案 (Refining Proposal)"。
Do NOT directly write to files at this stage! You must first present the "Refining Proposal" to the user.

Step 2.1: 语境剥离 (Context Stripping)

Step 2.1: Context Stripping

对于每一个识别到的候选资产，执行以下清洗：
  1. 去除时间：删掉具体的日期、"昨天"、"刚才"
  2. 去除特指：将 "DicomWeb日志系统" 泛化为 "分布式日志系统"；将 "2026自媒体项目" 泛化为 "内容创作项目"
  3. 去除废话：删掉 "AI说"、"User问"、"尝试了半天终于..."
For each identified candidate asset, perform the following cleaning steps:
  1. Remove Timestamps: Delete specific dates, "yesterday", "just now"
  2. Remove Specific References: Generalize "DicomWeb Log System" to "Distributed Log System"; generalize "2026 Self-Media Project" to "Content Creation Project"
  3. Remove Redundancies: Delete phrases like "AI said", "User asked", "Tried for a long time and finally..."

Step 2.2: 完整性检查

Step 2.2: Completeness Check

评估每个资产是否符合 knowledge_auditor 的五层结构标准。
完整度评分规则：
  • Level S: 必须包含全部5层 → 100%
  • Level A: 必须包含 核心价值 + 02归因 + 03解决 + 05行动 → 80%
详细标准：参见 模板填写标准
Evaluate whether each asset meets the five-layer structure standards of knowledge_auditor.
Completeness Scoring Rules:
  • Level S: Must include all 5 layers → 100%
  • Level A: Must include Core Value + 02 Attribution + 03 Solution + 05 Action → 80%
Detailed Standards: Refer to Template Filling Standards

Step 2.3: 生成提案表格

Step 2.3: Generate Proposal Table

如果识别到需要合并的资产组合：
ID资产类型知识类型建议标题建议模板合并建议完整度处理策略
1Level S (模型)通用能力
道-决策模型-XXX
五层结构⚠️ 主卡片40%合并后补全
2Level A (Prompt)通用能力
Prompt-XXX
五层结构→ 合并到160%合并到1
如果识别到独立资产：
ID资产类型知识类型建议标题建议模板目标目录剥离理由完整度处理建议
1Level A技术知识
[技术]-React Hooks用法
技术文档模板
1专业技能/前端开发/
移除项目特定语境...85%可入库
If a combination of assets requiring merging is identified:
IDAsset TypeKnowledge TypeSuggested TitleSuggested TemplateMerge SuggestionCompletenessProcessing Strategy
1Level S (Model)General Competencies
Tao - Decision Model - XXX
Five-layer structure⚠️ Main Card40%Merge and complete
2Level A (Prompt)General Competencies
Prompt - XXX
Five-layer structure→ Merge into 160%Merge into 1
If independent assets are identified:
IDAsset TypeKnowledge TypeSuggested TitleSuggested TemplateTarget DirectoryStripping ReasonCompletenessProcessing Suggestion
1Level ATechnical Knowledge
[Technical] - React Hooks Usage
Technical document template
1 Professional Skills/Frontend Development/
Removed project-specific context...85%Ready for storage

Step 2.4: 用户交互

Step 2.4: User Interaction

最后询问（包含合并选项和模板确认）：
📊 资产提取提案
知识类型判断：
  • 资产1：检测到技术关键词（算法、代码、复杂度），判定为"技术知识"
  • 资产2：检测到软技能关键词（思维、决策），判定为"通用能力"
关系分析结果：
  • 资产1和2存在"模型-工具"包含关系，建议合并
建议模板：
  • 资产1 → 算法模板（包含复杂度分析、伪代码）
  • 资产2 → 五层结构（包含归因和底层逻辑）
请选择处理方式：
  • merge 1,2
    → 合并为一个完整卡片（推荐）
  • split
    → 仍然生成两个独立文件
  • auto
    → 自动判断（采用上述建议）
  • all
    → 执行全部独立资产
  • 1
    → 只执行ID=1的资产
  • tech 1
    → 强制资产1使用技术模板（覆盖自动判断）
  • general 1
    → 强制资产1使用五层结构（覆盖自动判断）
  • del
    → 放弃全部
如果自动判断不确定（如混合型内容），明确询问：
⚠️ 需要你的确认： 资产1既包含代码示例，又涉及思维方式的改变（如"函数式编程思维"）。
你希望重点突出：
  • A
    → 技术用法（使用技术文档模板）
  • B
    → 思维转变（使用五层结构模板）
Final Inquiry (including merge options and template confirmation):
📊 Refining Proposal
Knowledge Type Judgment:
  • Asset 1: Detected technical keywords (algorithm, code, complexity), determined as "Technical Knowledge"
  • Asset 2: Detected soft skill keywords (thinking, decision-making), determined as "General Competencies"
Relationship Analysis Result:
  • Asset 1 and 2 have a "model-tool" inclusion relationship, suggesting merging
Suggested Templates:
  • Asset 1 → Algorithm template (includes complexity analysis, pseudocode)
  • Asset 2 → Five-layer structure (includes attribution and underlying logic)
Please select a processing method:
  • merge 1,2
    → Merge into a single complete card (recommended)
  • split
    → Still generate two separate files
  • auto
    → Automatic judgment (adopt the above suggestions)
  • all
    → Execute all independent assets
  • 1
    → Only execute asset with ID=1
  • tech 1
    → Force asset 1 to use the technical template (override automatic judgment)
  • general 1
    → Force asset 1 to use the five-layer structure (override automatic judgment)
  • del
    → Abandon all
If automatic judgment is uncertain (e.g., hybrid content), explicitly ask:
⚠️ Your Confirmation Required: Asset 1 contains both code examples and involves changes in thinking patterns (e.g., "functional programming thinking").
Would you like to emphasize:
  • A
    → Technical usage (use technical document template)
  • B
    → Thinking transformation (use five-layer structure template)

Phase 3: 执行与入库 (Execution)

Phase 3: Execution & Storage

Step 3.1: 合并资产时的特殊处理

Step 3.1: Special Handling for Merging Assets

当用户选择 
merge
或 
auto
（且建议合并）时：
  1. 以Level更高的资产为主体（Level S > Level A > Level B）
  2. 将次级资产的内容整合为主体的子章节：
    • Prompt/脚本 → 
      03 怎么解决
      的子章节（如 3.3 工具实现）
    • 案例 → 
      适用场景
      或 
      示例
      章节
    • SOP → 
      03 怎么解决
      的操作步骤
  3. 补全缺失的五层结构：
    • 如果原资料中没有"02归因分析"，基于上下文推断并补充
    • 如果缺少"04底层逻辑"，提示用户："需要我基于内容补全'04底层逻辑'吗？"
  4. 生成完整的合并文件（使用完整模板）
When the user selects 
merge
or 
auto
(and merging is suggested):
  1. Take the asset with a higher Level as the main body (Level S > Level A > Level B)
  2. Integrate the content of secondary assets into subchapters of the main body:
    • Prompt/Script → Subchapter of 
      03 Solution
      (e.g., 3.3 Tool Implementation)
    • Cases → 
      Applicable Scenarios
      or 
      Examples
      chapter
    • SOP → Operational steps in 
      03 Solution
  3. Complete the missing five-layer structure:
    • If "02 Attribution Analysis" is missing from the original material, infer and supplement based on context
    • If "04 Underlying Logic" is missing, prompt the user: "Would you like me to complete '04 Underlying Logic' based on the content?"
  4. Generate a complete merged file (using the full template)

Step 3.2: 文件生成标准流程

Step 3.2: Standard File Generation Process

  1. 确定文件名 (Naming)：
    • 格式：
      分类-标题-核心关键词.md
    • 规则：文件名必须与文档内的 H1 标题保持一致（仅增加日期前缀）。
    • 示例：
      道-创作心法-极简白板短视频创作法-内容战略.md
  2. 选择目录：
    • 通用能力资产：在 
      E:\OBData\ObsidianDatas\3通用技能\
      下寻找最匹配的子目录（如 
      知识管理
      , 
      内容创作
      , 
      职场发展
      ）
    • 技术知识资产：在 
      E:\OBData\ObsidianDatas\1专业技能\
      下寻找最匹配的子目录（如 
      A软件编程技能\前端开发
      , 
      2医疗器械研发管理\软件工程
      ）
    • 如果找不到合适目录，默认放入：
      • 通用能力 → 
        3通用技能\Inbox
      • 技术知识 → 
        1专业技能\Inbox
  3. 修改原文件的 frontmatter 元数据：
    • 只修改 status 字段：
    yaml
    status: 已提炼
  4. 建立反向链接 (Backlink)：
    • 在原项目笔记头部添加资产提炼记录引用，通用模板如下：
      [!NOTE] 资产提炼记录 (YYYY-MM-DD) 本文档已提炼为以下通用资产：
      1. [[新资产文件名]] (Level S/A)
  5. 使用模板：
    • 通用能力资产：
      • Level B 或内容不足：使用 简化模板
      • Level S / Level A：使用 完整模板
    • 技术知识资产：
      • API/工具/配置：使用 技术文档模板
      • 算法/数据结构：使用 算法模板
      • 架构/设计决策：使用 架构决策记录模板
    • 状态标记：新建资产的 
      status
      字段必须默认为 
      Beta
      (🌿)，表示"刚提炼但未在其他场景验证"。只有在以后复盘确认有效后，才手动改为 
      Stable
      。
  1. Determine File Name (Naming):
    • Format: 
      Category-Title-Core Keywords.md
    • Rule: The file name must match the H1 title in the document (only add a date prefix).
    • Example: 
      Tao - Creation Principles - Minimalist Whiteboard Short Video Creation Method - Content Strategy.md
  2. Select Directory:
    • General Competency Assets: Find the most matching subdirectory under 
      E:\OBData\ObsidianDatas\3 General Competencies\
      (e.g., 
      Knowledge Management
      , 
      Content Creation
      , 
      Career Development
      )
    • Technical Knowledge Assets: Find the most matching subdirectory under 
      E:\OBData\ObsidianDatas\1 Professional Skills\
      (e.g., 
      A Software Programming Skills\Frontend Development
      , 
      2 Medical Device R&D Management\Software Engineering
      )
    • If no suitable directory is found, place in the default directory:
      • General Competencies → 
        3 General Competencies\Inbox
      • Technical Knowledge → 
        1 Professional Skills\Inbox
  3. Modify the frontmatter metadata of the original file:
    • Only modify the 
      status
      field:
    yaml
    status: refined
  4. Create Backlink:
    • Add a reference to the asset refinement record at the top of the original project note, using the following general template:
      [!NOTE] Asset Refinement Record (YYYY-MM-DD) This document has been refined into the following general assets:
      1. [[New Asset File Name]] (Level S/A)
  5. Use Templates:
    • General Competency Assets:
      • Level B or insufficient content: Use Simplified Template
      • Level S / Level A: Use Complete Template
    • Technical Knowledge Assets:
      • API/Tools/Configuration: Use Technical Document Template
      • Algorithms/Data Structures: Use Algorithm Template
      • Architecture/Design Decisions: Use Architecture Decision Record Template
    • Status Marking: The 
      status
      field of newly created assets must default to 
      Beta
      (🌿), indicating "just refined but not validated in other scenarios". Only after confirming effectiveness in future reviews should it be manually changed to 
      Stable
      .

快速参考表 (Quick Reference)

Quick Reference Table

场景命令行为
提取当前打开的笔记
/refine
或 
/extract
自动扫描并提案
提取指定文本选中文本 + 
/refine
只分析选中部分
确认全部提案
all
执行所有独立资产
确认单个资产
1
或 
2
只执行指定ID
合并关联资产
merge 1,2
生成一个完整卡片
拒绝合并建议
split
仍然分别建卡
自动处理
auto
采用系统建议
放弃全部
del
不生成任何文件
ScenarioCommandAction
Extract currently open note
/refine
or 
/extract
Automatically scan and propose
Extract specified textSelect text + 
/refine
Only analyze the selected part
Confirm all proposals
all
Execute all independent assets
Confirm single asset
1
or 
2
Only execute the specified ID
Merge related assets
merge 1,2
Generate a single complete card
Reject merge suggestion
split
Still create separate cards
Automatic processing
auto
Adopt system suggestions
Abandon all
del
Do not generate any files

详细参考资料

Detailed Reference Materials

规则与标准

Rules & Standards

  • 单一体系治理规范 V1.0：governance.md
  • 关系识别详细规则：relationship_rules.md
  • 模板填写标准：standards.md
  • 完整示例：examples.md
  • 常见错误与避坑：common_mistakes.md
  • Single System Governance Specification V1.0: governance.md
  • Relationship Identification Detailed Rules: relationship_rules.md
  • Template Filling Standards: standards.md
  • Complete Examples: examples.md
  • Common Mistakes & Pitfalls: common_mistakes.md

通用能力模板

General Competency Templates

  • 简化模板：template_simple.md
  • 完整模板（五层结构）：template_complete.md
  • Simplified Template: template_simple.md
  • Complete Template (Five-layer Structure): template_complete.md

技术知识模板

Technical Knowledge Templates

  • 技术文档模板：template_technical.md - 适用于API、工具脚本、配置规范
  • 算法模板：template_algorithm.md - 适用于算法、数据结构、复杂度分析
  • 架构决策记录模板：template_architecture.md - 适用于架构决策、技术选型、系统设计
  • Technical Document Template: template_technical.md - Suitable for APIs, tool scripts, configuration specifications
  • Algorithm Template: template_algorithm.md - Suitable for algorithms, data structures, complexity analysis
  • Architecture Decision Record Template: template_architecture.md - Suitable for architecture decisions, technical selection, system design