← 返回 Skills 市场
122
总下载
0
收藏
1
当前安装
2
版本数
在 OpenClaw 中安装
/install knowledge-workflow
功能描述
知识管理工作流 - 完整的知识管理工作流:收集→打标→存储→发芽→产出。支持飞书/微信读书/URL,5 种发芽类型(灵光/心智模型/跨界/微习惯/潜意识)。
使用说明 (SKILL.md)
knowledge-workflow: 知识管理工作流 🌱
完整的知识管理工作流 - 收集→打标→存储→发芽→产出
版本: v2.0.0
最后更新: 2026-04-26
📋 功能描述
帮助用户系统化执行知识管理工作流。从各种来源收集知识,自动打标、存储、发芽,最终产出可发布的内容。
适用场景:
- 个人知识管理(飞书/微信读书/网页/文本)
- 团队知识沉淀(会议记录/项目文档)
- 自媒体内容生产(公众号文章/周报/月报)
- 知识发芽(灵光闪现/心智模型/跨界视角/微习惯/潜意识调整)
边界条件:
- 不替代深度思考(明确 AI 辅助边界)
- 发芽必须是高质量信息,用户可选择是否触发
- 需配合人工标注意义标签
🔄 核心工作流
| 步骤 | 功能 | 说明 | 输出 |
|---|---|---|---|
| 1 | collect(收集) | 从飞书/微信读书/URL/文本收集知识 | Markdown 笔记 |
| 2 | tag(打标) | 自动打标(主题 + 场景 + 行动) | 带标签笔记 |
| 3 | store(存储) | 存储到知识库,自动建立双链连接 | 存储路径 + 双链 |
| 4 | evolve(发芽) | 5 种发芽类型(用户选择触发) | 发芽内容 |
| 5 | output(产出) | 生成公众号文章/周报/月报 | 可发布内容 |
🌱 发芽功能(5 种类型)
| 发芽类型 | 说明 | 触发方式 | 输出 |
|---|---|---|---|
| spark | 灵光闪现 | 用户选择 | 核心洞察/洞察链条/跨界联想/问题启发/概念提炼 |
| model | 心智模型解读 | 用户选择 | 对应心智模型/模型对比/启发 |
| cross | 跨界视角 | 用户选择 | 跨领域视角/跨时空视角/跨界洞察 |
| habit | 微习惯 | 用户选择 | 可执行微习惯/习惯追踪/习惯养成建议 |
| subconscious | 潜意识调整 | 用户选择 | 潜意识模式/调整策略/自我反思问题 |
质量要求:
- 发芽必须是高质量信息
- 如果发芽质量不高,宁可不发芽
- 用户可选择是否保留发芽内容
- 发芽后标注质量等级(高/中/低)
📁 文件结构
knowledge-workflow/
├── SKILL.md # 技能文档(本文档)
├── main.py # 主程序
├── config.yaml # 配置文件
├── clawhub.yaml # ClawHub 发布配置
├── requirements.txt # Python 依赖
└── subfunctions/ # 子功能模块
├── collect.py # 收集功能
├── tag.py # 打标功能
├── store.py # 存储功能
├── evolve.py # 知识发芽(5 种类型)
└── output.py # 产出功能
🔧 使用示例
方式 1:一键调用(完整工作流)
# 处理飞书文档
python main.py run feishu PFAvdKEILouK29xCgNuc5b1bnnK
# 处理微信读书导出
python main.py run wechat "[微信读书导出文本]"
# 处理 URL
python main.py run url https://example.com/article
方式 2:分步调用
# 步骤 1: 收集
python main.py collect feishu PFAvdKEILouK29xCgNuc5b1bnnK
# 步骤 2: 打标
python main.py tag note-20260414160000
# 步骤 3: 知识发芽(用户选择触发)
python main.py evolve note-20260414160000 spark
# 步骤 4: 产出文章
python main.py output spark-20260414160000 article
⚠️ 注意事项
必须遵守:
- ✅ 发芽必须是高质量信息
- ✅ 用户选择触发发芽(不是自动)
- ✅ 发芽后标注质量等级(高/中/低)
- ❌ 不要自动发芽(必须用户选择)
- ❌ 不要低质量发芽(宁可不发芽)
- ❌ 不要只收集,不发芽
模糊请求处理:
如果用户请求模糊(如"帮我管理一下知识"):
→ 列出 5 个核心步骤供选择
→ 示例:"我可以帮你:1.收集 2.打标 3.存储 4.发芽 5.产出。你想做哪个?"
📊 成功指标
| 指标 | 目标值 | 说明 |
|---|---|---|
| 每日收集 | 3 件/天 | 飞书/微信读书/网页/文本 |
| 发芽率 | >50% | 收集后触发发芽的比例 |
| 产出率 | >30% | 发芽后产出文章的比例 |
| 质量等级 | 高 | 发芽内容质量等级为高 |
🔗 相关技能
context-manager- 个人上下文管理(前置技能)note-tagger- 笔记打标experience-memory-tracker- 体验记忆追踪
推荐组合:
context-manager → knowledge-workflow
(上下文管理) (知识生产)
维护者:燃冰 & ant
版本:v2.0.0
创建日期:2026-04-14
最后更新:2026-04-26
发布状态:待发布
安全使用建议
What to consider before installing or running this skill:
- The published bundle is incomplete: SKILL.md promises a full pipeline (main.py, collect/tag/store/output, Feishu/WeChat integration) but only evolve.py is included. Do not assume the integrations exist.
- As-is, the included code only implements the 'evolve' step: it searches your ~/kb for <note_id>.md and writes generated Markdown into ~/kb/outputs/sparks. If you run it, it will read local notes and create files in your home directory — back up sensitive notes first.
- The documentation shows examples using Feishu tokens and other remote sources, but no code here handles authentication or network integration. If you need those features, ask the maintainer for the missing modules or the repository URL and inspect the complete source before use.
- Before running any code: review the rest of the repository (main.py, collect/tag/store/output modules, and any code implementing _generate_* helpers) to confirm there are no unexpected network calls or secret exfiltration. Pay attention to any code that sends content to external endpoints or reads unrelated config files.
- If you must test now, run in a sandboxed environment or with a dedicated user account and non-sensitive sample notes, not your primary home directory.
If you can get the complete source (the referenced GitHub repo) and it matches the documentation, re-run this evaluation. If the author cannot provide the missing modules, treat the skill as only a local 'evolve' utility and not the advertised full workflow.
功能分析
Type: OpenClaw Skill
Name: knowledge-workflow
Version: 2.0.1
The skill bundle implements a structured knowledge management workflow for collecting, tagging, and 'sprouting' ideas from various sources like Feishu and WeChat Reading. The core logic in `subfunctions/evolve.py` focuses on reading local markdown files and generating new insights based on their content, with file operations restricted to a user-defined knowledge base directory (defaulting to `~/kb`). No evidence of data exfiltration, malicious execution, or prompt injection was found.
能力评估
Purpose & Capability
SKILL.md and clawhub.yaml describe a complete knowledge-management pipeline (main.py, collect.py, tag.py, store.py, output.py, config, requirements, Feishu/WeChat/URL integrations). The published package contains only subfunctions/evolve.py plus metadata and the README. That mismatch means most claimed capabilities (collection from Feishu/WeChat, tagging, storage, CLI entrypoint) are not present in the code bundle — this is an incoherence between stated purpose and actual capability.
Instruction Scope
Runtime instructions and usage examples reference running python main.py and passing tokens/URLs. No main.py or collect/tag/store/output modules are present. The included evolve.py reads and writes files under a user home path (default ~/kb) and generates Markdown outputs; that behavior is coherent with 'evolve' but the SKILL.md's broader runtime flow (fetching remote docs, handling Feishu/WeChat auth) is unsupported by the shipped code. The SKILL.md also instructs 'user choose to trigger evolve' which is policy-level, but there is no enforcement mechanism in the provided files.
Install Mechanism
There is no install spec (instruction-only), which lowers supply-chain risk. However, the package is incomplete relative to its documentation. Because there is a Python file included, an agent or user could execute it locally; that file performs file system reads and writes within the user's home directory. No remote downloads, no obscure installers, and no archive extraction are present in the package.
Credentials
The skill declares no required environment variables or credentials, yet the SKILL.md shows example calls that include Feishu tokens and mentions integrations that would normally require secrets. The absence of declared credential requirements is inconsistent with advertised integrations. The included evolve.py accesses the user's home directory (~/kb) for notes and writes outputs there — that is reasonable for a local knowledge tool but users should be aware their local notes would be read and new files created.
Persistence & Privilege
always:false and normal autonomous invocation settings. The code creates directories and writes Markdown files under ~/kb/outputs/sparks; it does not request elevated system privileges or modify other skills. Writing to a user's home directory is expected for a note-processing tool but still represents persistence of generated content on disk and potential exposure of local notes if misused.
如何使用
- 确保已安装 OpenClaw(本地或 Docker 部署)
- 在对话框中输入安装命令:
/install knowledge-workflow - 安装完成后,直接呼叫该 Skill 的名称或使用
/knowledge-workflow触发 - 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v2.0.1
完善 5 种发芽类型(spark/model/cross/habit/subconscious),全部实现真实可用的 AI 生成逻辑
v2.0.0
基于 Karpathy LLM Wiki 的完整自进化系统 - 新增规则提炼、信念更新、自我修复功能
元数据
常见问题
knowledge-workflow 是什么?
知识管理工作流 - 完整的知识管理工作流:收集→打标→存储→发芽→产出。支持飞书/微信读书/URL,5 种发芽类型(灵光/心智模型/跨界/微习惯/潜意识)。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 122 次。
如何安装 knowledge-workflow?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install knowledge-workflow」即可一键安装,无需额外配置。
knowledge-workflow 是免费的吗?
是的,knowledge-workflow 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。
knowledge-workflow 支持哪些平台?
knowledge-workflow 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。
谁开发了 knowledge-workflow?
由 lj22503(@lj22503)开发并维护,当前版本 v2.0.1。
推荐 Skills