Turn your domain expertise into a 24/7 AI consulting agent on OpenClaw. Clients interact in Feishu groups — your Agent answers accurately, safely, and around the clock.
将你的专业知识打造成 7×24 小时 AI 咨询 Agent。客户在飞书群中直接提问，Agent 基于你的知识库准确、安全地回答。
You have professional knowledge but limited time. Clients keep asking the same questions. You can't be online 24/7. This skill packages your expertise into an OpenClaw Agent that:
- Answers client questions based on YOUR knowledge, not hallucinations
- Operates 24/7 in Feishu groups without needing to be @mentioned
- Has safety constraints that prevent information leaks or identity exposure
- Can search for real-time information to supplement static knowledge
你有专业知识但时间有限，客户反复问同样的问题，你不可能 24 小时在线。这个 Skill 把你的知识打包成一个 OpenClaw Agent：
- 基于你的知识库回答，而非 AI 幻觉
- 在飞书群 7×24 小时运行，无需 @
- 内置安全约束，防止信息泄露和身份暴露
- 支持实时搜索补充静态知识
knowledge-agent/
├── SKILL.md # Entry point — 6-step interactive workflow / 入口 — 6步交互式工作流
├── references/
│ ├── knowledge-layers.md # Three-layer architecture / 三层知识架构
│ ├── safety-constraints.md # Security settings / 安全约束配置
│ ├── anti-hallucination.md # Quality controls / 反幻觉质量控制
│ ├── feishu-delivery.md # Feishu group setup / 飞书群配置
│ └── example-douyin.md # Complete example: Douyin consultant / 完整示例：抖音运营顾问
├── templates/
│ ├── AGENTS.template.md # AGENTS.md skeleton / AGENTS.md 骨架模板
│ ├── SOUL.template.md # Personality template / 人格与原则模板
│ ├── IDENTITY.template.md # Display name and image / 展示名称与形象
│ └── MEMORY.template.md # Long-term memory / 长期记忆结构
└── scripts/
└── setup-consultant.sh # Creates workspace / 创建工作区脚本
- Install the skill on your OpenClaw instance / 在你的 OpenClaw 实例上安装此 Skill
- Tell your main Agent: "Help me build a consulting Agent for [YOUR DOMAIN]" / 告诉主 Agent："帮我搭建一个 [你的领域] 咨询 Agent"
- Follow the 6-step interactive workflow — the Agent guides you through everything / 跟随 6 步交互式工作流，Agent 会引导你完成所有配置
- Share your knowledge materials when prompted / 按提示分享你的知识材料
- Review and confirm the generated configuration files / 审核并确认生成的配置文件
- Set up a Feishu group, invite clients, start serving / 创建飞书群，邀请客户，开始服务
- Independent workspace per agent / 每个 Agent 独立工作区 — never share workspace with other agents / 不要与其他 Agent 共享，否则会读到相同的 AGENTS.md
- Deterministic loading > hoping Agent reads it / 确定性加载 > 期望 Agent 自己去读 — critical knowledge goes in AGENTS.md / 关键知识放 AGENTS.md
- Safety constraints in AGENTS.md, not MEMORY.md / 安全约束放 AGENTS.md 而非 MEMORY.md — because MEMORY.md doesn't load in group chats / 因为 MEMORY.md 在群聊中不加载
- Search capability + knowledge base / 搜索能力 + 知识库 — Agent can verify with real-time search / Agent 可用实时搜索验证信息
- Domain-agnostic / 领域无关 — works for any field with a structured knowledge base / 适用于任何有结构化知识库的领域
- Professionals with domain expertise who want to scale their consulting / 想要规模化咨询服务的专业人士
- Anyone running paid knowledge services through Feishu groups / 在飞书群运营知识付费服务的人
- OpenClaw users who want to create a client-facing Agent with proper safety / 想创建面向客户的安全 Agent 的 OpenClaw 用户
@simonlin1212 — Based on production experience running paid consulting Agents (2026-04) / 基于实际运营付费咨询 Agent 的生产经验