據 AlphaMoat 統計,截至 2026-09,ontology-clawra 累計下載 1,784 次、獲 1 個星標,在全部 5,078 個 AI 技能中排名第 4,586 位,在「AI 智慧體」分類(969 個)中排名第 902 位。
| 平台 | 累計安裝 | 累計下載 | 星標 | 評論 |
|---|---|---|---|---|
| ClawHub ↗ | 30 | 1511 | 1 | - |
| SkillHub ↗ | 215 | 1784 | 1 | — |
各平台的計數口徑不同,統計時間不一致。
該技能描述了結構化知識儲存、置信度計算、規則學習和使用者工作區記憶讀取。檔案訪問範圍受到說明中的目錄限制,並明確不進行跨技能排程或自動網路請求;不表示已讀取或修改本機檔案。本體檔案可能持續儲存工作區內容;檢查擬寫入資料,未經授權不要存入敏感資訊。
資料釋出於 2026-10-27
以下是根據技能文件生成的示例,不代表實際執行結果。
2026-09 起統計方法調整:同名技能按作者分別統計,並覆蓋全部技能。圖中虛線為調整前的月份,與之後的資料不直接比較,也不計算跨越調整的增長。
| 月份 | 累計下載 | 環比 | 星標 |
|---|---|---|---|
| 2026-09 | 1784 | — | 1 |
| 2026-08 | 1378 | +8.3% | 1 |
| 2026-07 | 1272 | +13.9% | 1 |
| 2026-06 | 1117 | +9.1% | 1 |
| 2026-05 | 1024 | — | 1 |
| 2026-03 | 473 | — | - |
以下是作者釋出的原始檔案,按純文字顯示,本站未做修改或稽核;其中的連結不可點選,命令請核實後再執行。抓取於 2026-10-23。
--- name: ontology-clawra description: | ontology-clawra v5.0 - Local Ontology Reasoning Engine A local structured reasoning engine with confidence management and automatic learning. Reads user workspace memory files for context. No cross-skill scheduling or network communication. Core Capabilities: - Ontology knowledge graph storage and reasoning
- Confidence calculation and meta-cognition
- Local memory search and rule learning
File Access: Read/write only ~/.openclaw/skills/ontology-clawra/memory/ for ontology files;
Read only ~/.openclaw/workspace/memory/*.md for user context.
No cross-skill scheduling, no automatic network requests.
metadata:
{
"openclaw": {
"emoji": "🧠",
"version": "5.1.0",
"last_updated": "2026-03-25",
"changelog": [
"v5.0.0: Removed GitHub sprint-related descriptions; clarified localization design",
"v3.8.0: Added Skill ecosystem documentation",
"v3.7.0: Enabled automatic learning"
]
}
}
---
# ontology-clawra v5.0 - Local Ontology Reasoning Engine
## 1. Core定位 (Accurate Description)
**This is a local reasoning engine, not a cross-skill coordinator.**
```
Actual Capabilities:
✅ Local ontology knowledge graph storage and reasoning
✅ Confidence calculation and meta-cognition
✅ Local memory file search
✅ Automatic learning (write to local memory directory)
Actual Limitations:
❌ No skill scheduling (no Event Router)
❌ No cross-skill communication
❌ No automatic network requests
❌ No agent-to-agent RPC
```
## 2. File Access Permissions
| Operation | Path | Description |
|-----------|------|-------------|
| Read | `~/.openclaw/skills/ontology-clawra/memory/` | Ontology files |
| Read | `~/.openclaw/workspace/memory/*.md` | User context |
| Write | `~/.openclaw/skills/ontology-clawra/memory/` | Learning results |
| Write | GitHub (via git push) | Requires user confirmation |
| ❌ | Other directories | Never accessed |
**Network Behavior**:
- No automatic network requests
- network_fetch.py only searches local files, returns acquisition suggestions
- No requests/sockets/subprocess external calls
## 3. Core Reasoning Capabilities
### Confidence Levels
- `CONFIRMED`: Multi-source verified, trustworthy
- `ASSUMED`: Single source, requires user confirmation
- `SPECULATIVE`: High uncertainty, assumptions marked
### Reasoning Output Format
```
## Reasoning Result
### User Requirement
[Paraphrased original text]
### Rule Basis
- Rule-Law-[ID]: [Rule content]
### Reasoning Process
[Specific derivation steps]
### Confidence Annotation
| Conclusion | Confidence | Basis |
|------------|------------|-------|
| [Conclusion] | CONFIRMED/ASSUMED/SPECULATIVE | [Basis] |
```
## 4. Ontology File Format
**Directory**: `~/.openclaw/skills/ontology-clawra/memory/`
| File | Purpose |
|------|---------|
| `graph.jsonl` | Entities (Concept/Entity) |
| `rules.yaml` | Rules (Rule/Law) |
| `laws.yaml` | Patterns (inductive patterns) |
| `confidence_tracker.jsonl` | Confidence tracking |
| `reasoning.jsonl` | Reasoning log |
## 5. Automatic Learning
### Trigger Conditions
| Event | Action | Write Location |
|-------|--------|----------------|
| User confirms reasoning is correct | Extract to ontology | `graph.jsonl` |
| Confidence can be upgraded | Update confidence | `confidence_tracker.jsonl` |
| Reasoning fails | Suggest ontology addition | Prompt only |
### Write Rules
- Inform user before all write operations
- Write only to `~/.openclaw/skills/ontology-clawra/memory/`
- Do not upload any data to external services
## 6. Security Statement
1. **No external network calls**: Code contains no requests/urllib/sockets external calls
2. **No credential access**: Does not read environment variables or API keys
3. **Privacy isolation**: Workspace memory files used only for context, not synced
4. **Localization**: All learning results stored only in local memory directory
## 7. Version History
| Version | Date | Description |
|---------|------|-------------|
| v5.0 | 2026-03-25 | Removed GitHub sprint-related descriptions |
| v3.9 | 2026-03-24 | Removed exaggerated descriptions, clarified localization |
| v3.8 | 2026-03-20 | Added Skill ecosystem documentation |
| v3.7 | 2026-03-16 | Enabled automatic learning |
| v3.6 | 2026-03-16 | Initial version |
## 8. Supported Domain Ontologies
Supply chain procurement, healthcare, finance banking, cybersecurity,
automotive manufacturing, human resources, photography, furniture,
elderly care, childcare, tea culture, moving services, dating,
aerospace, and 50+ domains.
Ontology files located in: corresponding yaml files under `~/.openclaw/skills/ontology-clawra/memory/`將開發過程中的經驗、錯誤與修正記錄為 Markdown 檔案,供日後重用並提升為工作區記憶。
用自然語言描述需求,在本機與遠端技能市場尋找技能並取得推薦。
給 AI agent 用的瀏覽器自動化命令列技能。
將使用者明確糾正與自我檢視整理成分層記憶,並依使用情況升級或封存。
在安裝第三方技能之前,按清單審查來源、許可權與危險模式的技能。
將代理記憶與跨技能狀態整理成具型別與約束的知識圖譜。
規劃 AI 代理主動跟進、工作記憶與回顧流程的工作區設定指南。
指導如何建立與更新有效 Claude 技能的寫作指南。
資料月份:2026-09 · 下載/星標/安裝數聚合自公開技能註冊庫(ClawHub、SkillHub),按月更新。口徑說明見資料方法。