charlieviettq/awesome-agent-skill / .claude/skills/algo-net-community
"algo-net-community"
Verified"Implement Louvain community detection to discover densely connected groups in networks. Use this skill when the user needs to find communities or clusters in social/organizational networks, segment customers by interaction patterns, or analyze network modular structure — even if they say 'find groups in this network', 'community detection', or 'network clustering'.".
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 4.4 KB |
| examples/sample_scenario.md | markdown | 5.9 KB |
| references/leiden.md | markdown | 11.3 KB |
| references/multi-resolution.md | markdown | 11.7 KB |
examples/sample_scenario.md
View on GitHub# Example: SaaS 公司跨部門協作網絡分析
## Scenario
**公司:** Parchment Analytics(台北,150 人 SaaS 公司)
**情境:** 2026 年 Q1 組織重整後,HR 總監 Jenny Liu 想了解員工實際協作模式是否和組織架構圖吻合。她從 Slack workspace 匯出 2025-10-01 至 2025-12-31 三個月的訊息互動記錄(公開頻道訊息 + DM 次數統計,排除 bot)。
**用戶問題:**
> 「我有一份員工互動頻率表(CSV),想知道公司裡自然形成的『溝通圈』是什麼,看看和目前的 5 個部門分法是否一致。大約 130 名員工,3 個月資料。」
**資料格式(sample):**
```csv
source,target,messages
alice.chen,bob.wu,142
alice.chen,carol.lin,88
bob.wu,dave.ho,201
...
```
共 2,847 條邊,平均權重 47 訊息/季。
---
## Analysis
### Phase 1:輸入驗證 & 圖建構
**節點:** 130 名員工(排除離職者 8 人後剩 122 個有效節點)
**邊:** 2,847 條(僅保留雙向互動總次數 ≥ 10 的邊,過濾後剩 1,983 條)
**孤立節點:** 3 名(新進員工 < 30 天),標記但排除於社群偵測之外
**Gate ✓** — 圖已載入,平均 degree = 32.5,密度 0.27(中等密集,適合 Louvain)
**Resolution limit 檢查(Iron Law):**
- E = 1,983 條邊
- √(2E) ≈ √3,966 ≈ 63
- 結論:任何規模 < 63 人的社群理論上仍可偵測;但若存在規模 < 8-10 人的小圈子,可能被合併。→ 設 γ = 1.2 以抓較細粒度的社群。
### Phase 2:Louvain 執行
由於非確定性問題,執行 10 次取最高 modularity 的結果:
| 執行次數 | 社群數 | Modularity Q |
|---------|--------|-------------|
| Run 1 | 7 | 0.623 |
| Run 3 | 6 | 0.618 |
| Run 7 | 7 | **0.641** ← 選用 |
| Run 10 | 8 | 0.629 |
**最佳結果:7 個社群,Q = 0.641**
### Phase 3:驗證 & 人工核對
**Gate 1 — Q > 0 ✓**(0.641 屬於強模組化結構)
**社群大小分佈:**
| 社群 ID | 規模 | 內部密度 | 前 3 高互動成員 |
|--------|------|---------|--------------|
| C0 | 38 | 0.51 | alice.chen, bob.wu, erin.kao |
| C1 | 22 | 0.63 | frank.lin, grace.wu, henry.chang |
| C2 | 18 | 0.58 | iris.chen, jason.ho, karen.liu |
| C3 | 15 | 0.44 | leo.wang, mia.sun, nina.tsai |
| C4 | 12 | 0.72 | oscar.chen, penny.yeh, quinn.wu |
| C5 | 11 | 0.69 | ryan.lin, sara.ho, tina.chen |
| C6 | 6 | 0.81 | uma.wu, victor.lin, wendy.kao |
**Gate 2 — 大小分佈 ✓**(最大社群 38 人,無「一個超大 + 全部 singleton」的異常)
**與官方部門對比(混淆矩陣摘要):**
```
C0 C1 C2 C3 C4 C5 C6
Engineering(52) 31 5 4 8 2 1 1
Product(28) 4 18 2 1 0 2 1
Sales(20) 1 0 11 0 7 1 0
Data(15) 2 0 1 8 1 2 1
HR/Ops(25) 0 1 0 0 2 5 3 ← 分散在多社群
```
**關鍵發現:**
1. **C0 = Engineering + Data 混合**:Data team 的 8 人與 Engineering 核心高度混融,顯示 Data 實際上與 Eng 耦合,而非獨立運作
2. **Sales 分裂為 C2 + C4**:Sales team 有兩個截然不同的溝通圈,調查後發現對應「企業客戶組」與「中小企業組」,兩組主管有各自的 Slack 群組文化
3. **C6(6人高密度群)= 跨部門核心橋接者**:包含 HR 總監、CTO、CMO 及 3 位 PM,是組織橋接節點
4. **HR/Ops 分散**:HR 成員出現在 5 個不同社群,符合其「支援性」角色定位
### Phase 4:Resolution 敏感性檢查
| γ 值 | 社群數 | Q | 備注 |
|-----|-------|-------|-----|
| 0.8 | 4 | 0.587 | 過度合併,Sales 分裂消失 |
| 1.0 | 6 | 0.631 | 標準設定 |
| 1.2 | 7 | 0.641 | **選用**:揭露 Sales 內部分裂 |
| 1.5 | 11 | 0.608 | 過細,C6 被分解 |
γ = 1.2 是最佳解析度:Q 最高,且符合業務直覺。
---
## Result
```json
{
"communities": [
{"id": 0, "size": 38, "top_members": ["alice.chen", "bob.wu", "erin.kao"],
"internal_density": 0.51, "label": "Engineering-Data融合圈"},
{"id": 1, "size": 22, "top_members": ["frank.lin", "grace.wu", "henry.chang"],
"internal_density": 0.63, "label": "Product Core"},
{"id": 2, "size": 18, "top_members": ["iris.chen", "jason.ho", "karen.liu"],
"internal_density": 0.58, "label": "Sales企業客戶組"},
{"id": 3, "size": 15, "top_members": ["leo.wang", "mia.sun", "nina.tsai"],
"internal_density": 0.44, "label": "Engineering Platform子團隊"},
{"id": 4, "size": 12, "top_members": ["oscar.chen", "penny.yeh", "quinn.wu"],
"internal_density": 0.72, "label": "Sales中小企業組"},
{"id": 5, "size": 11, "top_members": ["ryan.lin", "sara.ho", "tina.chen"],
"internal_density": 0.69, "label": "HR-Ops核心"},
{"id": 6, "size": 6, "top_members": ["uma.wu", "victor.lin", "wendy.kao"],
"internal_density": 0.81, "label": "跨部門橋接者(高層)"}
],
"summary": {
"num_communities": 7,
"modularity": 0.641,
"largest": 38,
"smallest": 6,
"resolution_gamma": 1.2
},
"metadata": {
"algorithm": "louvain",
"runs": 10,
"nodes": 122,
"edges": 1983,
"period": "2025-10-01 to 2025-12-31"
}
}
```
**給 Jenny Liu 的建議:**
1. **組織架構與實際溝通有顯著落差**:偵測到 7 個通訊社群 vs 官方 5 個部門,主要差異在 Data team 實質上隸屬 Engineering 溝通圈,以及 Sales 已自然分成兩個次文化群。
2. **Q1 重整建議**:可考慮正式將 Data team 納入 Engineering 匯報線,或至少設置共同 PM;Sales 分組管理制度化。
3. **C6 橋接者需要保護**:這 6 人是組織的溝通樞紐,任何一人離職都會大幅降低跨部門資訊流通。
4. **限制**:本分析基於 Slack 訊息頻率,**不**代表工作品質或信任度,解讀社群意義需搭配 1:1 訪談確認。
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |