SuperAPI AI Directory
Partner中文AI newsFavorites
Skillslark-workflow-wiki-auditor

lark-workflow-wiki-auditor

Office & productivity

View repository

Updated 10-16 08:52

About this skill

Knowledge Base Health Audit Workflow

CRITICAL — Before starting, you MUST first use the Read tool to read lark-shared/SKILL.md, which contains authentication, permission handling, and security rules.

Applicable Scenarios

  • "Check the status of the knowledge base" / "Which documents are outdated"
  • "Knowledge base health report" / "Audit the knowledge base"
  • "Are there any documents that haven't been updated for a long time" / "Knowledge base structure analysis"
  • "Knowledge base optimization suggestions" / "Wiki audit"
  • "Clean up the knowledge base" / "Which documents may need updating"

Together with existing Skills, it forms the "Knowledge Management Trio":

  • knowledge-qa：Search Q&A — Use the knowledge base
  • doc-summarizer：Document Summary — Understand the knowledge base
  • wiki-auditor(this Skill):Health Audit — Maintain the knowledge base

Prerequisites

Supports user identity. Before execution, ensure authorization has been granted:

lark-cli auth login --domain wiki,drive

Workflow

指定知识空间 ──► wiki spaces list（列出可用空间）
                       │
                       ▼
              wiki nodes（遍历节点树）
                       │
                       ▼
              drive metas（获取文档元数据）
                       │
                       ▼
              AI 健康度分析
              ┌────────┼────────┐
              ▼        ▼        ▼
           过时检测  结构分析  覆盖评估
                       │
                       ▼
              健康度审计报告

Step 1: List knowledge spaces

Refer to lark-wiki/SKILL.md。

# 列出所有可见的知识空间
lark-cli wiki spaces list --format json

If there are multiple spaces, list them for the user to choose:

找到 3 个知识空间：
1. 产品团队知识库（space_id: xxx）
2. 技术文档库（space_id: yyy）
3. 公司公共知识库（space_id: zzz）

请选择要审计的知识空间（输入序号，或"全部"）

Step 2: Traverse knowledge nodes

# 获取空间根节点下的子节点
lark-cli wiki spaces nodes list --params '{"space_id":"<space_id>","page_size":50}' --format json

# 获取子节点的子节点（递归）
lark-cli wiki spaces nodes list --params '{"space_id":"<space_id>","parent_node_token":"<parent_token>","page_size":50}' --format json

For each node, record:

  • node_token: node token
  • obj_token: associated document token
  • obj_type: document type
  • title: title
  • has_child: whether it has child nodes
  • Node hierarchy depth (calculated based on recursion depth)

⚠️ Data volume control: A large knowledge base may have hundreds of nodes. It is recommended to limit the maximum traversal depth to 3-4 levels; when exceeded, prompt the user to narrow the scope.

Step 3: Obtain document metadata

Refer to lark-drive/SKILL.md。

# 批量获取文档元数据（每次最多 15 个）
lark-cli drive metas batch_query --data '{
  "request_docs": [
    {"doc_type": "docx", "doc_token": "<token1>"},
    {"doc_type": "docx", "doc_token": "<token2>"}
  ],
  "with_url": true
}' --format json

Extract from the metadata:

  • latest_modify_time: last modified time
  • create_time: creation time
  • owner: document owner
  • url: document link

Step 4: AI health analysis

4.1 Outdated document detection:

Outdated level Time since last update Mark
Fresh < 30 days 🟢
Normal 30-90 days 🟡
Stale 90-180 days 🟠
Outdated > 180 days 🔴

4.2 Structure analysis:

  • Hierarchy depth distribution (too deep > 4 levels is not conducive to discovery)
  • Isolated nodes (proportion of leaf nodes that have documents but no child nodes and no sibling nodes)
  • Empty nodes (nodes exist but the associated document is empty or has been deleted)
  • Directory balance (too many documents under a single directory > 20)

4.3 Coverage assessment:

  • Total number of documents
  • Distribution by document type (docx / sheet / bitable, etc.)
  • Distribution by creator (whether highly concentrated among a few people)

Step 5: Generate audit report

## 📚 知识库健康度报告

### 空间信息
- **知识空间**：{空间名称}
- **审计时间**：{当前日期}
- **总文档数**：{N} 篇
- **节点层级**：最深 {M} 层

### 🏥 健康度评分

| 维度 | 评分 | 说明 |
|------|------|------|
| 内容新鲜度 | {X}/100 | 基于文档更新时间分布 |
| 结构合理性 | {Y}/100 | 基于层级深度和目录均衡度 |
| 覆盖完整性 | {Z}/100 | 基于文档数量和类型分布 |
| **综合评分** | **{总分}/100** | **{优秀/良好/需改进/较差}** |

### 🔴 过时文档清单（需关注）
| # | 文档标题 | 最后更新 | 距今 | 拥有者 | 链接 |
|---|---------|---------|------|--------|------|
| 1 | XX 接口文档 v1.0 | 2025-06-15 | 292天 | 张三 | [查看]({链接}) |
| 2 | YY 项目设计方案 | 2025-08-20 | 226天 | 李四 | [查看]({链接}) |
| 3 | ZZ 操作手册 | 2025-09-10 | 205天 | 王五 | [查看]({链接}) |

### 📊 新鲜度分布

🟢 Fresh (<30 days) ████████░░ 35% ({N} articles) 🟡 Normal (30-90 days) ██████░░░░ 28% ({N} articles) 🟠 Stale (90-180 days) ████░░░░░░ 22% ({N} articles) 🔴 Outdated (>180 days) ███░░░░░░░ 15% ({N} articles)


### 🗂️ 结构分析
- **层级分布**：1层 {N}个 | 2层 {N}个 | 3层 {N}个 | 4层+ {N}个
- ⚠️ 「{目录名}」下有 {N} 篇文档，建议拆分子目录
- ⚠️ 发现 {N} 个空节点（可能是已删除的文档引用）
- ⚠️ 层级深度 > 4 层的分支有 {N} 个

### 👤 贡献者分布
| 作者 | 文档数 | 占比 |
|------|--------|------|
| 张三 | 25 | 35% |
| 李四 | 18 | 25% |
| 其他 | 28 | 40% |

### 💡 优化建议
1. **{建议 1}**：{针对过时文档的具体建议}
2. **{建议 2}**：{针对结构问题的具体建议}
3. **{建议 3}**：{针对贡献者集中度的建议}

Step 6: Automatically create remediation suggestion Wiki (optional, v1.0.7+)

When the user wants to write the audit results to the knowledge base:

# 在知识库中创建审计报告节点（v1.0.7+ shortcut）
lark-cli wiki +node-create --space-id "<space_id>" \
  --parent-node-token "<parent_token>" \
  --title "知识库审计报告 (<date>)"

Tip: The audit report can be directly embedded in the knowledge base, making it convenient for the team to view and track remediation progress.

Step 7: Automatically fix misplaced documents (optional, v1.0.10+)

When the audit finds that a document is placed in an unreasonable location, it can be directly moved to the correct directory:

# 将错位文档移动到正确目录（v1.0.10+ shortcut）
lark-cli wiki +move --space-id "<space_id>" \
  --node-token "<misplaced_node_token>" \
  --target-parent-token "<correct_parent_token>"

⚠️ Safety rule: Before moving a document, the change plan must be displayed and user confirmation obtained. Tip：wiki +move Supports asynchronous task polling, so long-running operations will not time out.

Fault tolerance mechanism

Exception scenario Handling method
No access permission to the knowledge space Prompt the user to join the space or contact the administrator
Number of nodes > 200 Limit the traversal depth to 2 levels and prompt the user to narrow the scope
Failed to retrieve document metadata Record the failed nodes and analyze the rest normally
No documents in the space Prompt "This knowledge space currently has no documents"
drive domain not authorized Unable to retrieve modification time; perform structural analysis only

Permission table

Step Command Required scope Required or not
List spaces wiki spaces list wiki:wiki:readonly ✅ Required
Traverse nodes wiki spaces nodes list wiki:wiki:readonly ✅ Required
Document metadata drive metas batch_query drive:drive:readonly Recommended
Create audit report wiki +node-create wiki:wiki Optional (when writing audit results)
Move misplaced documents wiki +move wiki:wiki Optional (v1.0.10+)

Reference

  • lark-shared — Authentication, permissions (required reading)
  • lark-wiki — Detailed usage for knowledge space traversal
  • lark-drive — Document metadata query