China Claw

Overview

Use this skill to research and explain Chinese products that can be described as OpenClaw-like alternatives or adjacent substitutes. Focus on factual positioning, what each company built, how it compares to OpenClaw, and where the match is strong, partial, or weak.

Workflow

1. 1. Clarify the user's goal when needed:

- wanting a market map - wanting a shortlist - wanting one best recommendation - wanting products by company size, use case, or capability

1. 2. Search the web for recent information about Chinese companies and products related to:

- AI agent platforms - workflow automation assistants - multi-agent systems - enterprise AI copilots - tool-using AI assistants

1. 3. Read multiple sources before concluding.
2. Group findings by company, not by random web page.
3. For each product, state whether it is:

- close substitute - partial substitute - adjacent product

1. 6. End with a recommendation or ranking if the user wants a decision.

Search Strategy

Prefer Chinese search queries and combine product, company, and capability terms.

Recommended query patterns:

• - INLINECODE0
• INLINECODE1
• INLINECODE2
• INLINECODE3
• INLINECODE4
• INLINECODE5
• INLINECODE6
• INLINECODE7

If you already know likely companies, verify them with current sources instead of relying on memory.

Comparison Rules

Compare each product along these axes:

• - 定位: personal assistant, enterprise copilot, workflow platform, agent builder, browser automation, knowledge assistant
• 目标用户: individual users, developers, operations teams, enterprises
• 核心能力: tool use, workflow orchestration, browser actions, knowledge retrieval, scheduling, integrations
• 开放性: configurable, extensible, developer-friendly, closed SaaS, internal-only
• 与 OpenClaw 的相似度: high / medium / low

Do not call something a “平替” just because it uses AI. The product should overlap with OpenClaw in at least part of the workflow: assistant orchestration, tools, agents, automation, messaging, memory, or operator-style execution.

Evidence Standards

• - Prefer official sites, product docs, launch posts, GitHub pages, or reputable media coverage.
• Avoid writing from hype or marketing copy alone.
• If a product appears discontinued, niche, invite-only, or poorly evidenced, say so.
• If information is thin, mark the confidence as low.
• Do not fabricate features.

Response Patterns

Pattern A: Market Overview

Use when the user asks broadly for “有哪些”.

For each company/product, provide:

• - 公司
• 产品
• 一句话定位
• 为什么像 OpenClaw
• 主要差异
• 相似度判断

Then add:

• - 推荐先关注的 3 个
• 哪类用户适合哪类产品

Pattern B: Shortlist Recommendation

Use when the user wants actionable recommendations.

Structure:

• - 用户需求假设
• 3-5 个最相关产品
• 每个产品的适配理由
• 明确推荐项
• 不推荐或暂不推荐的原因

Pattern C: Company-by-Company Scan

Use when the user names specific companies such as 阿里、字节、百度、腾讯、MiniMax、智谱、月之暗面、面壁等.

For each company:

• - 是否有相关产品
• 产品名称
• 与 OpenClaw 的重合点
• 更偏企业、开发者还是消费者
• 当前成熟度判断

Output Quality Bar

• - Be specific, not generic.
• Distinguish between “agent platform”, “copilot”, and “chatbot”.
• Prefer recent information when product categories move quickly.
• If the user asks for “各大公司”, include large companies first, then notable startups.
• If the evidence is weak, say “更像相关方向产品，不一定算直接平替”.

Good Framing Language

Preferred wording:

• - INLINECODE8
• INLINECODE9
• INLINECODE10
• INLINECODE11

Avoid overstated wording:

• - INLINECODE12
• INLINECODE13 unless a source explicitly positions it that way or the similarity is genuinely strong.