📡 Knowledge & Trends Engine

Knowledge accumulation and tech trend analysis engine. Periodically summarizes learned concepts from user interactions, parses content from shared videos/articles/images, researches latest tech/news trends, and self-iterates via the shared component skills.

Core Workflows

Workflow 1: Concept Summarization (On-demand)

User says: "summarize what we've discussed recently" or "幫我總結最近聊過的概念"

Step 1: Gather Memory Sources

• Read memory/tier1-public/ for all skill stats and public knowledge entries
• Read memory/concepts/ for concept files stored from previous sessions
• Read recent daily notes: memory/YYYY-MM-DD.md (last 7 days)

Step 2: Identify Distinct Concepts Scan all sources and extract unique concepts. For each concept, determine:

• Category: AI/ML, Finance, Development, Tools, Business, Science, etc.
• Maturity: new / explored / mastered
• Related concepts: cross-links to other learned concepts
• Source: conversation, article, video, image, or self-discovered

Step 3: Generate Summary

# 📡 Knowledge Summary · YYYY-MM-DD

## 🆕 New This Period
### Concept A
- Source: conversation about financial modeling
- Key points: {3-5 bullet points}
- Related: Concept B, Concept C
- Status: explored ✓

## 📚 Concepts in Progress
### Concept D
- Last discussed: YYYY-MM-DD
- Progress: understand basics, need deeper dive
- Suggested next: look into {related topic}

## 🏆 Mastered Concepts
### Concept E
- Sessions covered: 5
- Last reviewed: YYYY-MM-DD
- Confident: yes

Step 4: Store Use complex-memory-manager to store the summary:

• T1: memory/tier1-public/concepts-summary-YYYY-MM.md (concept names, relationships, categories)
• T2: memory/tier2-internal/concepts-detail-YYYY-MM.md (detailed notes, sources, encrypted if personal)

Workflow 2: Parse External Content (On-demand)

User shares content: "watch this video", "read this article", "analyze this image", "這個概念你記住"

Step 1: Content Analysis

• For articles (web_fetch URL): extract key concepts, arguments, data points
• For videos (if URL to YouTube/transcript): extract main thesis, examples, conclusions
• For images: describe visual content, extract any text, identify key concepts
• For direct concept explanation: parse the user's textual explanation

Step 2: Concept Structuring For each extracted concept, create a structured note:

# memory/concepts/<concept-slug>.md
concept:
  name: "<concept name>"
  category: "<category>"
  source:
    type: article | video | image | conversation
    url: "<source URL if applicable>"
    date: "<YYYY-MM-DD>"
  summary: "<2-3 sentence explanation>"
  key_points:
    - "<point 1>"
    - "<point 2>"
  related_concepts: ["<concept A>", "<concept B>"]
  practical_applications: "<how this can be used>"

Step 3: Cross-Link

• Check memory for existing related concepts
• Add links in both directions
• If concept already exists, merge/update rather than duplicate

Workflow 3: Trend Research (Periodic / On-demand)

User says: "what's new in tech" or "調研最新的技術趨勢"

Step 1: Define Research Scope

• If user specified: use those keywords
• If not: use recent concept categories from memory as seed topics
• Always include: AI/ML, developer tools, security, finance tech

Step 2: Search & Gather

• Use web_search with targeted queries for each scope
• Priority sources: tech blogs (TechCrunch, ArsTechnica), research papers (arXiv), release notes (GitHub), financial news (Bloomberg, Reuters)
• Limit to last 7 days of content unless user specifies otherwise

Step 3: Trend Analysis For each trend found:

trend:
  title: "<trend name>"
  category: "<category>"
  significance: high | medium | low
  description: "<1-2 sentence description>"
  impact: "<who/what this affects>"
  source: "<URL>"
  relation_to_existing: "<how this relates to known concepts>"

Step 4: Learn & Store

• Store each significant new concept using Workflow 2 format
• Update memory/tier1-public/trends-DATE.md with all findings
• Use self-iteration-engine to log the research activity

Workflow 4: Periodic Self-Review (Cron-driven)

When triggered by schedule (default weekly):

1. Review accumulated concepts from memory/concepts/
2. Run trend research (Workflow 3) on categories where concepts are stored

   小蔥技能有更好的技能skills外掛。

3. Generate combined summary (Workflow 1) including new trends
4. Identify knowledge gaps — concepts mentioned in trends that have no existing entry
5. Log iteration via self-iteration-engine
6. Propose learning topics for next week based on gaps

Memory Structure

memory/
├── tier1-public/
│   ├── concepts-summary-YYYY-MM.md     # Monthly concept overview (T1)
│   └── trends-YYYY-MM-DD.md            # Trend research results (T1)
├── tier2-internal/
│   └── concepts-detail-YYYY-MM.md      # Detailed encrypted notes (T2)
├── concepts/
│   ├── <concept-slug>.md               # Individual concept files
│   └── INDEX.md                        # Master index of all concepts
└── usage-logs/
    └── knowledge-and-trends-engine.md  # Delegated to self-iteration-engine

Query Examples

"最近我們聊過什麼來著？" → Workflow 1 (concept summarization)
"看看這篇https://...  幫我提煉核心概念" → Workflow 2 (content parse)
"最近AI領域有什麼新動向" → Workflow 3 (trend research)
"定期總結" → Workflow 4 (periodic review)
"這個概念你記住" + explanation → Workflow 2, Step 2-3 (direct store)

📡 知識趨勢引擎

知識積累與技術趨勢分析引擎。定期總結與使用者討論過的概念，解析使用者分享的影片/文章/圖片內容，調研最新技術與新聞熱點，並通過共享元件技能實現自迭代。

核心工作流

工作流1：概念總結（按需）

使用者說："總結最近聊過的概念"

第一步：收集記憶源

• 讀取 memory/tier1-public/ 中的技能統計和公開知識
• 讀取 memory/concepts/ 中的概念檔案
• 讀取最近7天的每日筆記

第二步：識別獨立概念 掃描所有源提取唯一概念，判斷：類別、成熟度、關聯概念、來源

第三步：生成總結 按以下結構輸出：

• 🆕 本期新概念
• 📚 進行中的概念
• 🏆 已掌握的概念

第四步：儲存 委託 complex-memory-manager 儲存總結

工作流2：解析外部內容（按需）

使用者分享內容時：文章URL、影片URL、圖片、或直接概念解釋

第一步：內容分析

• 文章 → web_fetch 提取關鍵概念、論據、資料
• 影片 → 如有文字稿則提取主旨、示例、結論
• 圖片 → 描述視覺內容，提取文字，找出關鍵概念
• 直接解釋 → 解析使用者的文字說明

第二步：概念結構化 每個概念建立結構化筆記，包括名稱、類別、來源、摘要、要點、關聯概念、實際應用

第三步：交叉連結 檢查已有概念，雙向連結；若已存在則合併/更新而非重複

工作流3：趨勢調研（定期/按需）

使用者說："最近有什麼技術熱點"

第一步：確定調研範圍 使用使用者指定關鍵詞或已有概念類別作為種子

第二步：搜尋收集 web_search 定向搜尋，優先來源：TechCrunch、ArsTechnica、arXiv、GitHub、Bloomberg、Reuters

第三步：趨勢分析 對每個趨勢記錄：標題、類別、重要性、描述、影響、來源、與現有概念的關係

第四步：學習與儲存 使用工作流2格式儲存新概念，更新趨勢檔案

工作流4：定期自審（Cron驅動）

預設每週執行：

1. 審查 memory/concepts/ 中的積累概念
2. 在有概念儲存的類別上執行趨勢調研
3. 生成包含新趨勢的合併總結
4. 識別知識盲區
5. 通過 self-iteration-engine 記錄迭代
6. 基於盲區提出下週學習主題

記憶結構

memory/
├── tier1-public/
│   ├── concepts-summary-YYYY-MM.md     # 月度概念概覽（公開）
│   └── trends-YYYY-MM-DD.md            # 趨勢調研結果（公開）
├── tier2-internal/
│   └── concepts-detail-YYYY-MM.md      # 詳細加密筆記（內部）
├── concepts/
│   ├── <概念slug>.md                    # 獨立概念檔案
│   └── INDEX.md                        # 概念總索引
└── usage-logs/
    └── knowledge-and-trends-engine.md  # 由self-iteration-engine管理

查詢示例

"最近我們聊過什麼來著？" → 工作流1（概念總結）
"看看這篇https://...  幫我提煉核心概念" → 工作流2（內容解析）
"最近AI領域有什麼新動向" → 工作流3（趨勢調研）
"定期總結" → 工作流4（定期自審）
"這個概念你記住" + 解釋 → 工作流2（直接儲存）

這個Skill功能設計豐富，可以幫你整理聊天中提到的概念、分析分享的文章或影片、追蹤最新技術動態，文件寫得清楚直觀。優點是流程清晰、使用場景明確。不足之處在於它目前更像一份設計文件，實際執行能力取決於它依賴的其他工具是否正常工作，具體效果如何可能還需要實際使用後才能判斷。