金石知识库

金石知识库管理技能。监控钉钉多维表格中的项目管理事项状态，当事项状态为已完成时自动归档并生成知识文档。

Security Scan

The skill's code, declared requirements, and runtime instructions are consistent with a DingTalk table -> local document archival workflow and do not ask for unrelated credentials or system access.

Purpose & Capability

Name/description, required binary 'mcporter', and the DINGTALK_MCP_URL environment variable align with a DingTalk multi-dimensional table integration. The code uses mcporter to call the dingtalk-ai-table provider and generates local markdown documents from records, which matches the stated purpose. Minor note: the Python code itself does not directly read DINGTALK_MCP_URL (the CLI/connector likely does), but this is plausible rather than suspicious.

Instruction Scope

SKILL.md and archive_checker.py only operate on the declared DingTalk base/table/view and on files inside the skill workspace (docs/ and archived-items.json). There are no instructions to read unrelated system files, exfiltrate data to external endpoints, or perform broad system reconnaissance. The script invokes 'mcporter' via subprocess to query records and writes generated docs/state to its workspace.

Install Mechanism

No install spec is provided (instruction-only with included script), so nothing is downloaded or extracted at install time. This is a low-risk install model; the only runtime dependency the skill enforces is the external 'mcporter' binary which must be installed separately by the operator.

Credentials

The skill requires a single env var DINGTALK_MCP_URL which is appropriate for a DingTalk connector. It does not request unrelated secrets or multiple credentials. Note: the Python code does not directly reference that env var — expected because the mcporter/dingtalk-ai-table integration likely reads it; operators should confirm how mcporter obtains credentials (env, config file, or other).

Persistence & Privilege

always:false and user-invocable:true (default) — no forced global presence. The skill stores its own state and generated docs under its workspace only and does not modify other skills or system-wide agent settings.

Assessment

This skill appears to do what it claims: query a DingTalk table and write local markdown docs. Before installing, confirm the following: (1) mcporter is a trusted binary in your environment and you understand how it gets configured (does it read DINGTALK_MCP_URL or other credentials from a config file?), (2) the workspace where docs and archived-items.json are written is an appropriate location (avoid running it in a directory that contains sensitive secrets), (3) inspect the dingtalk-ai-table/mcporter setup to see what network access and credentials they will use, and (4) run the script manually first (python3 skills/knowledge-base/archive_checker.py) as a non-privileged user to verify behavior and outputs. The code uses subprocess.run without shell=True and writes only to its own workspace, so there are no immediate red flags, but you should vet the mcporter/dingtalk connector configuration before granting it access to live systems.

项目知识库管理

功能

1. 监控项目管理事项：定期检查钉钉多维表格中的已完成事项
2. 自动归档：当事项状态为已完成时，自动生成知识文档
3. 知识文档生成：
   • 问题类 → 问题分析报告（含AI分析输出）
   • 需求类 → 需求说明文档
   • 任务类 → 任务总结
   • 事件类 → 事件报告

配置

环境变量

export DINGTALK_MCP_URL="<钉钉AI表格MCP服务器地址>"

钉钉表格配置

• Base: 金谷信托项目管理系统
• Base ID: QOG9lyrgJP3Oo757S9wn4AyZVzN67Mw4
• Table: 项目事项管理表
• Table ID: atLkTeV
• 监控视图: 上周已完成事项 (viewId: dGoortH)

使用方式

手动执行归档检查

python3 skills/knowledge-base/archive_checker.py

定时任务

系统会自动在每个工作日上午9点执行归档检查。

输出

生成的文档保存在 skills/knowledge-base/docs/ 目录下，文件命名格式：

{事项类型}_{标题}_{recordId}_{日期}.md

依赖技能

• dingtalk-ai-table: 钉钉多维表格操作